perltidy - a perl script indenter and reformatter
perltidy [ options ] file1 file2 file3 ...
(output goes to file1.tdy, file2.tdy, file3.tdy, ...)
perltidy [ options ] file1 -o outfile
perltidy [ options ] file1 -st >outfile
perltidy [ options ] <infile >outfilePerltidy reads a perl script and writes an indented, reformatted script. The formatting process involves converting the script into a string of tokens, removing any non-essential whitespace, and then rewriting the string of tokens with whitespace using whatever rules are specified, or defaults. This happens in a series of operations which can be controlled with the parameters described in this document.
Perltidy is a commandline frontend to the module Perl::Tidy. For documentation describing how to call the Perl::Tidy module from other applications see the separate documentation for Perl::Tidy. It is the file Perl::Tidy.pod in the source distribution.
Many users will find enough information in "EXAMPLES" to get started. New users may benefit from the short tutorial which can be found at http://perltidy.sourceforge.net/tutorial.html
A convenient aid to systematically defining a set of style parameters can be found at http://perltidy.sourceforge.net/stylekey.html
Perltidy can produce output on either of two modes, depending on the existence of an -html flag. Without this flag, the output is passed through a formatter. The default formatting tries to follow the recommendations in perlstyle(1), but it can be controlled in detail with numerous input parameters, which are described in "FORMATTING OPTIONS".
When the -html flag is given, the output is passed through an HTML formatter which is described in "HTML OPTIONS".
Here are some example perltidy commands followed by their meanings:
perltidy somefile.plThis will produce a file somefile.pl.tdy containing the script reformatted using the default options, which approximate the style suggested in perlstyle(1). The source file somefile.pl is unchanged.
perltidy *.plExecute perltidy on all .pl files in the current directory with the default options. The output will be in files with an appended .tdy extension. For any file with an error, there will be a file with extension .ERR.
perltidy -b file1.pl file2.plModify file1.pl and file2.pl in place, and backup the originals to file1.pl.bak and file2.pl.bak. If file1.pl.bak and/or file2.pl.bak already exist, they will be overwritten.
perltidy -b -bext='/' file1.pl file2.plSame as the previous example except that the backup files file1.pl.bak and file2.pl.bak will be deleted if there are no errors.
perltidy -gnu somefile.plExecute perltidy on file somefile.pl with a style which approximates the GNU Coding Standards for C programs. The output will be somefile.pl.tdy.
perltidy -i=3 somefile.plExecute perltidy on file somefile.pl, with 3 columns for each level of indentation (-i=3) instead of the default 4 columns. There will not be any tabs in the reformatted script, except for any which already exist in comments, pod documents, quotes, and here documents. Output will be somefile.pl.tdy.
perltidy -i=3 -et=8 somefile.plSame as the previous example, except that leading whitespace will be entabbed with one tab character per 8 spaces.
perltidy -ce -l=72 somefile.plExecute perltidy on file somefile.pl with all defaults except use "cuddled elses" (-ce) and a maximum line length of 72 columns (-l=72) instead of the default 80 columns.
perltidy -g somefile.plExecute perltidy on file somefile.pl and save a log file somefile.pl.LOG which shows the nesting of braces, parentheses, and square brackets at the start of every line.
perltidy -dbs -dbl=10 somefile.pl >blocks.csvThis will dump a table of comma-separated metrics for subroutines longer than 10 lines to blocks.csv.
perltidy -duv somefile.pl >vars.txtThis will dump a list of unused and reused lexical variable names to vars.txt.
perltidy -html somefile.plThis will produce a file somefile.pl.html containing the script with html markup. The output file will contain an embedded style sheet in the <HEAD> section which may be edited to change the appearance.
perltidy -html -css=mystyle.css somefile.plThis will produce a file somefile.pl.html containing the script with html markup. This output file will contain a link to a separate style sheet file mystyle.css. If the file mystyle.css does not exist, it will be created. If it exists, it will not be overwritten.
perltidy -html -pre somefile.plWrite an html snippet with only the PRE section to somefile.pl.html. This is useful when code snippets are being formatted for inclusion in a larger web page. No style sheet will be written in this case.
perltidy -html -ss >mystyle.cssWrite a style sheet to mystyle.css and exit.
perltidy -html -frm mymodule.pmWrite html with a frame holding a table of contents and the source code. The output files will be mymodule.pm.html (the frame), mymodule.pm.toc.html (the table of contents), and mymodule.pm.src.html (the source code).
The entire command line is scanned for options, and they are processed before any files are processed. As a result, it does not matter whether flags are before or after any filenames. However, the relative order of parameters is important, with later parameters overriding the values of earlier parameters.
For each parameter, there is a long name and a short name. The short names are convenient for keyboard input, while the long names are self-documenting and therefore useful in scripts. It is customary to use two leading dashes for long names, but one may be used.
Most parameters which serve as on/off flags can be negated with a leading "n" (for the short name) or a leading "no" or "no-" (for the long name). For example, the flag to outdent long quotes is -olq or --outdent-long-quotes. The flag to skip this is -nolq or --nooutdent-long-quotes or --no-outdent-long-quotes.
Options may not be bundled together. In other words, options -q and -g may NOT be entered as -qg.
Option names may be terminated early as long as they are uniquely identified. For example, instead of --dump-token-types, it would be sufficient to enter --dump-tok, or even --dump-t, to uniquely identify this command.
The following parameters concern the files which are read and written.
Show summary of usage and exit.
Show perltidy VERSION number and exit.
Name of the output file (only if a single input file is being processed). If no output file is specified, and output is not redirected to the standard output (see -st), the output will go to filename.tdy. [Note: - does not redirect to standard output. Use -st instead.]
Perltidy must be able to operate on an arbitrarily large number of files in a single run, with each output being directed to a different output file. Obviously this would conflict with outputting to the single standard output device, so a special flag, -st, is required to request outputting to the standard output. For example,
perltidy somefile.pl -st >somefile.new.plThis option may only be used if there is just a single input file. The default is -nst or --nostandard-output.
If perltidy detects an error when processing file somefile.pl, its default behavior is to write error messages to file somefile.pl.ERR. Use -se to cause all error messages to be sent to the standard error output stream instead. This directive may be negated with -nse. Thus, you may place -se in a .perltidyrc and override it when desired with -nse on the command line.
Change the extension of the output file to be ext instead of the default tdy (or html in case the --html option is used). See "Specifying File Extensions".
When perltidy creates a filename for an output file, by default it merely appends an extension to the path and basename of the input file. This parameter causes the path to be changed to path instead.
The path should end in a valid path separator character, but perltidy will try to add one if it is missing.
For example
perltidy somefile.pl -opath=/tmp/will produce /tmp/somefile.pl.tdy. Otherwise, somefile.pl.tdy will appear in whatever directory contains somefile.pl.
If the path contains spaces, it should be placed in quotes.
This parameter will be ignored if output is being directed to standard output, or if it is being specified explicitly with the -outfile=s parameter.
Modify the input file or files in-place and save the original with the extension .bak. Any existing .bak file will be deleted. See next item for changing the default backup extension, and for eliminating the backup file altogether.
Please Note: Writing back to the input file increases the risk of data loss or corruption in the event of a software or hardware malfunction. Before using the -b parameter please be sure to have backups and verify that it works correctly in your environment and operating system.
A -b flag will be ignored if input is from standard input or goes to standard output, or if the -html flag is set.
In particular, if you want to use both the -b flag and the --perl-best-practices (-pbp) flag, then you must put a -nostandard-output (-nst) flag after the -pbp flag because it contains a --standard-output flag as one of its components, which means that output will go to the standard output stream.
This parameter serves two purposes: (1) to change the extension of the backup file to be something other than the default .bak, and (2) to indicate that no backup file should be saved.
To change the default extension to something other than .bak see "Specifying File Extensions".
A backup file of the source is always written, but you can request that it be deleted at the end of processing if there were no errors. This is risky unless the source code is being maintained with a source code control system.
To indicate that the backup should be deleted include one forward slash, /, in the extension. If any text remains after the slash is removed it will be used to define the backup file extension (which is always created and only deleted if there were no errors).
Here are some examples:
Parameter Extension Backup File Treatment
<-bext=bak> F<.bak> Keep (same as the default behavior)
<-bext='/'> F<.bak> Delete if no errors
<-bext='/backup'> F<.backup> Delete if no errors
<-bext='original/'> F<.original> Delete if no errorsThis parameter should not normally be used but is available in the event that problems arise as a transition is made from an older implementation of the backup logic to a newer implementation. The newer implementation is the default and is specified with -bm='copy'. The older implementation is specified with -bm='move'. The difference is that the older implementation made the backup by moving the input file to the backup file, and the newer implementation makes the backup by copying the input file. The newer implementation preserves the file system inode value. This may avoid problems with other software running simultaneously. This change was made as part of issue git #103 at github.
Setting -w causes any non-critical warning messages to be reported as errors. These include messages about possible pod problems, possibly bad starting indentation level, and cautions about indirect object usage. The default, -nw or --nowarning-output, is not to include these warnings.
Deactivate error messages (for running under an editor).
For example, if you use a vi-style editor, such as vim, you may execute perltidy as a filter from within the editor using something like
:n1,n2!perltidy -qwhere n1,n2 represents the selected text. Without the -q flag, any error message may mess up your screen, so be prepared to use your "undo" key.
Save the .LOG file, which has many useful diagnostics. Perltidy always creates a .LOG file, but by default it is deleted unless a program bug is suspected. Setting the -log flag forces the log file to be saved.
Set maximum interval between input code lines in the logfile. This purpose of this flag is to assist in debugging nesting errors. The value of n is optional. If you set the flag -g without the value of n, it will be taken to be 1, meaning that every line will be written to the log file. This can be helpful if you are looking for a brace, paren, or bracket nesting error.
Setting -g also causes the logfile to be saved, so it is not necessary to also include -log.
If no -g flag is given, a value of 50 will be used, meaning that at least every 50th line will be recorded in the logfile. This helps prevent excessively long log files.
Setting a negative value of n is the same as not setting -g at all.
Ignore any .perltidyrc command file. Normally, perltidy looks first in your current directory for a .perltidyrc file of parameters. (The format is described below). If it finds one, it applies those options to the initial default values, and then it applies any that have been defined on the command line. If no .perltidyrc file is found, it looks for one in your home directory.
If you set the -npro flag, perltidy will not look for this file.
To simplify testing and switching .perltidyrc files, this command may be used to specify a configuration file which will override the default name of .perltidyrc. There must not be a space on either side of the '=' sign. For example, the line
perltidy -pro=testcfgwould cause file testcfg to be used instead of the default .perltidyrc.
A pathname begins with three dots, e.g. ".../.perltidyrc", indicates that the file should be searched for starting in the current directory and working upwards. This makes it easier to have multiple projects each with their own .perltidyrc in their root directories.
Write a list of all options used to the .LOG file. Please see --dump-options for a simpler way to do this.
Force perltidy to process binary files. To avoid producing excessive error messages, perltidy skips files identified by the system as non-text. However, valid perl scripts containing binary data may sometimes be identified as non-text, and this flag forces perltidy to process them.
This flag asserts that the input and output code streams are identical, or in other words that the input code is already 'tidy' according to the formatting parameters. If this is not the case, an error message noting this is produced. This error message will cause the process to return a non-zero exit code. The test for this is made by comparing an MD5 hash value for the input and output code streams. This flag has no other effect on the functioning of perltidy. This might be useful for certain code maintenance operations. Note: you will not see this message if you have error messages turned off with the -quiet flag.
This flag asserts that the input and output code streams are different, or in other words that the input code is 'untidy' according to the formatting parameters. If this is not the case, an error message noting this is produced. This flag has no other effect on the functioning of perltidy.
When the standard input supplies the input stream, and the input has not been received within n seconds, perltidy will end with a timeout message. The intention is to catch a situation where perltidy is accidentally invoked without a file to process and therefore waits for input from the system standard input (stdin), which never arrives. The default is n=10 seconds. This check can be turned off with n=0.
If --format=tidy (the default) then perltidy will reformat the input file, and if --format=html then perltidy will produce html output.
For convenience, the abbreviation --tidy is equivalent to --format=tidy, and --html is equivalent to --format=html.
This flag disables all formatting and causes the input to be copied unchanged to the output except for possible changes in line ending characters and any pre- and post-filters. This can be useful in conjunction with a hierarchical set of .perltidyrc files to avoid unwanted code tidying. See also "Skipping Selected Sections of Code" for a way to avoid tidying specific sections of code.
Use n columns per indentation level (default n=4).
The default maximum line length is n=80 characters. Perltidy will try to find line break points to keep lines below this length. However, long quotes and side comments may cause lines to exceed this length. And long lines may sometimes be broken at a length less than n characters because some of the line break decisions employ small tolerances to prevent formatting instabilities.
The default length of 80 comes from the past when this was the standard CRT screen width. Many programmers prefer to increase this to something like 120.
Setting -l=0 is equivalent to setting -l=(a very large number). But this is not recommended because, for example, a very long list will be formatted in a single long line.
A problem arises using a fixed maximum line length with very deeply nested code and data structures because eventually the amount of leading whitespace used for indicating indentation takes up most or all of the available line width, leaving little or no space for the actual code or data. One solution is to use a very long line length. Another solution is to use the -vmll flag, which basically tells perltidy to ignore leading whitespace when measuring the line length.
To be precise, when the -vmll parameter is set, the maximum line length of a line of code will be M+L*I, where
M is the value of --maximum-line-length=M (-l=M), default 80,
I is the value of --indent-columns=I (-i=I), default 4,
L is the indentation level of the line of codeWhen this flag is set, the choice of breakpoints for a block of code should be essentially independent of its nesting depth. However, the absolute line lengths, including leading whitespace, can still be arbitrarily large. This problem can be avoided by including the next parameter.
The default is not to do this (-nvmll).
This flag also addresses problems with very deeply nested code and data structures. When the nesting depth exceeds the value n the leading whitespace will be reduced and start at a depth of 1 again. The result is that blocks of code will shift back to the left rather than moving arbitrarily far to the right. This occurs cyclically to any depth.
For example if one level of indentation equals 4 spaces (-i=4, the default), and one uses -wc=15, then if the leading whitespace on a line exceeds about 4*15=60 spaces it will be reduced back to 4*1=4 spaces and continue increasing from there. If the whitespace never exceeds this limit the formatting remains unchanged.
The combination of --variable-maximum-line-length and --whitespace-cycle=n provides a solution to the problem of displaying arbitrarily deep data structures and code in a finite window, although --whitespace-cycle=n may of course be used without --variable-maximum-line-length.
The default is not to use this, which can also be indicated using -wc=0.
Using tab characters will almost certainly lead to future portability and maintenance problems, so the default and recommendation is not to use them. For those who prefer tabs, however, there are two different options.
Except for possibly introducing tab indentation characters, as outlined below, perltidy does not introduce any tab characters into your file, and it removes any tabs from the code (unless requested not to do so with --freeze-whitespace). If you have any tabs in your comments, quotes, or here-documents, they will remain.
This flag causes each n leading space characters produced by the formatting process to be replaced by one tab character. The formatting process itself works with space characters. The -et=n parameter is applied as a last step, after formatting is complete, to convert leading spaces into tabs. Before starting to use tabs, it is essential to first get the indentation controls set as desired without tabs, particularly the two parameters --indent-columns=n (or -i=n) and --continuation-indentation=n (or -ci=n).
The value of the integer n can be any value but can be coordinated with the number of spaces used for indentation. For example, -et=4 -ci=4 -i=4 will produce one tab for each indentation level and and one for each continuation indentation level. You may want to coordinate the value of n with what your display software assumes for the spacing of a tab.
The default is not to use this, which can also be indicated using -et=0.
This flag causes one leading tab character to be inserted for each level of indentation. Certain other features are incompatible with this option, and if these options are also given, then a warning message will be issued and this flag will be unset. One example is the --line-up-parentheses option. This flag is retained for backwards compatibility, but if you use tabs, the --entab-leading-whitespace=n flag is recommended. If both --tabs and --entab-leading-whitespace=n are set, then --entab-leading-whitespace=n is used.
If the first line of code passed to perltidy contains leading tabs but no tab scheme is specified for the output stream then perltidy must guess how many spaces correspond to each leading tab. This number of spaces n corresponding to each leading tab of the input stream may be specified with -dt=n. The default is n=8.
This flag has no effect if a tab scheme is specified for the output stream, because then the input stream is assumed to use the same tab scheme and indentation spaces as for the output stream (any other assumption would lead to unstable editing).
This flag is used to deactivate all whitespace and line break changes within non-blank lines of code. When it is in effect, the only change to the script will be to the indentation and to the number of blank lines. And any flags controlling whitespace and newlines will be ignored. You might want to use this if you are perfectly happy with your whitespace and line breaks, and merely want perltidy to handle the indentation. (This also speeds up perltidy by well over a factor of two, so it might be useful when perltidy is merely being used to help find a brace error in a large script).
Setting this flag is equivalent to setting --freeze-newlines and --freeze-whitespace.
If you also want to keep your existing blank lines exactly as they are, you can add --freeze-blank-lines.
With this option perltidy is still free to modify the indenting (and outdenting) of code and comments as it normally would. If you also want to prevent long comment lines from being outdented, you can add either --nooutdent-long-lines (-noll) or --maximum-line-length=0 (l=0).
Setting this flag will prevent perltidy from doing any special operations on closing side comments. You may still delete all side comments however when this flag is in effect.
This flag indicates if the input data stream uses a character encoding. Perltidy does not look for the encoding directives in the source stream, such as use utf8, and instead relies on this flag to determine the encoding. (This is because perltidy often works on snippets of code rather than complete files, so it cannot rely on use utf8 directives). Consequently perltidy is likely to encounter problems formatting a file which is only partially encoded.
The possible values for s are:
-enc=none if no encoding is used, or
-enc=utf8 for encoding in utf8
-enc=guess if perltidy should guess between these two possibilities.The value none causes the stream to be processed without special encoding assumptions. This is appropriate for files which are written in single-byte character encodings such as latin-1.
The value utf8 causes the stream to be read and written as UTF-8. If the input stream cannot be decoded with this encoding then processing is not done.
The value guess tells perltidy to guess between either utf8 encoding or no encoding (meaning one character per byte). The guess option uses the Encode::Guess module which has been found to be reliable at detecting if a file is encoded in utf8 or not.
The current default is guess.
The abbreviations -utf8 or -UTF8 are equivalent to -enc=utf8, and the abbreviation -guess is equivalent to -enc=guess. So to process a file named file.pl which is encoded in UTF-8 you can use:
perltidy -utf8 file.plor
perltidy -guess file.plor simply
perltidy file.plsince -guess is the default.
To process files with an encoding other than UTF-8, it would be necessary to write a short program which calls the Perl::Tidy module with some pre- and post-processing to handle decoding and encoding.
This flag was added to resolve an issue involving the interface between Perl::Tidy and calling programs, and in particular Code::TidyAll (tidyall).
If you only run the perltidy binary this flag has no effect. If you run a program which calls the Perl::Tidy module and receives a string in return, then the meaning of the flag is as follows:
The setting -eos means Perl::Tidy should encode any string which it decodes. This is the default because it makes perltidy behave well as a filter, and is the correct setting for most programs.
The setting -neos means that a string should remain decoded if it was decoded by Perl::Tidy. This is only appropriate if the calling program will handle any needed encoding before outputting the string.
The default was changed from -neos to -eos in versions after 20220217. If this change causes a program to start running incorrectly on encoded files, an emergency fix might be to set -neos. Additional information can be found in the man pages for the Perl::Tidy module and also in https://github.com/perltidy/perltidy/blob/master/docs/eos_flag.md.
This flag controls whether or not perltidy may use module Unicode::GCString to obtain accurate display widths of wide characters. The default is --nouse-unicode-gcstring.
If this flag is set, and text is encoded, perltidy will look for the module Unicode::GCString and, if found, will use it to obtain character display widths. This can improve displayed vertical alignment for files with wide characters. It is a nice feature but it is off by default to avoid conflicting formatting when there are multiple developers. Perltidy installation does not require Unicode::GCString, so users wanting to use this feature need set this flag and also to install Unicode::GCString separately.
If this flag is set and perltidy does not find module Unicode::GCString, a warning message will be produced and processing will continue but without the potential benefit provided by the module.
Also note that actual vertical alignment depends upon the fonts used by the text display software, so vertical alignment may not be optimal even when Unicode::GCString is used.
where s=win, dos, unix, or mac. This flag tells perltidy to output line endings for a specific system. Normally, perltidy writes files with the line separator character of the host system. The win and dos flags have an identical result.
This flag tells perltidy to write its output files with the same line endings as the input file, if possible. It should work for dos, unix, and mac line endings. It will only work if perltidy input comes from a filename (rather than stdin, for example). If perltidy has trouble determining the input file line ending, it will revert to the default behavior of using the line ending of the host system.
This flag, which is enabled by default, allows perltidy to terminate the last line of the output stream with a newline character, regardless of whether or not the input stream was terminated with a newline character. If this flag is negated, with -natnl, then perltidy will add a terminal newline to the the output stream only if the input stream is terminated with a newline.
Negating this flag may be useful for manipulating one-line scripts intended for use on a command line.
This flag causes perltidy to do n complete iterations. The reason for this flag is that code beautification is an iterative process and in some cases the output from perltidy can be different if it is applied a second time. For most purposes the default of n=1 should be satisfactory. However n=2 can be useful when a major style change is being made, or when code is being beautified on check-in to a source code control system. It has been found to be extremely rare for the output to change after 2 iterations. If a value n is greater than 2 is input then a convergence test will be used to stop the iterations as soon as possible, almost always after 2 iterations. See the next item for a simplified iteration control.
This flag has no effect when perltidy is used to generate html.
This flag is equivalent to -it=4 and is included to simplify iteration control. For all practical purposes one either does or does not want to be sure that the output is converged, and there is no penalty to using a large iteration limit since perltidy will check for convergence and stop iterating as soon as possible. The default is -nconv (no convergence check). Using -conv will approximately double run time since typically one extra iteration is required to verify convergence. No extra iterations are required if no new line breaks are made, and two extra iterations are occasionally needed when reformatting complex code structures, such as deeply nested ternary statements.
Continuation indentation is extra indentation spaces applied when a long line is broken. The default is n=2, illustrated here:
my $level = # -ci=2
( $max_index_to_go >= 0 ) ? $levels_to_go[0] : $last_output_level;The same example, with n=0, is a little harder to read:
my $level = # -ci=0
( $max_index_to_go >= 0 ) ? $levels_to_go[0] : $last_output_level;The value given to -ci is also used by some commands when a small space is required. Examples are commands for outdenting labels, --outdent-labels (-ola), and control keywords, --outdent-keywords (-okw).
When default values are not used, it is recommended that either
(1) the value n given with -ci=n be no more than about one-half of the number of spaces assigned to a full indentation level on the -i=n command, or
(2) the flag --extended-continuation-indentation is used (see next section).
This flag allows perltidy to use some improvements which have been made to its indentation model. One of the things it does is "extend" continuation indentation deeper into structures, hence the name. The improved indentation is particularly noticeable when the flags -ci=n and -i=n use the same value of n. There are no significant disadvantages to using this flag, but to avoid disturbing existing formatting the default is not to use it, -nxci.
Please see the section "-pbp, --perl-best-practices" for an example of how this flag can improve the formatting of ternary statements. It can also improve indentation of some multi-line qw lists as shown below.
# perltidy
foreach $color (
qw(
AntiqueWhite3 Bisque1 Bisque2 Bisque3 Bisque4
SlateBlue3 RoyalBlue1 SteelBlue2 DeepSkyBlue3
),
qw(
LightBlue1 DarkSlateGray1 Aquamarine2 DarkSeaGreen2
SeaGreen1 Yellow1 IndianRed1 IndianRed2 Tan1 Tan4
)
)
# perltidy -xci
foreach $color (
qw(
AntiqueWhite3 Bisque1 Bisque2 Bisque3 Bisque4
SlateBlue3 RoyalBlue1 SteelBlue2 DeepSkyBlue3
),
qw(
LightBlue1 DarkSlateGray1 Aquamarine2 DarkSeaGreen2
SeaGreen1 Yellow1 IndianRed1 IndianRed2 Tan1 Tan4
)
)This flag allows perltidy to remove continuation indentation in some special cases where it is not really unnecessary. For a simple example, the default formatting for the following snippet is:
# perltidy -nmci
$self->blurt( "Error: No INPUT definition for type '$type', typekind '"
. $type->xstype
. "' found" );The second and third lines are one level deep in a container, and are also statement continuations, so they get indented by the sum of the full indentation -i value and the continuation indentation -ci value. If this flag is set, the indentation is reduced by -ci spaces, giving
# perltidy -mci
$self->blurt( "Error: No INPUT definition for type '$type', typekind '"
. $type->xstype
. "' found" );This flag is off by default.
By default, perltidy examines the input file and tries to determine the starting indentation level. While it is often zero, it may not be zero for a code snippet being sent from an editing session.
To guess the starting indentation level perltidy simply assumes that indentation scheme used to create the code snippet is the same as is being used for the current perltidy process. This is the only sensible guess that can be made. It should be correct if this is true, but otherwise it probably won't. For example, if the input script was written with -i=2 and the current perltidy flags have -i=4, the wrong initial indentation will be guessed for a code snippet which has non-zero initial indentation. Likewise, if an entabbing scheme is used in the input script and not in the current process then the guessed indentation will be wrong.
If the default method does not work correctly, or you want to change the starting level, use -sil=n, to force the starting level to be n.
These flags provide an alternative indentation method for list data. The original flag for this is -lp, but it has some limitations (explained below) which are avoided with the newer -xlp flag. So -xlp is probably the better choice for new work, but the -lp flag is retained to minimize changes to existing formatting. If you enter both -lp and -xlp, then -xlp will be used.
In the default indentation method perltidy indents lists with 4 spaces, or whatever value is specified with -i=n. Here is a small list formatted in this way:
# perltidy (default)
@month_of_year = (
'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun',
'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec'
);The -lp or -xlp flags add extra indentation to cause the data to begin past the opening parentheses of a sub call or list, or opening square bracket of an anonymous array, or opening curly brace of an anonymous hash. With this option, the above list would become:
# perltidy -lp or -xlp
@month_of_year = (
'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun',
'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec'
);If the available line length (see -l=n ) does not permit this much space, perltidy will use less. For alternate placement of the closing paren, see the next section.
These flags have no effect on code BLOCKS, such as if/then/else blocks, which always use whatever is specified with -i=n.
Some limitations on these flags are:
A limitation on -lp, but not -xlp, occurs in situations where perltidy does not have complete freedom to choose line breaks. Then it may temporarily revert to its default indentation method. This can occur for example if there are blank lines, block comments, multi-line quotes, or side comments between the opening and closing parens, braces, or brackets. It will also occur if a multi-line anonymous sub occurs within a container since that will impose specific line breaks (such as line breaks after statements).
For both the -lp and -xlp flags, any parameter which significantly restricts the ability of perltidy to choose newlines will conflict with these flags and will cause them to be deactivated. These include --indent-only, --freeze-newlines, -noadd-newlines, and -nodelete-old-newlines.
The -lp and -xlp options may not be used together with the -t tabs option. They may, however, be used with the -et=n tab method
There are some potential disadvantages of this indentation method compared to the default method that should be noted:
The available line length can quickly be used up if variable names are long. This can cause deeply nested code to quickly reach the line length limit, and become badly formatted, much sooner than would occur with the default indentation method.
Since the indentation depends on the lengths of variable names, small changes in variable names can cause changes in indentation over many lines in a file. This means that minor name changes can produce significant file differences. This can be annoying and does not occur with the default indentation method.
Some things that can be done to minimize these problems are:
Increase --maximum-line-length=n above the default n=80 characters if necessary.
If you use -xlp then long side comments can limit the indentation over multiple lines. Consider adding the flag --ignore-side-comment-lengths to prevent this, or minimizing the use of side comments.
Apply this style in a limited way. By default, it applies to all list containers (not just lists in parentheses). The next section describes how to limit this style to, for example, just function calls. The default indentation method will be applied elsewhere.
The following discussion is written for -lp but applies equally to the newer -xlp version. By default, the -lp flag applies to as many containers as possible. The set of containers to which the -lp style applies can be reduced by either one of these two flags:
Use --line-up-parentheses-inclusion-list=s (-lpil=s) to specify the containers to which -lp applies, or
use --line-up-parentheses-exclusion-list=s (-lpxl=s) to specify the containers to which -lp does NOT apply.
Only one of these two flags may be used. Both flags can achieve the same result, but the -lpil=s flag is much easier to describe and use and is recommended. The -lpxl=s flag was the original implementation and is only retained for backwards compatibility.
This list s for these parameters is a string with space-separated items. Each item consists of up to three pieces of information in this order: (1) an optional letter code (2) a required container type, and (3) an optional numeric code.
The only required piece of information is a container type, which is one of '(', '[', or '{'. For example the string
-lpil='('means use -lp formatting only on lists within parentheses, not lists in square-brackets or braces. The same thing could alternatively be specified with
-lpxl = '[ {'which says to exclude lists within square-brackets and braces. So what remains is lists within parentheses.
A second optional item of information which can be given for parentheses is an alphanumeric letter which is used to limit the selection further depending on the type of token immediately before the paren. The possible letters are currently 'k', 'K', 'f', 'F', 'w', and 'W', with these meanings for matching whatever precedes an opening paren:
'k' matches if the previous nonblank token is a perl built-in keyword
(such as 'if', 'while'),
'K' matches if 'k' does not, meaning that the previous token is not a keyword.
'f' matches if the previous token is a function other than a keyword.
'F' matches if 'f' does not.
'w' matches if either 'k' or 'f' match.
'W' matches if 'w' does not.For example:
-lpil = 'f('means only apply -lp to function calls, and
-lpil = 'w('means only apply -lp to parenthesized lists which follow a function or a keyword.
This last example could alternatively be written using the -lpxl=s flag as
-lpxl = '[ { W('which says exclude -lp for lists within square-brackets, braces, and parens NOT preceded by a keyword or function. Clearly, the -lpil=s method is easier to understand.
An optional numeric code may follow any of the container types to further refine the selection based on container contents. The numeric codes are:
'0' or blank: no restriction is placed on container contents
'1' the container contents must be a simple list without sublists
'2' the container contents must be a simple list without sublists, without
code blocks, and without ternary operatorsFor example,
-lpil = 'f(2'means only apply -lp to function calls with simple lists (not containing any sublists, code blocks or ternary expressions).
The -cti=n flag controls the indentation of a line beginning with a ), ], or a non-block }. Such a line receives:
-cti = 0 no extra indentation (default)
-cti = 1 extra indentation such that the closing token
aligns with its opening token.
-cti = 2 one extra indentation level if the line looks like:
); or ]; or };
-cti = 3 one extra indentation level alwaysThe flags -cti=1 and -cti=2 work well with the --line-up-parentheses (-lp) flag (previous section).
# perltidy -lp -cti=1
@month_of_year = (
'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun',
'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec'
);
# perltidy -lp -cti=2
@month_of_year = (
'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun',
'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec'
);These flags are merely hints to the formatter and they may not always be followed. In particular, if -lp is not being used, the indentation for cti=1 is constrained to be no more than one indentation level.
If desired, this control can be applied independently to each of the closing container token types. In fact, -cti=n is merely an abbreviation for -cpi=n -csbi=n -cbi=n, where: -cpi or --closing-paren-indentation controls )'s, -csbi or --closing-square-bracket-indentation controls ]'s, -cbi or --closing-brace-indentation controls non-block }'s.
The -icp flag is equivalent to -cti=2, described in the previous section. The -nicp flag is equivalent -cti=0. They are included for backwards compatibility.
The -icb option gives one extra level of indentation to a brace which terminates a code block . For example,
if ($task) {
yyy();
} # -icb
else {
zzz();
}The default is not to do this, indicated by -nicb.
Normally, lines of code contained within a pair of block braces receive one additional level of indentation. This flag, which is enabled by default, causes perltidy to look for opening block braces which are followed by a special side comment. This special side comment is #<<< by default. If found, the code between this opening brace and its corresponding closing brace will not be given the normal extra indentation level. For example:
{ #<<< a closure to contain lexical vars
my $var; # this line does not get one level of indentation
...
}
# this line does not 'see' $var;This can be useful, for example, when combining code from different files. Different sections of code can be placed within braces to keep their lexical variables from being visible to the end of the file. To keep the new braces from causing all of their contained code to be indented if you run perltidy, and possibly introducing new line breaks in long lines, you can mark the opening braces with this special side comment.
Only the opening brace needs to be marked, since perltidy knows where the closing brace is. Braces contained within marked braces may also be marked as non-indenting.
If your code happens to have some opening braces followed by '#<<<', and you don't want this behavior, you can use -nnib to deactivate it. To make it easy to remember, the default string is the same as the string for starting a format-skipping section. There is no confusion because in that case it is for a block comment rather than a side-comment.
The special side comment can be changed with the next parameter.
The -nibp=string parameter may be used to change the marker for non-indenting braces. The default is equivalent to -nibp='#<<<'. The string that you enter must begin with a # and should be in quotes as necessary to get past the command shell of your system. This string is the leading text of a regex pattern that is constructed by prepending a '^' and appending a'\s', so you must also include backslashes for characters to be taken literally rather than as patterns.
For example, to match the side comment '#++', the parameter would be
-nibp='#\+\+'When -olq is set, lines which is a quoted string longer than the value maximum-line-length will have their indentation removed to make them more readable. This is the default. To prevent such out-denting, use -nolq.
This command is equivalent to the combination --outdent-long-quotes and --outdent-long-comments, and it is included for compatibility with previous versions of perltidy. The negation of this also works, -noll or --nooutdent-long-lines, and is equivalent to setting -nolq and -nolc.
This command will cause labels to be outdented by the number of spaces defined by --continuation-indentation=n, if possible. This is the default. For example:
my $i;
LOOP: while ( $i = <FOTOS> ) {
chomp($i);
next unless $i;
fixit($i);
}Use -nola to prevent this. To control line breaks after labels see "-bal=n, --break-after-labels=n".
The command -okw will cause certain leading control keywords to be outdented by the number of spaces defined by --continuation-indentation=nspaces, if possible. By default, these keywords are redo, next, last, goto, and return. The intention is to make these control keywords easier to see. To change this list of keywords being outdented, see the next section.
For example, using perltidy -okw on the previous example gives:
my $i;
LOOP: while ( $i = <FOTOS> ) {
chomp($i);
next unless $i;
fixit($i);
}Notice that the keyword next has been outdented. The default is not to do this.
This command can be used to change the keywords which are outdented with the -okw command. The parameter string is a required list of perl keywords, which should be placed in quotes if there are more than one. By itself, it does not cause any outdenting to occur, so the -okw command is still required.
For example, the commands -okwl="next last redo goto" -okw will cause those four keywords to be outdented.
Whitespace refers to the blank space between variables, operators, and other code tokens.
This flag causes your original whitespace to remain unchanged, and causes the rest of the whitespace commands in this section, the Code Indentation section, and the Comment Control section to be ignored.
Here the term "tightness" will mean the closeness with which pairs of enclosing tokens, such as parentheses, contain the quantities within. A numerical value of 0, 1, or 2 defines the tightness, with 0 being least tight and 2 being most tight. Spaces within containers are always symmetric, so if there is a space after a ( then there will be a space before the corresponding ).
The -pt=n or --paren-tightness=n parameter controls the space within parens. The example below shows the effect of the three possible values, 0, 1, and 2:
if ( ( my $len_tab = length( $tabstr ) ) > 0 ) { # -pt=0
if ( ( my $len_tab = length($tabstr) ) > 0 ) { # -pt=1 (default)
if ((my $len_tab = length($tabstr)) > 0) { # -pt=2When n is 0, there is always a space to the right of a '(' and to the left of a ')'. For n=2 there is never a space. For n=1, the default, there is a space unless the quantity within the parens is a single token, such as an identifier or quoted string.
Likewise, the parameter -sbt=n or --square-bracket-tightness=n controls the space within square brackets, as illustrated below.
$width = $col[ $j + $k ] - $col[ $j ]; # -sbt=0
$width = $col[ $j + $k ] - $col[$j]; # -sbt=1 (default)
$width = $col[$j + $k] - $col[$j]; # -sbt=2Curly braces which do not contain code blocks are controlled by the parameter -bt=n or --brace-tightness=n.
$obj->{ $parsed_sql->{ 'table' }[0] }; # -bt=0
$obj->{ $parsed_sql->{'table'}[0] }; # -bt=1 (default)
$obj->{$parsed_sql->{'table'}[0]}; # -bt=2And finally, curly braces which contain blocks of code are controlled by the parameter -bbt=n or --block-brace-tightness=n as illustrated in the example below.
%bf = map { $_ => -M $_ } grep { /\.deb$/ } dirents '.'; # -bbt=0 (default)
%bf = map { $_ => -M $_ } grep {/\.deb$/} dirents '.'; # -bbt=1
%bf = map {$_ => -M $_} grep {/\.deb$/} dirents '.'; # -bbt=2To simplify input in the case that all of the tightness flags have the same value <n>, the parameter <-act=n> or --all-containers-tightness=n is an abbreviation for the combination <-pt=n -sbt=n -bt=n -bbt=n>.
The previous section described two controls for spacing within curly braces, namely -block-brace-tightness=n for code block braces and -brace-tightness=n for all other braces.
There is a little fuzziness in this division of brace types though because the curly braces considered by perltidy to contain code blocks for formatting purposes, such as highlighting code structure, exclude some of the small code blocks used by Perl mainly for isolating terms. These include curly braces following a keyword where an indirect object might occur, or curly braces following a type symbol. For example, perltidy does not mark the following braces as code block braces:
print {*STDERR} $message;
return ${$foo};Consequently, the spacing within these small braced containers by default follows the flag --brace-tightness=n rather than --block-brace-tightness=n, as one might expect.
If desired, small blocks such as these can be made to instead follow the spacing defined by the --block-brace-tightness=n flag by setting --extended-block-tightness. The specific types of small blocks to which this parameter applies is controlled by a companion control parameter, described in the next section.
Note that if the two flags --block-brace-tightness=n and --brace-tightness=n have the same value n then there would be no reason to set this flag.
The previous parameter --extended-block-tightness (-xbt) can be made to apply to curly braces preceded by any of the keywords
print printf exec system sayand/or the special symbols
$ @ % & * $#The parameter string s may contain a selection of these keywords and symbols to indicate the brace types to which --extended-block-tightness applies. For convenience, all of the keywords can be selected with 'k', and all of the special symbols can be selected with 't'. The default is equivalent to -xbtl='k', which selects all of the keywords.
Examples:
-xbtl='k' # selects just the keywords [DEFAULT]
-xbtl="t" # selects just the special type symbols
-xbtl="k t" # selects all keywords and symbols, or more simply
-xbtl="kt" # selects all keywords and symbols
-xbtl="print say" # selects just keywords B<print> and B<say>:Here are some formatting examples using the default values of --brace-tightness=n and --block-brace-tightness=n. Note that in these examples $ref is in block braces but $key is not.
# default formatting
print {*STDERR} $message;
my $val = ${$ref}{$key};
# perltidy -xbt or
# perltidy -xbt -xbtl=k
print { *STDERR } $message;
my $val = ${$ref}{$key};
# perltidy -xbt -xbtl=t
print {*STDERR} $message;
my $val = ${ $ref }{$key};
# perltidy -xbt -xbtl=kt
print { *STDERR } $message;
my $val = ${ $ref }{$key};Finally, note that this parameter merely changes the way that the parameter --extended-block-tightness works. It has no effect unless --extended-block-tightness is actually set.
The flag -tso causes certain perl token sequences (secret operators) which might be considered to be a single operator to be formatted "tightly" (without spaces). The operators currently modified by this flag are:
0+ +0 ()x!! ~~<> ,=> =( )=For example the sequence 0 +, which converts a string to a number, would be formatted without a space: 0+ when the -tso flag is set. This flag is off by default.
Some programmers prefer a space before all terminal semicolons. The default is for no such space, and is indicated with -nsts or --nospace-terminal-semicolon.
$i = 1 ; # -sts
$i = 1; # -nsts (default)Semicolons within for loops may sometimes be hard to see, particularly when commas are also present. This option places spaces on both sides of these special semicolons, and is the default. Use -nsfs or --nospace-for-semicolon to deactivate it.
for ( @a = @$ap, $u = shift @a ; @a ; $u = $v ) { # -sfs (default)
for ( @a = @$ap, $u = shift @a; @a; $u = $v ) { # -nsfsSetting -asc allows perltidy to add any missing optional semicolon at the end of a line which is followed by a closing curly brace on the next line. This is the default, and may be deactivated with -nasc or --noadd-semicolons.
Setting -dsm allows perltidy to delete extra semicolons which are simply empty statements. This is the default, and may be deactivated with -ndsm or --nodelete-semicolons. (Such semicolons are not deleted, however, if they would promote a side comment to a block comment).
Setting this option allows perltidy to add certain whitespace to improve code readability. This is the default. If you do not want any whitespace added, but are willing to have some whitespace deleted, use -naws. (Use --freeze-whitespace (-fws) to leave whitespace completely unchanged).
Setting this option allows perltidy to remove some old whitespace between characters, if necessary. This is the default. If you do not want any old whitespace removed, use -ndws or --nodelete-old-whitespace.
For those who want more detailed control over the whitespace around tokens, there are four parameters which can directly modify the default whitespace rules built into perltidy for any token. They are:
-wls=s or --want-left-space=s,
-nwls=s or --nowant-left-space=s,
-wrs=s or --want-right-space=s,
-nwrs=s or --nowant-right-space=s.
These parameters are each followed by a quoted string, s, containing a list of token types. No more than one of each of these parameters should be specified, because repeating a command-line parameter always overwrites the previous one before perltidy ever sees it.
To illustrate how these are used, suppose it is desired that there be no space on either side of the token types = + - / *. The following two parameters would specify this desire:
-nwls="= + - / *" -nwrs="= + - / *"(Note that the token types are in quotes, and that they are separated by spaces). With these modified whitespace rules, the following line of math:
$root = -$b + sqrt( $b * $b - 4. * $a * $c ) / ( 2. * $a );becomes this:
$root=-$b+sqrt( $b*$b-4.*$a*$c )/( 2.*$a );These parameters should be considered to be hints to perltidy rather than fixed rules, because perltidy must try to resolve conflicts that arise between them and all of the other rules that it uses. One conflict that can arise is if, between two tokens, the left token wants a space and the right one doesn't. In this case, the token not wanting a space takes priority.
It is necessary to have a list of all token types in order to create this type of input. Such a list can be obtained by the command --dump-token-types. Also try the -D flag on a short snippet of code and look at the .DEBUG file to see the tokenization.
To illustrate, suppose we do not want a space after a colon which introduces a sub attribute. We need to know its type. It is not a colon because that is the type of a ternary operator. The ouput of --dump-token-types states that it is token type 'A'. To verify this, we can run perltidy -D on a short piece of code containing such a colon, such as
sub foo : lvalue;This produces a .DEBUG file which contains
1: sub foo : lvalue;
1: SSSSSSSbAbwwwwww;The top line is the input, and the bottom line shows the token types. The 'A' beneath the colon verifies that it is type A.
So to prevent a space on the right of this colon we can use
# perltidy -nwrs='A'
sub foo :lvalue;WARNING Be sure to put these tokens in quotes to avoid having them misinterpreted by your command shell.
The various parameters controlling whitespace within a program are requests which perltidy follows as well as possible, but there are a number of situations where changing whitespace could change program behavior and is not done. Some of these are obvious; for example, we should not remove the space between the two plus symbols in '$x+ +$y' to avoid creating a '++' operator. Some are more subtle and involve the whitespace around bareword symbols and locations of possible filehandles. For example, consider the problem of formatting the following subroutine:
sub print_div {
my ($x,$y)=@_;
print $x/$y;
}Suppose the user requests that / signs have a space to the left but not to the right. Perltidy will refuse to do this, but if this were done the result would be
sub print_div {
my ($x,$y)=@_;
print $x /$y;
}If formatted in this way, the program will not run (at least with recent versions of perl) because the $x is taken to be a filehandle and / is assumed to start a quote. In a complex program, there might happen to be a / which terminates the multiline quote without a syntax error, allowing the program to run, but not as intended.
Related issues arise with other binary operator symbols, such as + and -, and in older versions of perl there could be problems with ternary operators. So to avoid changing program behavior, perltidy has the simple rule that whitespace around possible filehandles is left unchanged. Likewise, whitespace around barewords is left unchanged. The reason is that if the barewords are defined in other modules, or in code that has not even been written yet, perltidy will not have seen their prototypes and must treat them cautiously.
In perltidy this is implemented in the tokenizer by marking token following a print keyword as a special type Z. When formatting is being done, whitespace following this token type is generally left unchanged as a precaution against changing program behavior. This is excessively conservative but simple and easy to implement. Keywords which are treated similarly to print include printf, sort, exec, system. Changes in spacing around parameters following these keywords may have to be made manually. For example, the space, or lack of space, after the parameter $foo in the following line will be unchanged in formatting.
system($foo );
system($foo);To find if a token is of type Z you can use perltidy -DEBUG. For the first line above the result is
1: system($foo );
1: kkkkkk{ZZZZb};which shows that system is type k (keyword) and $foo is type Z.
Despite these precautions, it is still possible to introduce syntax errors with some asymmetric whitespace rules, particularly when call parameters are not placed in containing parens or braces. For example, the following two lines will be parsed by perl without a syntax error:
# original programming, syntax ok
my @newkeys = map $_-$nrecs+@data, @oldkeys;
# perltidy default, syntax ok
my @newkeys = map $_ - $nrecs + @data, @oldkeys;But the following will give a syntax error:
# perltidy -nwrs='-'
my @newkeys = map $_ -$nrecs + @data, @oldkeys;For another example, the following two lines will be parsed without syntax error:
# original programming, syntax ok
for my $severity ( reverse $SEVERITY_LOWEST+1 .. $SEVERITY_HIGHEST ) { ... }
# perltidy default, syntax ok
for my $severity ( reverse $SEVERITY_LOWEST + 1 .. $SEVERITY_HIGHEST ) { ... }But the following will give a syntax error:
# perltidy -nwrs='+', syntax error:
for my $severity ( reverse $SEVERITY_LOWEST +1 .. $SEVERITY_HIGHEST ) { ... }To avoid subtle parsing problems like this, it is best to avoid spacing a binary operator asymmetrically with a space on the left but not on the right.
When an opening paren follows a Perl keyword, no space is introduced after the keyword, unless it is (by default) one of these:
my local our state and or xor err eq ne if else elsif until unless
while for foreach return switch case given when catchThese defaults can be modified with two commands:
-sak=s or --space-after-keyword=s adds keywords.
-nsak=s or --nospace-after-keyword=s removes keywords.
where s is a list of keywords (in quotes if necessary). For example,
my ( $a, $b, $c ) = @_; # default
my( $a, $b, $c ) = @_; # -nsak="my local our"The abbreviation -nsak='*' is equivalent to including all of the keywords in the above list.
When both -nsak=s and -sak=s commands are included, the -nsak=s command is executed first. For example, to have space after only the keywords (my, local, our) you could use -nsak="*" -sak="my local our".
To put a space after all keywords, see the next item.
When an opening paren follows a function or keyword, no space is introduced after the keyword except for the keywords noted in the previous item. To always put a space between a function or keyword and its opening paren, use the command:
-skp or --space-keyword-paren
You may also want to use the flag -sfp (next item) too.
When an opening paren follows a function the default and recommended formatting is not to introduce a space. To cause a space to be introduced use:
-sfp or --space-function-paren
myfunc( $a, $b, $c ); # default
myfunc ( $a, $b, $c ); # -sfpYou will probably also want to use the flag -skp (previous item) too.
The parameter is not recommended because spacing a function paren can make a program vulnerable to parsing problems by Perl. For example, the following two-line program will run as written but will have a syntax error if reformatted with -sfp:
if ( -e filename() ) { print "I'm here\n"; }
sub filename { return $0 }In this particular case the syntax error can be removed if the line order is reversed, so that Perl parses 'sub filename' first.
A side-effect of using the -sfp flag is that the parens may become vertically aligned. For example,
# perltidy -sfp
myfun ( $aaa, $b, $cc );
mylongfun ( $a, $b, $c );This is the default behavior. To prevent this alignment use -nfpva:
# perltidy -sfp -nfpva
myfun ( $aaa, $b, $cc );
mylongfun ( $a, $b, $c );This flag can be used to control whether a function prototype is preceded by a space. For example, the following prototype does not have a space.
sub usage();This integer n may have the value 0, 1, or 2 as follows:
-spp=0 means no space before the paren
-spp=1 means follow the example of the source code [DEFAULT]
-spp=2 means always put a space before the parenThe default is -spp=1, meaning that a space will be used if and only if there is one in the source code. Given the above line of code, the result of applying the different options would be:
sub usage(); # n=0 [no space]
sub usage(); # n=1 [default; follows input]
sub usage (); # n=2 [space]This flag is analogous to the previous except that it applies to the space before the opening paren of a sub signature rather than a sub prototype.
For example, consider the following line:
sub circle( $xc, $yc, $rad )This space before the opening paren can be controlled with integer n which may have the value 0, 1, or 2 with these meanings:
-ssp=0 means no space before the paren
-ssp=1 means follow the example of the source code [DEFAULT]
-ssp=2 means always put a space before the parenThe default is -ssp=1, meaning that will be a space in the output if, and only if, there is one in the input. Given the above line of code, the result of applying the different options would be:
sub circle( $xc, $yc, $rad ) # n=0 [no space]
sub circle( $xc, $yc, $rad ) # n=1 [default; same as input]
sub circle ( $xc, $yc, $rad ) # n=2 [space]The space inside of an opening paren, which itself follows a certain keyword, can be controlled by this parameter. The space on the inside of the corresponding closing paren will be treated in the same (balanced) manner. This parameter has precedence over any other paren spacing rules. The values of n are as follows:
-kpit=0 means always put a space (not tight)
-kpit=1 means ignore this parameter [default]
-kpit=2 means never put a space (tight)To illustrate, the following snippet is shown formatted in three ways:
if ( seek( DATA, 0, 0 ) ) { ... } # perltidy (default)
if (seek(DATA, 0, 0)) { ... } # perltidy -pt=2
if ( seek(DATA, 0, 0) ) { ... } # perltidy -pt=2 -kpit=0In the second case the -pt=2 parameter makes all of the parens tight. In the third case the -kpit=0 flag causes the space within the 'if' parens to have a space, since 'if' is one of the keywords to which the -kpit flag applies by default. The remaining parens are still tight because of the -pt=2 parameter.
The set of keywords to which this parameter applies are by default are:
if elsif unless while until for foreachThese can be changed with the parameter -kpitl=s described in the next section.
This command can be used to change the keywords to which the previous parameter, -kpit=n, applies. The parameter string is a required list either keywords or functions, which should be placed in quotes if there are more than one. By itself, this parameter does not cause any change in spacing, so the -kpit=n command is still required.
For example, the commands -kpitl="if else while" -kpit=2 will cause the just the spaces inside parens following 'if', 'else', and 'while' keywords to follow the tightness value indicated by the -kpit=2 flag.
In the following example some extra space has been inserted on the second line between the two open parens. This extra space is called "logical padding" and is intended to help align similar things vertically in some logical or ternary expressions.
# perltidy [default formatting]
$same =
( ( $aP eq $bP )
&& ( $aS eq $bS )
&& ( $aT eq $bT )
&& ( $a->{'title'} eq $b->{'title'} )
&& ( $a->{'href'} eq $b->{'href'} ) );Note that this is considered to be a different operation from "vertical alignment" because space at just one line is being adjusted, whereas in "vertical alignment" the spaces at all lines are being adjusted. So it is sort of a local version of vertical alignment.
Here is an example involving a ternary operator:
# perltidy [default formatting]
$bits =
$top > 0xffff ? 32
: $top > 0xff ? 16
: $top > 1 ? 8
: 1;This behavior is controlled with the flag --logical-padding, which is set 'on' by default. If it is not desired it can be turned off using --nological-padding or -nlop. The above two examples become, with -nlop:
# perltidy -nlop
$same =
( ( $aP eq $bP )
&& ( $aS eq $bS )
&& ( $aT eq $bT )
&& ( $a->{'title'} eq $b->{'title'} )
&& ( $a->{'href'} eq $b->{'href'} ) );
# perltidy -nlop
$bits =
$top > 0xffff ? 32
: $top > 0xff ? 16
: $top > 1 ? 8
: 1;qw quotes-tqw or --trim-qw provide the default behavior of trimming spaces around multi-line qw quotes and indenting them appropriately.
-ntqw or --notrim-qw cause leading and trailing whitespace around multi-line qw quotes to be left unchanged. This option will not normally be necessary, but was added for testing purposes, because in some versions of perl, trimming qw quotes changes the syntax tree.
lines like
$str1=\"string1";
$str2=\'string2';can confuse syntax highlighters unless a space is included between the backslash and the single or double quotation mark.
this can be controlled with the value of n as follows:
-sbq=0 means no space between the backslash and quote
-sbq=1 means follow the example of the source code
-sbq=2 means always put a space between the backslash and quoteThe default is -sbq=1, meaning that a space will be used if there is one in the source code.
-trp or --trim-pod will remove trailing whitespace from lines of POD. The default is not to do this.
Perltidy has a number of ways to control the appearance of both block comments and side comments. The term block comment here refers to a full-line comment, whereas side comment will refer to a comment which appears on a line to the right of some code.
Perltidy does not do any word wrapping of commented text to match a selected maximum line length. This is because there is no way to determine if this is appropriate for the given content. However, an interactive program named perlcomment.pl is available in the examples folder of the perltidy distribution which can assist in doing this.
Block comments normally look best when they are indented to the same level as the code which follows them. This is the default behavior, but you may use -nibc to keep block comments left-justified. Here is an example:
# this comment is indented (-ibc, default)
if ($task) { yyy(); }The alternative is -nibc:
# this comment is not indented (-nibc)
if ($task) { yyy(); }See also the next item, -isbc, as well as -sbc, for other ways to have some indented and some outdented block comments.
If there is no leading space on the line, then the comment will not be indented, and otherwise it may be.
If both -ibc and -isbc are set, then -isbc takes priority.
When -olc is set, lines which are full-line (block) comments longer than the value maximum-line-length will have their indentation removed. This is the default; use -nolc to prevent outdenting.
Side comments look best when lined up several spaces to the right of code. Perltidy will try to keep comments at least n spaces to the right. The default is n=4 spaces.
This parameter tells perltidy to line up side comments in column number n whenever possible. The default, n=0, will not do this.
This parameter causes perltidy to ignore the length of side comments when setting line breaks. The default, -niscl, is to include the length of side comments when breaking lines to stay within the length prescribed by the -l=n maximum line length parameter. For example, the following long single line would remain intact with -l=80 and -iscl:
perltidy -l=80 -iscl
$vmsfile =~ s/;[\d\-]*$//; # Clip off version number; we can use a newer version as wellwhereas without the -iscl flag the line will be broken:
perltidy -l=80
$vmsfile =~ s/;[\d\-]*$//
; # Clip off version number; we can use a newer version as wellPerltidy, by default, will look for side comments beginning with ## no critic and ignore their lengths when making line break decisions, even if the user has not set -iscl. The reason is that an unwanted line break can make these special comments ineffective in controlling perlcritic.
Setting --ignore-perlcritic-comments tells perltidy not to look for these ## no critic comments.
By default, perltidy tries to identify and align "hanging side comments", which are something like this:
my $IGNORE = 0; # This is a side comment
# This is a hanging side comment
# And so is thisA comment is considered to be a hanging side comment if (1) it immediately follows a line with a side comment, or another hanging side comment, and (2) there is some leading whitespace on the line. To deactivate this feature, use -nhsc or --nohanging-side-comments. If block comments are preceded by a blank line, or have no leading whitespace, they will not be mistaken as hanging side comments.
A closing side comment is a special comment which perltidy can automatically create and place after the closing brace of a code block. They can be useful for code maintenance and debugging. The command -csc (or --closing-side-comments) adds or updates closing side comments. For example, here is a small code snippet
sub message {
if ( !defined( $_[0] ) ) {
print("Hello, World\n");
}
else {
print( $_[0], "\n" );
}
}And here is the result of processing with perltidy -csc:
sub message {
if ( !defined( $_[0] ) ) {
print("Hello, World\n");
}
else {
print( $_[0], "\n" );
}
} ## end sub messageA closing side comment was added for sub message in this case, but not for the if and else blocks, because they were below the 6 line cutoff limit for adding closing side comments. This limit may be changed with the -csci command, described below.
The command -dcsc (or --delete-closing-side-comments) reverses this process and removes these comments.
Several commands are available to modify the behavior of these two basic commands, -csc and -dcsc:
where n is the minimum number of lines that a block must have in order for a closing side comment to be added. The default value is n=6. To illustrate:
# perltidy -csci=2 -csc
sub message {
if ( !defined( $_[0] ) ) {
print("Hello, World\n");
} ## end if ( !defined( $_[0] ))
else {
print( $_[0], "\n" );
} ## end else [ if ( !defined( $_[0] ))
} ## end sub messageNow the if and else blocks are commented. However, now this has become very cluttered.
where string is the prefix used before the name of the block type. The default prefix, shown above, is ## end. This string will be added to closing side comments, and it will also be used to recognize them in order to update, delete, and format them. Any comment identified as a closing side comment will be placed just a single space to the right of its closing brace.
where string is a list of block types to be tagged with closing side comments. By default, all code block types preceded by a keyword or label (such as if, sub, and so on) will be tagged. The -cscl command changes the default list to be any selected block types; see "Specifying Block Types". For example, the following command requests that only sub's, labels, BEGIN, and END blocks be affected by operations which add (-csc) or delete (-dcsc) closing side comments:
--closing-side-comment-list='sub : BEGIN END'where string is a list of block types which should NOT be tagged with closing side comments. If a block type appears in both -cscl and -cscxl, then -cscxl has priority and the block will not be tagged.
For example, the following command requests that anonymous subs should not be affected by any -csc or -dcsc operation:
--closing-side-comment-exclusion-list='asub'By default, no block types are excluded.
The text appended to certain block types, such as an if block, is whatever lies between the keyword introducing the block, such as if, and the opening brace. Since this might be too much text for a side comment, there needs to be a limit, and that is the purpose of this parameter. The default value is n=20, meaning that no additional tokens will be appended to this text after its length reaches 20 characters. Omitted text is indicated with .... (Tokens, including sub names, are never truncated, however, so actual lengths may exceed this). To illustrate, in the above example, the appended text of the first block is ( !defined( $_[0] ).... The existing limit of n=20 caused this text to be truncated, as indicated by the .... See the next flag for additional control of the abbreviated text.
As discussed in the previous item, when the closing-side-comment-maximum-text limit is exceeded the comment text must be truncated. Older versions of perltidy terminated with three dots, and this can still be achieved with -ncscb:
perltidy -csc -ncscb
} ## end foreach my $foo (sort { $b cmp $a ...However this causes a problem with editors which cannot recognize comments or are not configured to do so because they cannot "bounce" around in the text correctly. The -cscb flag has been added to help them by appending appropriate balancing structure:
perltidy -csc -cscb
} ## end foreach my $foo (sort { $b cmp $a ... })The default is -cscb.
The default, n=0, places the text of the opening if statement after any terminal else.
If n=2 is used, then each elsif is also given the text of the opening if statement. Also, an else will include the text of a preceding elsif statement. Note that this may result some long closing side comments.
If n=1 is used, the results will be the same as n=2 whenever the resulting line length is less than the maximum allowed.
When using closing-side-comments, and the closing-side-comment-maximum-text limit is exceeded, then the comment text must be abbreviated. It is terminated with three dots if the -cscb flag is negated:
perltidy -csc -ncscb
} ## end foreach my $foo (sort { $b cmp $a ...This causes a problem with older editors which do not recognize comments because they cannot "bounce" around in the text correctly. The -cscb flag tries to help them by appending appropriate terminal balancing structures:
perltidy -csc -cscb
} ## end foreach my $foo (sort { $b cmp $a ... })The default is -cscb.
This parameter is intended to help make the initial transition to the use of closing side comments. It causes two things to happen if a closing side comment replaces an existing, different closing side comment: first, an error message will be issued, and second, the original side comment will be placed alone on a new specially marked comment line for later attention.
The intent is to avoid clobbering existing hand-written side comments which happen to match the pattern of closing side comments. This flag should only be needed on the first run with -csc.
Important Notes on Closing Side Comments:
Closing side comments are only placed on lines terminated with a closing brace. Certain closing styles, such as the use of cuddled elses (-ce), preclude the generation of some closing side comments.
Please note that adding or deleting of closing side comments takes place only through the commands -csc or -dcsc. The other commands, if used, merely modify the behavior of these two commands.
It is recommended that the -cscw flag be used along with -csc on the first use of perltidy on a given file. This will prevent loss of any existing side comment data which happens to have the csc prefix.
Once you use -csc, you should continue to use it so that any closing side comments remain correct as code changes. Otherwise, these comments will become incorrect as the code is updated.
If you edit the closing side comments generated by perltidy, you must also change the prefix to be different from the closing side comment prefix. Otherwise, your edits will be lost when you rerun perltidy with -csc. For example, you could simply change ## end to be ## End, since the test is case sensitive. You may also want to use the -ssc flag to keep these modified closing side comments spaced the same as actual closing side comments.
Temporarily generating closing side comments is a useful technique for exploring and/or debugging a perl script, especially one written by someone else. You can always remove them with -dcsc.
Static block comments are block comments with a special leading pattern, ## by default, which will be treated slightly differently from other block comments. They effectively behave as if they had glue along their left and top edges, because they stick to the left edge and previous line when there is no blank spaces in those places. This option is particularly useful for controlling how commented code is displayed.
When -sbc is used, a block comment with a special leading pattern, ## by default, will be treated specially.
Comments so identified are treated as follows:
If there is no leading space on the line, then the comment will not be indented, and otherwise it may be,
no new blank line will be inserted before such a comment, and
such a comment will never become a hanging side comment.
For example, assuming @month_of_year is left-adjusted:
@month_of_year = ( # -sbc (default)
'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct',
## 'Dec', 'Nov'
'Nov', 'Dec');Without this convention, the above code would become
@month_of_year = ( # -nsbc
'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct',
## 'Dec', 'Nov'
'Nov', 'Dec'
);which is not as clear. The default is to use -sbc. This may be deactivated with -nsbc.
This parameter defines the prefix used to identify static block comments when the -sbc parameter is set. The default prefix is ##, corresponding to -sbcp=##. The prefix is actually part of a perl pattern used to match lines and it must either begin with # or ^#. In the first case a prefix ^\s* will be added to match any leading whitespace, while in the second case the pattern will match only comments with no leading whitespace. For example, to identify all comments as static block comments, one would use -sbcp=#. To identify all left-adjusted comments as static block comments, use -sbcp='^#'.
Please note that -sbcp merely defines the pattern used to identify static block comments; it will not be used unless the switch -sbc is set. Also, please be aware that since this string is used in a perl regular expression which identifies these comments, it must enable a valid regular expression to be formed.
A pattern which can be useful is:
-sbcp=^#{2,}[^\s#]This pattern requires a static block comment to have at least one character which is neither a # nor a space. It allows a line containing only '#' characters to be rejected as a static block comment. Such lines are often used at the start and end of header information in subroutines and should not be separated from the intervening comments, which typically begin with just a single '#'.
The command -osbc will cause static block comments to be outdented by the number of --continuation-spaces=n, if possible.
Static side comments are side comments with a special leading pattern. This option can be useful for controlling how commented code is displayed when it is a side comment.
When -ssc is used, a side comment with a static leading pattern, which is ## by default, will be spaced only a single space from previous character, and it will not be vertically aligned with other side comments.
The default is -nssc.
This parameter defines the prefix used to identify static side comments when the -ssc parameter is set. The default prefix is ##, corresponding to -sscp=##.
Please note that -sscp merely defines the pattern used to identify static side comments; it will not be used unless the switch -ssc is set. Also, note that this string is used in a perl regular expression which identifies these comments, so it must enable a valid regular expression to be formed.
Selected lines of code may be passed verbatim to the output without any formatting by marking the starting and ending lines with special comments. There are two options for doing this. The first option is called --format-skipping or -fs, and the second option is called --code-skipping or -cs.
In both cases the lines of code will be output without any changes. The difference is that in --format-skipping perltidy will still parse the marked lines of code and check for errors, whereas in --code-skipping perltidy will simply pass the lines to the output without any checking.
Both of these features are enabled by default and are invoked with special comment markers. --format-skipping uses starting and ending markers '#<<<' and '#>>>', like this:
#<<< format skipping: do not let perltidy change my nice formatting
my @list = (1,
1, 1,
1, 2, 1,
1, 3, 3, 1,
1, 4, 6, 4, 1,);
#>>>--code-skipping uses starting and ending markers '#<<V' and '#>>V', like this:
#<<V code skipping: perltidy will pass this verbatim without error checking
token ident_digit {
[ [ <?word> | _ | <?digit> ] <?ident_digit>
| <''>
]
};
#>>V(The last character V is like a < or > rotated 90 degrees). Additional text may appear on the special comment lines provided that it is separated from the marker by at least one space to highlight the sign, as in the above examples.
Any number of code-skipping or format-skipping sections may appear in a file. If an opening code-skipping or format-skipping comment is not followed by a corresponding closing comment, then skipping continues to the end of the file. If a closing code-skipping or format-skipping comment appears in a file but does not follow a corresponding opening comment, then it is treated as an ordinary comment without any special meaning.
It is recommended to use --code-skipping only if you need to hide a block of an extended syntax which would produce errors if parsed by perltidy, and use --format-skipping otherwise. This is because the --format-skipping option provides the benefits of error checking, and there are essentially no limitations on which lines to which it can be applied. The --code-skipping option, on the other hand, does not do error checking and its use is more restrictive because the code which remains, after skipping the marked lines, must be syntactically correct code with balanced containers.
These features should be used sparingly to avoid littering code with markers, but they can be helpful for working around occasional problems.
Note that it may be possible to avoid the use of --format-skipping for the specific case of a comma-separated list of values, as in the above example, by simply inserting a blank or comment somewhere between the opening and closing parens. See the section "Controlling List Formatting".
The following sections describe the available controls for these options. They should not normally be needed.
As explained above, this flag, which is enabled by default, causes any code between special beginning and ending comment markers to be passed to the output without formatting. The code between the comments is still checked for errors however. The default beginning marker is #<<< and the default ending marker is #>>>.
Format skipping begins when a format skipping beginning comment is seen and continues until a format-skipping ending comment is found.
This feature can be disabled with -nfs. This should not normally be necessary.
This and the next parameter allow the special beginning and ending comments to be changed. However, it is recommended that they only be changed if there is a conflict between the default values and some other use. If they are used, it is recommended that they only be entered in a .perltidyrc file, rather than on a command line. This is because properly escaping these parameters on a command line can be difficult.
If changed comment markers do not appear to be working, use the -log flag and examine the .LOG file to see if and where they are being detected.
The -fsb=string parameter may be used to change the beginning marker for format skipping. The default is equivalent to -fsb='#<<<'. The string that you enter must begin with a # and should be in quotes as necessary to get past the command shell of your system. It is actually the leading text of a pattern that is constructed by appending a '\s', so you must also include backslashes for characters to be taken literally rather than as patterns.
Some examples show how example strings become patterns:
-fsb='#\{\{\{' becomes /^#\{\{\{\s/ which matches #{{{ but not #{{{{
-fsb='#\*\*' becomes /^#\*\*\s/ which matches #** but not #***
-fsb='#\*{2,}' becomes /^#\*{2,}\s/ which matches #** and #*****The -fse=string is the corresponding parameter used to change the ending marker for format skipping. The default is equivalent to -fse='#<<<'.
The beginning and ending strings may be the same, but it is preferable to make them different for clarity.
As explained above, this flag, which is enabled by default, causes any code between special beginning and ending comment markers to be directly passed to the output without any error checking or formatting. Essentially, perltidy treats it as if it were a block of arbitrary text. The default beginning marker is #<<V and the default ending marker is #>>V.
This feature can be disabled with -ncs. This should not normally be necessary.
This may be used to change the beginning comment for a --code-skipping section, and its use is similar to the -fsb=string. The default is equivalent to -csb='#<<V'.
This may be used to change the ending comment for a --code-skipping section, and its use is similar to the -fse=string. The default is equivalent to -cse='#>>V'.
A command --line-range-tidy=n1:n2 is available to process just a selected range of lines of an input stream with perltidy. This command is mainly of interest for programming interactive code editors. When it is used, the entire input stream is read but just the selected range of lines of the input file are processed by the perltidy tokenizer and formatter, and then the stream is reassembled for output. The selected lines need to contain a complete statement or balanced container. Otherwise, a syntax error will occur and the code will not be tidied. There are a couple of limitations on the use of this command: (1) it may not be applied to multiple files, and (2) it only applies to code tidying and not, for example, html formatting.
The range of lines is specified by integers n1 and n2, where n1 is the first line number to be formatted (start counting with 1) and n2 is the last line number to be formatted. If n2 is not given, or exceeds the actual number of lines, then formatting continues to the end of the file.
Examples:
-line-range-tidy=43:109 # tidy lines 43 through 109
-line-range-tidy=' 43 : 109' # tidy lines 43 through 109 (spaces ok in quotes)
-line-range-tidy=1: # tidy all lines
-line-range-tidy=0:90 # ERROR (n1 must be >= 1)The parameters in this and the next sections control breaks after non-blank lines of code. Blank lines are controlled separately by parameters in the section "Blank Line Control".
By default, perltidy first deletes all old line break locations, and then it looks for good break points to match the desired line length. Use -ndnl or --nodelete-old-newlines to force perltidy to retain all old line break points.
By default, perltidy will add line breaks when necessary to create continuations of long lines and to improve the script appearance. Use -nanl or --noadd-newlines to prevent any new line breaks.
This flag does not prevent perltidy from eliminating existing line breaks; see --freeze-newlines to completely prevent changes to line break points.
If you do not want any changes to the line breaks within lines of code in your script, set -fnl, and they will remain fixed, and the rest of the commands in this section and sections "Controlling List Formatting", "Retaining or Ignoring Existing Line Breaks". You may want to use -nooutdent-long-lines with this.
Note: If you also want to keep your blank lines exactly as they are, you can use the --freeze-blank-lines flag which is described in the section "Blank Line Control".
Enable the "cuddled else" style, in which else and elsif are follow immediately after the curly brace closing the previous block. The default is not to use cuddled elses, and is indicated with the flag -nce or --nocuddled-else. Here is a comparison of the alternatives:
# -ce
if ($task) {
yyy();
} else {
zzz();
}
# -nce (default)
if ($task) {
yyy();
}
else {
zzz();
}In this example the keyword else is placed on the same line which begins with the preceding closing block brace and is followed by its own opening block brace on the same line. Other keywords and function names which are formatted with this "cuddled" style are elsif, continue, catch, finally.
Other block types can be formatted by specifying their names on a separate parameter --cuddled-block-list, described in a later section.
Cuddling between a pair of code blocks requires that the closing brace of the first block start a new line. If this block is entirely on one line in the input file, it is necessary to decide if it should be broken to allow cuddling. This decision is controlled by the flag --cuddled-break-options=n (-cbo=n) discussed below. The default and recommended value of -cbo=1 bases this decision on the first block in the chain. If it spans multiple lines then cuddling is made and continues along the chain, regardless of the sizes of subsequent blocks. Otherwise, short lines remain intact.
So for example, the --cuddled-else flag would not have any effect if the above snippet is rewritten as
if ($task) { yyy() }
else { zzz() }If the first block spans multiple lines, then cuddling can be done and will continue for the subsequent blocks in the chain, as illustrated in the previous snippet.
If there are blank lines between cuddled blocks they will be eliminated. If there are comments after the closing brace where cuddling would occur then cuddling will be prevented. If this occurs, cuddling will restart later in the chain if possible.
This flag is equivalent to --cuddled-else (-ce).
The built-in default cuddled block types are else, elsif, continue, catch, finally.
Additional block types to which the -cuddled-blocks style applies can be defined by this parameter. This parameter is a character string, giving a list of block types separated by commas or spaces. For example, to cuddle code blocks of type sort, map and grep, in addition to the default types, the string could be set to
-cbl="sort map grep"or equivalently
-cbl=sort,map,grepNote however that these particular block types are typically short so there might not be much opportunity for the cuddled format style.
Using commas avoids the need to protect spaces with quotes.
As a diagnostic check, the flag --dump-cuddled-block-list or -dcbl can be used to view the hash of values that are generated by this flag.
Finally, note that the --cuddled-block-list parameter by itself merely specifies which blocks are formatted with the cuddled format. It has no effect unless this formatting style is activated with --cuddled-else.
When cuddled else formatting is selected with -cuddled-else, setting this flag causes perltidy to ignore its built-in defaults and rely exclusively on the block types specified on the --cuddled-block-list flag described in the previous section. For example, to avoid using cuddled catch and finally, which are among the defaults, the following set of parameters could be used:
perltidy -ce -cbl='else elsif continue' -cblxCuddled formatting is only possible between a pair of code blocks if the closing brace of the first block starts a new line. If a block is encountered which is entirely on a single line, and cuddled formatting is selected, it is necessary to make a decision as to whether or not to "break" the block, meaning to cause it to span multiple lines. This parameter controls that decision. The options are:
cbo=0 Never force a short block to break.
cbo=1 If the first of a pair of blocks is broken in the input file,
then break the second [DEFAULT].
cbo=2 Break open all blocks for maximal cuddled formatting.The default and recommended value is cbo=1. With this value, if the starting block of a chain spans multiple lines, then a cascade of breaks will occur for remaining blocks causing the entire chain to be cuddled.
The option cbo=0 can produce erratic cuddling if there are numerous one-line blocks.
The option cbo=2 produces maximal cuddling but will not allow any short blocks.
Use the flag -bl to place an opening block brace on a new line:
if ( $input_file eq '-' )
{
...
}By default it applies to all structural blocks except sort map grep eval and anonymous subs.
The default is -nbl which places an opening brace on the same line as the keyword introducing it if possible. For example,
# default
if ( $input_file eq '-' ) {
...
}When -bl is set, the blocks to which this applies can be controlled with the parameters --brace-left-list and -brace-left-exclusion-list described in the next sections.
Use this parameter to change the types of block braces for which the -bl flag applies; see "Specifying Block Types". For example, -bll='if elsif else sub' would apply it to only if/elsif/else and named sub blocks. The default is all blocks, -bll='*'.
Use this parameter to exclude types of block braces for which the -bl flag applies; see "Specifying Block Types". For example, the default settings -bll='*' and -blxl='sort map grep eval asub' mean all blocks except sort map grep eval and anonymous sub blocks.
Note that the lists -bll=s and -blxl=s control the behavior of the -bl flag but have no effect unless the -bl flag is set. These two lists provide complete control for this flag, but two shortcut flags are available and described in the next sections.
The flag -sbl provides a shortcut way to turn on -bl just for named subs. The same effect can be achieved by turning on -bl with the block list set as -bll='sub'. To avoid conflicts, it is recommended to either use the more general list method described above to control -bl, or this shortcut method, but not both.
For example,
perltidy -sblproduces this result:
sub message
{
if (!defined($_[0])) {
print("Hello, World\n");
}
else {
print($_[0], "\n");
}
}This negative version of this flag, -nsbl, turns off -bl for named subs. The same effect can be achieved with the exclusion list method, -blxl=sub.
The flag -asbl is like the -sbl flag except that it applies to anonymous sub's instead of named subs. The same effect can be achieved by turning on -bl with the block list set to include -bll='asub'.
For example
perltidy -asblproduces this result:
$a = sub
{
if ( !defined( $_[0] ) ) {
print("Hello, World\n");
}
else {
print( $_[0], "\n" );
}
};This negative version of this flag, -nasbl, turns off -bl for anonymous subs.
The flag -bli is similar to the -bl flag but in addition it causes one unit of continuation indentation ( see --continuation-indentation ) to be placed before an opening and closing block braces.
For example
# perltidy -bli
if ( $input_file eq '-' )
{
important_function();
}By default, this extra indentation occurs for block types: if, elsif, else, unless, while, for, foreach, do, and also named subs and blocks preceded by a label. The next item shows how to change this.
Note: The -bli flag is similar to the -bl flag, with the difference being that braces get indented. But these two flags are implemented independently, and have different default settings for historical reasons. If desired, a mixture of effects can be achieved if desired by turning them both on with different -list settings. In the event that both settings are selected for a certain block type, the -bli style has priority.
Use this parameter to change the types of block braces for which the -bli flag applies; see "Specifying Block Types".
The default is -blil='if else elsif unless while for foreach do : sub'.
Use this parameter to exclude types of block braces for which the -bli flag applies; see "Specifying Block Types".
This might be useful in conjunction with selecting all blocks -blil='*'. The default setting is -blixl=' ', which does not exclude any blocks.
Note that the two parameters -blil and -blixl control the behavior of the -bli flag but have no effect unless the -bli flag is set.
The default style, -nbl places the opening code block brace on a new line if it does not fit on the same line as the opening keyword, like this:
if ( $bigwasteofspace1 && $bigwasteofspace2
|| $bigwasteofspace3 && $bigwasteofspace4 )
{
big_waste_of_time();
}To force the opening brace to always be on the right, use the -bar flag. In this case, the above example becomes
if ( $bigwasteofspace1 && $bigwasteofspace2
|| $bigwasteofspace3 && $bigwasteofspace4 ) {
big_waste_of_time();
}A conflict occurs if both --opening-brace-on_new-line (-bl) and -bar are specified.
A related parameter, --cuddled-paren-brace, causes perltidy to join two lines which otherwise would be
)
{to be
) {For example:
# default
foreach my $dir (
'05_lexer', '07_token', '08_regression', '11_util',
'13_data', '15_transform'
)
{
...
}
# perltidy -cpb
foreach my $dir (
'05_lexer', '07_token', '08_regression', '11_util',
'13_data', '15_transform'
) {
...;
}The -otr flag is a hint that perltidy should not place a break between a comma and an opening token. For example:
# default formatting
push @{ $self->{$module}{$key} },
{
accno => $ref->{accno},
description => $ref->{description}
};
# perltidy -otr
push @{ $self->{$module}{$key} }, {
accno => $ref->{accno},
description => $ref->{description}
};The flag -otr is actually an abbreviation for three other flags which can be used to control parens, hash braces, and square brackets separately if desired:
-opr or --opening-paren-right
-ohbr or --opening-hash-brace-right
-osbr or --opening-square-bracket-rightWhen a list of items spans multiple lines, the default formatting is to place the opening brace (or other container token) at the end of the starting line, like this:
$romanNumerals = {
one => 'I',
two => 'II',
three => 'III',
four => 'IV',
};This flag can change the default behavior to cause a line break to be placed before the opening brace according to the value given to the integer n:
-bbhb=0 never break [default]
-bbhb=1 stable: break if the input script had a break
-bbhb=2 break if list is 'complex' (see note below)
-bbhb=3 always breakFor example,
# perltidy -bbhb=3
$romanNumerals =
{
one => 'I',
two => 'II',
three => 'III',
four => 'IV',
};There are several points to note about this flag:
This parameter only applies if the opening brace is preceded by an '=' or '=>'.
This parameter only applies if the contents of the container looks like a list. The contents need to contain some commas or '=>'s at the next interior level to be considered a list.
For the n=2 option, a list is considered 'complex' if it is part of a nested list structure which spans multiple lines in the input file.
If multiple opening tokens have been 'welded' together with the -wn parameter, then this parameter has no effect.
The indentation of the braces will normally be one level of continuation indentation by default. This can be changed with the parameter -bbhbi=n in the next section.
Similar flags for controlling parens and square brackets are given in the subsequent section.
This flag is a companion to -bbhb=n for controlling the indentation of an opening hash brace which is placed on a new line by that parameter. The indentation is as follows:
-bbhbi=0 one continuation level [default]
-bbhbi=1 outdent by one continuation level
-bbhbi=2 indent one full indentation levelFor example:
# perltidy -bbhb=3 -bbhbi=1
$romanNumerals =
{
one => 'I',
two => 'II',
three => 'III',
four => 'IV',
};
# perltidy -bbhb=3 -bbhbi=2
$romanNumerals =
{
one => 'I',
two => 'II',
three => 'III',
four => 'IV',
};Note that this parameter has no effect unless -bbhb=n is also set.
This flag is similar to the flag described above, except it applies to lists contained within square brackets.
-bbsb=0 never break [default]
-bbsb=1 stable: break if the input script had a break
-bbsb=2 break if list is 'complex' (part of nested list structure)
-bbsb=3 always breakThis flag is a companion to -bbsb=n for controlling the indentation of an opening square bracket which is placed on a new line by that parameter. The indentation is as follows:
-bbsbi=0 one continuation level [default]
-bbsbi=1 outdent by one continuation level
-bbsbi=2 indent one full indentation levelThis flag is similar to -bbhb=n, described above, except it applies to lists contained within parens.
-bbp=0 never break [default]
-bbp=1 stable: break if the input script had a break
-bpb=2 break if list is 'complex' (part of nested list structure)
-bbp=3 always breakThis flag is a companion to -bbp=n for controlling the indentation of an opening paren which is placed on a new line by that parameter. The indentation is as follows:
-bbpi=0 one continuation level [default]
-bbpi=1 outdent by one continuation level
-bbpi=2 indent one full indentation levelSome types of closing block braces, such as eval, may be followed by additional code. A line break may be inserted between such a closing brace and the following code depending on the parameter n and the length of the trailing code, as follows:
If the trailing code fits on a single line, then
-bfvt=0 Follow the input style regarding break/no-break
-bfvt=1 Follow the input style regarding break/no-break [Default]
-bfvt=2 Do not insert a line breakIf the trailing code requires multiple lines, then
-bfvt=0 Insert a line break
-bfvt=1 Insert a line break except for a cuddled block chain [Default]
-bfvt=2 Do not insert a line breakSo the most compact code is achieved with -bfvt=2.
Example (non-cuddled, multiple lines ):
# -bfvt=0 or -bvft=1 [DEFAULT]
eval {
( $line, $cond ) = $self->_normalize_if_elif($line);
1;
}
or die sprintf "Error at line %d\nLine %d: %s\n%s",
( $line_info->start_line_num() ) x 2, $line, $@;
# -bfvt=2
eval {
( $line, $cond ) = $self->_normalize_if_elif($line);
1;
} or die sprintf "Error at line %d\nLine %d: %s\n%s",
( $line_info->start_line_num() ) x 2, $line, $@;Example (cuddled, multiple lines):
# -bfvt=0
eval {
#STUFF;
1; # return true
}
or do {
##handle error
};
# -bfvt=1 [DEFAULT] or -bfvt=2
eval {
#STUFF;
1; # return true
} or do {
##handle error
};The -wn flag causes closely nested pairs of opening and closing container symbols (curly braces, brackets, or parens) to be "welded" together, meaning that they are treated as if combined into a single unit, with the indentation of the innermost code reduced to be as if there were just a single container symbol.
For example:
# default formatting
do {
{
next if $x == $y;
}
} until $x++ > $z;
# perltidy -wn
do { {
next if $x == $y;
} } until $x++ > $z;When this flag is set perltidy makes a preliminary pass through the file and identifies all nested pairs of containers. To qualify as a nested pair, the closing container symbols must be immediately adjacent and the opening symbols must either (1) be adjacent as in the above example, or (2) have an anonymous sub declaration following an outer opening container symbol which is not a code block brace, or (3) have an outer opening paren separated from the inner opening symbol by any single non-container symbol or something that looks like a function evaluation, as illustrated in the next examples. An additional option (4) which can be turned on with the flag --weld-fat-comma is when the opening container symbols are separated by a hash key and fat comma (=>).
Any container symbol may serve as both the inner container of one pair and as the outer container of an adjacent pair. Consequently, any number of adjacent opening or closing symbols may join together in weld. For example, here are three levels of wrapped function calls:
# default formatting
my (@date_time) = Localtime(
Date_to_Time(
Add_Delta_DHMS(
$year, $month, $day, $hour, $minute, $second,
'0', $offset, '0', '0'
)
)
);
# perltidy -wn
my (@date_time) = Localtime( Date_to_Time( Add_Delta_DHMS(
$year, $month, $day, $hour, $minute, $second,
'0', $offset, '0', '0'
) ) );Notice how the indentation of the inner lines are reduced by two levels in this case. This example also shows the typical result of this formatting, namely it is a sandwich consisting of an initial opening layer, a central section of any complexity forming the "meat" of the sandwich, and a final closing layer. This predictable structure helps keep the compacted structure readable.
The inner sandwich layer is required to be at least one line thick. If this cannot be achieved, welding does not occur. This constraint can cause formatting to take a couple of iterations to stabilize when it is first applied to a script. The --converge flag can be used to insure that the final format is achieved in a single run.
Here is an example illustrating a welded container within a welded containers:
# default formatting
$x->badd(
bmul(
$class->new(
abs(
$sx * int( $xr->numify() ) & $sy * int( $yr->numify() )
)
),
$m
)
);
# perltidy -wn
$x->badd( bmul(
$class->new( abs(
$sx * int( $xr->numify() ) & $sy * int( $yr->numify() )
) ),
$m
) );The welded closing tokens are by default on a separate line but this can be modified with the --vertical-tightness-closing=n (-vtc=n) flag (described in the next section). For example, the same example adding -vtc=2 is
# perltidy -wn -vtc=2
$x->badd( bmul(
$class->new( abs(
$sx * int( $xr->numify() ) & $sy * int( $yr->numify() ) ) ),
$m ) );This format option is quite general but there are some limitations.
One limitation is that any line length limit still applies and can cause long welded sections to be broken into multiple lines.
Also, the stacking of containers defined by this flag have priority over any other container stacking flags. This is because any welding is done first.
When the -wfc flag is set, along with -wn (--weld-nested-containers), perltidy is allowed to weld an opening paren to an inner opening container when they are separated by a hash key and fat comma (=>). for example
# perltidy -wn -wfc
elf->call_method( method_name_foo => {
some_arg1 => $foo,
some_other_arg3 => $bar->{'baz'},
} );This option is off by default.
The -wnxl=s flag provides some control over the types of containers which can be welded. The -wn flag by default is "greedy" in welding adjacent containers. If it welds more types of containers than desired, this flag provides a capability to reduce the amount of welding by specifying a list of things which should not be welded.
The logic in perltidy to apply this is straightforward. As each container token is being considered for joining a weld, any exclusion rules are consulted and used to reject the weld if necessary.
This list is a string with space-separated items. Each item consists of up to three pieces of information: (1) an optional position, (2) an optional preceding type, and (3) a container type.
The only required piece of information is a container type, which is one of '(', '[', '{' or 'q'. The first three of these are container tokens and the last represents a quoted list. For example the string
-wnxl='[ { q'means do NOT include square-brackets, braces, or quotes in any welds. The only unspecified container is '(', so this string means that only welds involving parens will be made.
To illustrate, following welded snippet consists of a chain of three welded containers with types '(' '[' and 'q':
# perltidy -wn
skip_symbols( [ qw(
Perl_dump_fds
Perl_ErrorNo
Perl_GetVars
PL_sys_intern
) ] );Even though the qw term uses parens as the quote delimiter, it has a special type 'q' here. If it appears in a weld it always appears at the end of the welded chain.
Any of the container types '[', '{', and '(' may be prefixed with a position indicator which is either '^', to indicate the first token of a welded sequence, or '.', to indicate an interior token of a welded sequence. (Since a quoted string 'q' always ends a chain it does need a position indicator).
For example, if we do not want a sequence of welded containers to start with a square bracket we could use
-wnxl='^['In the above snippet, there is a square bracket but it does not start the chain, so the formatting would be unchanged if it were formatted with this restriction.
A third optional item of information which can be given is an alphanumeric letter which is used to limit the selection further depending on the type of token immediately before the container. If given, it goes just before the container symbol. The possible letters are currently 'k', 'K', 'f', 'F', 'w', and 'W', with these meanings:
'k' matches if the previous nonblank token is a perl built-in keyword (such as 'if', 'while'),
'K' matches if 'k' does not, meaning that the previous token is not a keyword.
'f' matches if the previous token is a function other than a keyword.
'F' matches if 'f' does not.
'w' matches if either 'k' or 'f' match.
'W' matches if 'w' does not.For example, compare
# perltidy -wn
if ( defined( $_Cgi_Query{
$Config{'methods'}{'authentication'}{'remote'}{'cgi'}{'username'}
} ) )with
# perltidy -wn -wnxl='^K( {'
if ( defined(
$_Cgi_Query{ $Config{'methods'}{'authentication'}{'remote'}{'cgi'}
{'username'} }
) )The first case does maximum welding. In the second case the leading paren is retained by the rule (it would have been rejected if preceded by a non-keyword) but the curly brace is rejected by the rule.
Here are some additional example strings and their meanings:
'^(' - the weld must not start with a paren
'.(' - the second and later tokens may not be parens
'.w(' - the second and later tokens may not keyword or function call parens
'(' - no parens in a weld
'^K(' - exclude a leading paren preceded by a non-keyword
'.k(' - exclude a secondary paren preceded by a keyword
'[ {' - exclude all brackets and braces
'[ ( ^K{' - exclude everything except nested structures like do {{ ... }}These parameters control what shall be called vertical tightness. Here are the main points:
Opening tokens (except for block braces) are controlled by -vt=n, or --vertical-tightness=n, where
-vt=0 always break a line after opening token (default).
-vt=1 do not break unless this would produce more than one
step in indentation in a line.
-vt=2 never break a line after opening tokenYou must also use the -lp (--line-up-parentheses) flag when you use the -vt flag; the reason is explained below.
Closing tokens (except for block braces) are controlled by -vtc=n, or --vertical-tightness-closing=n, where
-vtc=0 always break a line before a closing token (default),
-vtc=1 do not break before a closing token which is followed
by a semicolon or another closing token, and is not in
a list environment.
-vtc=2 never break before a closing token.
-vtc=3 Like -vtc=1 except always break before a closing token
if the corresponding opening token follows an = or =>.The rules for -vtc=1 and -vtc=3 are designed to maintain a reasonable balance between tightness and readability in complex lists.
Different controls may be applied to different token types, and it is also possible to control block braces; see below.
Finally, please note that these vertical tightness flags are merely hints to the formatter, and it cannot always follow them. Things which make it difficult or impossible include comments, blank lines, blocks of code within a list, and possibly the lack of the --line-up-parentheses parameter. Also, these flags may be ignored for very small lists (2 or 3 lines in length).
Here are some examples:
# perltidy -lp -vt=0 -vtc=0
%romanNumerals = (
one => 'I',
two => 'II',
three => 'III',
four => 'IV',
);
# perltidy -lp -vt=1 -vtc=0
%romanNumerals = ( one => 'I',
two => 'II',
three => 'III',
four => 'IV',
);
# perltidy -lp -vt=1 -vtc=1
%romanNumerals = ( one => 'I',
two => 'II',
three => 'III',
four => 'IV', );
# perltidy -vtc=3
my_function(
one => 'I',
two => 'II',
three => 'III',
four => 'IV', );
# perltidy -vtc=3
%romanNumerals = (
one => 'I',
two => 'II',
three => 'III',
four => 'IV',
);In the last example for -vtc=3, the opening paren is preceded by an equals so the closing paren is placed on a new line.
The difference between -vt=1 and -vt=2 is shown here:
# perltidy -lp -vt=1
$init->add(
mysprintf( "(void)find_threadsv(%s);",
cstring( $threadsv_names[ $op->targ ] )
)
);
# perltidy -lp -vt=2
$init->add( mysprintf( "(void)find_threadsv(%s);",
cstring( $threadsv_names[ $op->targ ] )
)
);With -vt=1, the line ending in add( does not combine with the next line because the next line is not balanced. This can help with readability, but -vt=2 can be used to ignore this rule.
The tightest, and least readable, code is produced with both -vt=2 and -vtc=2:
# perltidy -lp -vt=2 -vtc=2
$init->add( mysprintf( "(void)find_threadsv(%s);",
cstring( $threadsv_names[ $op->targ ] ) ) );Notice how the code in all of these examples collapses vertically as -vt increases, but the indentation remains unchanged. This is because perltidy implements the -vt parameter by first formatting as if -vt=0, and then simply overwriting one output line on top of the next, if possible, to achieve the desired vertical tightness. The -lp (--line-up-parentheses) indentation style has been designed to allow this vertical collapse to occur, which is why it is required for the -vt parameter.
The -vt=n and -vtc=n parameters apply to each type of container token. If desired, vertical tightness controls can be applied independently to each of the closing container token types.
The parameters for controlling parentheses are -pvt=n or --paren-vertical-tightness=n, and -pvtc=n or --paren-vertical-tightness-closing=n.
Likewise, the parameters for square brackets are -sbvt=n or --square-bracket-vertical-tightness=n, and -sbvtc=n or --square-bracket-vertical-tightness-closing=n.
Finally, the parameters for controlling non-code block braces are -bvt=n or --brace-vertical-tightness=n, and -bvtc=n or --brace-vertical-tightness-closing=n.
In fact, the parameter -vt=n is actually just an abbreviation for -pvt=n -bvt=n sbvt=n, and likewise -vtc=n is an abbreviation for -pvtc=n -bvtc=n -sbvtc=n.
The -bbvt=n flag is just like the -vt=n flag but applies to opening code block braces.
-bbvt=0 break after opening block brace (default).
-bbvt=1 do not break unless this would produce more than one
step in indentation in a line.
-bbvt=2 do not break after opening block brace.It is necessary to also use either -bl or -bli for this to work, because, as with other vertical tightness controls, it is implemented by simply overwriting a line ending with an opening block brace with the subsequent line. For example:
# perltidy -bli -bbvt=0
if ( open( FILE, "< $File" ) )
{
while ( $File = <FILE> )
{
$In .= $File;
$count++;
}
close(FILE);
}
# perltidy -bli -bbvt=1
if ( open( FILE, "< $File" ) )
{ while ( $File = <FILE> )
{ $In .= $File;
$count++;
}
close(FILE);
}By default this applies to blocks associated with keywords if, elsif, else, unless, for, foreach, sub, while, until, and also with a preceding label. This can be changed with the parameter -bbvtl=string, or --block-brace-vertical-tightness-list=string, where string is a space-separated list of block types. For more information on the possible values of this string, see "Specifying Block Types"
For example, if we want to just apply this style to if, elsif, and else blocks, we could use perltidy -bli -bbvt=1 -bbvtl='if elsif else'.
There is no vertical tightness control for closing block braces; with one exception they will be placed on separate lines. The exception is that a cascade of closing block braces may be stacked on a single line. See --stack-closing-block-brace.
The -sot flag tells perltidy to "stack" opening tokens when possible to avoid lines with isolated opening tokens.
For example:
# default
$opt_c = Text::CSV_XS->new(
{
binary => 1,
sep_char => $opt_c,
always_quote => 1,
}
);
# -sot
$opt_c = Text::CSV_XS->new( {
binary => 1,
sep_char => $opt_c,
always_quote => 1,
}
);For detailed control of individual closing tokens the following controls can be used:
-sop or --stack-opening-paren
-sohb or --stack-opening-hash-brace
-sosb or --stack-opening-square-bracket
-sobb or --stack-opening-block-braceThe flag -sot is an abbreviation for -sop -sohb -sosb.
The flag -sobb is an abbreviation for -bbvt=2 -bbvtl='*'. This will case a cascade of opening block braces to appear on a single line, although this an uncommon occurrence except in test scripts.
The -sct flag tells perltidy to "stack" closing tokens when possible to avoid lines with isolated closing tokens.
For example:
# default
$opt_c = Text::CSV_XS->new(
{
binary => 1,
sep_char => $opt_c,
always_quote => 1,
}
);
# -sct
$opt_c = Text::CSV_XS->new(
{
binary => 1,
sep_char => $opt_c,
always_quote => 1,
} );The -sct flag is somewhat similar to the -vtc (--vertical-tightness-closing flags, and in some cases it can give a similar result. The difference is that the -vtc flags try to avoid lines with leading opening tokens by "hiding" them at the end of a previous line, whereas the -sct flag merely tries to reduce the number of lines with isolated closing tokens by stacking them but does not try to hide them. For example:
# -vtc=2
$opt_c = Text::CSV_XS->new(
{
binary => 1,
sep_char => $opt_c,
always_quote => 1, } );For detailed control of the stacking of individual closing tokens the following controls can be used:
-scp or --stack-closing-paren
-schb or --stack-closing-hash-brace
-scsb or --stack-closing-square-bracket
-scbb or --stack-closing-block-braceThe flag -sct is an abbreviation for stacking the non-block closing tokens, -scp -schb -scsb.
Stacking of closing block braces, -scbb, causes a cascade of isolated closing block braces to be combined into a single line as in the following example:
# -scbb:
for $w1 (@w1) {
for $w2 (@w2) {
for $w3 (@w3) {
for $w4 (@w4) {
push( @lines, "$w1 $w2 $w3 $w4\n" );
} } } }To simplify input even further for the case in which both opening and closing non-block containers are stacked, the flag -sac or --stack-all-containers is an abbreviation for --stack-opening-tokens --stack-closing-tokens.
Please note that if both opening and closing tokens are to be stacked, then the newer flag -weld-nested-containers may be preferable because it insures that stacking is always done symmetrically. It also removes an extra level of unnecessary indentation within welded containers. It is able to do this because it works on formatting globally rather than locally, as the --stack-opening-tokens and --stack-closing-tokens flags do.
Four command line parameters provide some control over whether a line break should be before or after specific token types. Two parameters give detailed control:
-wba=s or --want-break-after=s, and
-wbb=s or --want-break-before=s.
These parameters are each followed by a quoted string, s, containing a list of token types (separated only by spaces). No more than one of each of these parameters should be specified, because repeating a command-line parameter always overwrites the previous one before perltidy ever sees it.
By default, perltidy breaks after these token types: % + - * / x != == >= <= =~ !~ < > | & = **= += *= &= <<= &&= -= /= |= >>= ||= //= .= %= ^= x=
And perltidy breaks before these token types by default: . << >> -> && || //
To illustrate, to cause a break after a concatenation operator, '.', rather than before it, the command line would be
-wba="."As another example, the following command would cause a break before math operators '+', '-', '/', and '*':
-wbb="+ - / *"These commands should work well for most of the token types that perltidy uses (use --dump-token-types for a list). Also try the -D flag on a short snippet of code and look at the .DEBUG file to see the tokenization. However, for a few token types there may be conflicts with hardwired logic which cause unexpected results. One example is curly braces, which should be controlled with the parameter --opening-brace-on-new-line provided for that purpose.
WARNING Be sure to put these tokens in quotes to avoid having them misinterpreted by your command shell.
Two additional parameters are available which, though they provide no further capability, can simplify input are:
-baao or --break-after-all-operators,
-bbao or --break-before-all-operators.
The -baao sets the default to be to break after all of the following operators:
% + - * / x != == >= <= =~ !~ < > | &
= **= += *= &= <<= &&= -= /= |= >>= ||= //= .= %= ^= x=
. : ? && || and or err xorand the -bbao flag sets the default to break before all of these operators. These can be used to define an initial break preference which can be fine-tuned with the --want-break-after and --want-break-before flags. For example, to break before all operators except an = one could use --bbao -wba='=' rather than listing every single perl operator except = on a -wbb flag.
This flag controls whether or not a line break occurs after a label. There are three possible values for n:
-bal=0 break if there is a break in the input [DEFAULT]
-bal=1 always break after a label
-bal=2 never break after a labelFor example,
# perltidy -bal=1
RETURN:
return;
# perltidy -bal=2
RETURN: return;Perltidy attempts to format lists of comma-separated values in tables which look good. Its default algorithms usually work well, but sometimes they don't. In this case, there are several methods available to control list formatting.
A very simple way to prevent perltidy from changing the line breaks within a comma-separated list of values is to insert a blank line, comment, or side-comment anywhere between the opening and closing parens (or braces or brackets). This causes perltidy to skip over its list formatting logic. (The reason is that any of these items put a constraint on line breaks, and perltidy needs complete control over line breaks within a container to adjust a list layout). For example, let us consider
my @list = (1,
1, 1,
1, 2, 1,
1, 3, 3, 1,
1, 4, 6, 4, 1,);The default formatting, which allows a maximum line length of 80, will flatten this down to one line:
# perltidy (default)
my @list = ( 1, 1, 1, 1, 2, 1, 1, 3, 3, 1, 1, 4, 6, 4, 1, );This formatting loses the nice structure. The original line breaks can be retained by adding comment or a blank line somewhere between the two parens. For example,
my @list = (
1, # a side comment forces the original line breakpoints to be kept
1, 1,
1, 2, 1,
1, 3, 3, 1,
1, 4, 6, 4, 1,
);We could achieve the same result with a blank line or full comment anywhere between the opening and closing parens. Vertical alignment of the list items will still occur if possible. The blank line method is shown here:
my @list = (
1,
1, 1,
1, 2, 1,
1, 3, 3, 1,
1, 4, 6, 4, 1,
);For another possibility see the -fs flag in "Skipping Selected Sections of Code".
The -boc flag is another way to prevent comma-separated lists from being reformatted. Using -boc on the above example, plus additional flags to retain the original style, yields
# perltidy -boc -lp -pt=2 -vt=1 -vtc=1
my @list = (1,
1, 1,
1, 2, 1,
1, 3, 3, 1,
1, 4, 6, 4, 1,);A disadvantage of this flag compared to the methods discussed above is that all tables in the file must already be nicely formatted.
If n is a positive number, and the computed number of fields for any table exceeds n, then it will be reduced to n. This parameter might be used on a small section of code to force a list to have a particular number of fields per line, and then either the -boc flag could be used to retain this formatting, or a blank line or comment could be introduced somewhere to freeze the formatting in future applications of perltidy. For example
# perltidy -mft=2
@month_of_year = (
'Jan', 'Feb',
'Mar', 'Apr',
'May', 'Jun',
'Jul', 'Aug',
'Sep', 'Oct',
'Nov', 'Dec'
);The default value is n=0, which does not place a limit on the number of fields in a table.
A comma which follows a comma arrow, '=>', is given special consideration. In a long list, it is common to break at all such commas. This parameter can be used to control how perltidy breaks at these commas. (However, it will have no effect if old comma breaks are being forced because -boc is used). The possible values of n are:
n=0 break at all commas after =>
n=1 stable: break at all commas after => if container is open,
EXCEPT FOR one-line containers
n=2 break at all commas after =>, BUT try to form the maximum
one-line container lengths
n=3 do not treat commas after => specially at all
n=4 break everything: like n=0 but ALSO break a short container with
a => not followed by a comma when -vt=0 is used
n=5 stable: like n=1 but ALSO break at open one-line containers when
-vt=0 is used (default)For example, given the following single line, perltidy by default will not add any line breaks because it would break the existing one-line container:
bless { B => $B, Root => $Root } => $package;Using -cab=0 will force a break after each comma-arrow item:
# perltidy -cab=0:
bless {
B => $B,
Root => $Root
} => $package;If perltidy is subsequently run with this container broken, then by default it will break after each '=>' because the container is now broken. To reform a one-line container, the parameter -cab=2 could be used.
The flag -cab=3 can be used to prevent these commas from being treated specially. In this case, an item such as "01" => 31 is treated as a single item in a table. The number of fields in this table will be determined by the same rules that are used for any other table. Here is an example.
# perltidy -cab=3
my %last_day = (
"01" => 31, "02" => 29, "03" => 31, "04" => 30,
"05" => 31, "06" => 30, "07" => 31, "08" => 31,
"09" => 30, "10" => 31, "11" => 30, "12" => 31
);This option causes repeated commas to be removed. For example, given this list with a repeated comma
ignoreSpec( $file, "file",, \%spec, \%Rspec );we can remove it with -drc
# perltidy -drc:
ignoreSpec( $file, "file", \%spec, \%Rspec );This parameter also deletes repeated fat commas, '=>'. The complete list of actions taken when this flag is set are as follows:
Repeated commas like ',,' are removed with a warning.
Repeated fat commas like '=> =>' are removed with a warning.
The combination '=>,' produces a warning but is not changed (it is likely an error but only its author would know how to fix it).
The remaining combination ',=>' (sometimes called a 'winking fat comma') is ignored by this parameter.
These warnings are only output if the --warning-output, or -w, flag is set.
This option is on by default. Use -ndrc to turn it off.
A trailing comma is a comma following the last item of a list. Perl allows trailing commas but they are not required. Including them can sometimes simplify the maintenance of large or complex lists. By default, perltidy does not add or delete trailing commas, but it is possible to manipulate them with the following set of related parameters:
--want-trailing-commas=s, -wtc=s - defines where trailing commas are wanted
--add-trailing-commas, -atc - gives permission to add trailing commas to match the style wanted
--delete-trailing-commas, -dtc - gives permission to delete trailing commas which do not match the style wanted
--add-lone-trailing-commas, -altc - gives permission to add a comma if it will be the only comma. This is off by default and explained below.
--delete-lone-trailing-commas, -dltc - gives permission to delete the only comma in a list. This is on by default and explained below.
The parameter --want-trailing-commas=s, or -wtc=s, defines a preferred style. The string s indicates which lists should get trailing commas, as follows:
s=0 : no list should have a trailing comma
s=1 or * : every list should have a trailing comma
s=m a multi-line list should have a trailing commas
s=b trailing commas should be 'bare' (comma followed by newline)
s=h lists of key=>value pairs, with about one one '=>' and one ',' per line,
with a bare trailing comma
s=i lists with about one comma per line, with a bare trailing comma
s=' ' or -wtc not defined : leave trailing commas unchanged [DEFAULT].This parameter by itself only indicates the where trailing commas are wanted. Perltidy only adds these trailing commas if the flag --add-trailing-commas, or -atc is set. And perltidy only removes unwanted trailing commas if the flag --delete-trailing-commas, or -dtc is set.
Here are some example parameter combinations and their meanings
-wtc=0 -dtc : delete all trailing commas
-wtc=1 -atc : all lists get trailing commas
-wtc=m -atc : all multi-line lists get trailing commas, but
single line lists remain unchanged.
-wtc=m -dtc : multi-line lists remain unchanged, but
any trailing commas on single line lists are removed.
-wtc=m -atc -dtc : all multi-line lists get trailing commas, and
any trailing commas on single line lists are removed.For example, given the following input without a trailing comma
bless {
B => $B,
Root => $Root
} => $package;we can add a trailing comma after the variable $Root using
# perltidy -wtc=m -atc
bless {
B => $B,
Root => $Root,
} => $package;This could also be achieved in this case with -wtc=b instead of -wtc=m because the trailing comma here is bare (separated from its closing brace by a newline). And it could also be achieved with -wtc=h because this particular list is a list of key=>value pairs.
The above styles should cover the main of situations of interest, but it is possible to apply a different style to each type of container token by including an opening token ahead of the style character in the above table. For example
-wtc='(m [b'means that lists within parens should have multi-line trailing commas, and that lists within square brackets have bare trailing commas. Since there is no specification for curly braces in this example, their trailing commas would remain unchanged.
For parentheses, an additional item of information which can be given is an alphanumeric letter which is used to limit the selection further depending on the type of token immediately before the opening paren. The possible letters are currently 'k', 'K', 'f', 'F', 'w', and 'W', with these meanings for matching whatever precedes an opening paren:
'k' matches if the previous nonblank token is a perl built-in keyword
(such as 'if', 'while'),
'K' matches if 'k' does not, meaning that the previous token is not a keyword.
'f' matches if the previous token is a function other than a keyword.
'F' matches if 'f' does not.
'w' matches if either 'k' or 'f' match.
'W' matches if 'w' does not.These are the same codes used for --line-up-parentheses-inclusion-list. For example,
-wtc = 'w(m'means that trailing commas are wanted for multi-line parenthesized lists following a function call or keyword.
Some points to note regarding adding and deleting trailing commas:
For the implementation of these parameters, a list is basically taken to be a container of items (parens, square brackets, or braces), which is not a code block, with one or more commas or fat commas. These parameters only apply to something that fits this definition of a list.
Note that a paren-less list of parameters is not a list by this definition, so these parameters have no effect on a paren-less list.
By multiline list is meant a list for which the first comma and trailing comma are on different lines.
A bare trailing comma is a comma which is at the end of a line. That is, the closing container token follows on a different line. So a list with a bare trailing comma is a special case of a multi-line list.
The decision regarding whether or not a list is multi-line or bare is made based on the input stream if only one iteration is made, which is the default.
It can sometimes be preferable to base decisions on trailing commas on the final line breaks rather than the initial line breaks. This can be accomplished by telling perltidy to perform two or more internal iterations, for example with the --converge parameter. In this case any comma addition or deletion operations are postponed until the start of the second iteration, after changes in line breaks have been made. For a discussion see https://github.com/perltidy/perltidy/issues/156.
A parameter --delay-trailing-comma-operations, or -dtco, is available to control behavior if desired. Negating this parameter, with -ndtco, tells perltidy to always use the starting state to make decisions regarding comma addition and deletion, even when iterations are requested.
When using these parameters for the first time it is a good idea to practice on some test scripts and verify that the results are as expected.
Since the default behavior is not to add or delete commas, these parameters can be useful on a temporary basis for reformatting a script.
Special Considerations for Lone Trailing Commas
Adding or deleting the only comma in a list can have some implications which need to be explained and possibly controlled.
The main issue with deleting a lone comma is that if it is deleted, then it might not be possible add it back automatically since perltidy uses the existance of commas to help locate lists. For example, given
my ( $self, ) = @_;and if we remove the comma with
# perltidy -wtc=m -dtc
my ( $self ) = @_;then we cannot use the trailing comma controls to add this comma back. The parameter --delete-lone-trailing-commas allows such a comma to be deleted, and is on by default, but can be turned off to prevent this. This might be useful if one is experimenting with formatting options and wants to restrict testing to operations which are reversible. Note that this parameter is a fine-tuning control for --delete-trailing-commas which must also be set for it to have any effect.
Perltidy can add a lone comma to certain lists which themselves contain lists. For example, a comma can be added after the closing hash brace in the following snippet:
$self->make_grammar(
{
iterator => $self->_iterator,
parser => $self,
version => $self->version,
}
);However, the parameter --add-lone-trailing-commas must be set to allow such a comma to be added. The reason is that adding such a comma would prevent the --weld-nested-containers flag from operating on this structure. This can be confusing, so this parameter is off by default. Note that this parameter is a fine-tuning control for --add-trailing-commas which must also be set for it to have any effect. To illustrate its use on the above example:
# perltidy -atc -altc -wtc=b
$self->make_grammar(
{
iterator => $self->_iterator,
parser => $self,
version => $self->version,
},
);This comma can be removed to allow welding with the control described in the next section.
If the closing tokens of two nested containers are separated by a comma, then welding requested with --weld-nested-containers cannot occur. Any commas in this situation are optional trailing commas and can be removed with -dwic. For example, a comma in this script prevents welding:
# perltidy -wn
$self->make_grammar(
{
iterator => $self->_iterator,
parser => $self,
version => $self->version,
},
);Adding -dwic removes the comma and allows welding:
# perltidy -wn -dwic
$self->make_grammar( {
iterator => $self->_iterator,
parser => $self,
version => $self->version,
} );This feature is off by default. Here are some points to note about the -dwic parameter
This operation is not always reversible, so please check results of using this parameter carefully.
Removing this type of isolated trailing comma is necessary for welding to be possible, but not sufficient. So welding will not always occur where these commas are removed.
This operation is independent of --add-trailing-commas and --delete-trailing-commas. If it conflicts with any of those settings, it has priority.
In the following expression, the arrow operator '->' between the closing and opening brackets of hash keys and array indexes are optional:
return $self->{'commandline'}->{'arg_list'}->[0]->[0]->{'hostgroups'};These will be called interbracket arrows here, for lack of a better term. Perltidy will not change them by default, but they can be added or removed with the following parameters.
This parameter deletes interbracket arrows. Applied to the above example we have
# perltidy -dia
return $self->{'commandline'}{'arg_list'}[0][0]{'hostgroups'};By default this applies to all interbracket arrows, but selective deletion is possible with controls described below.
This parameter adds interbracket arrows. Applied to the line of code above, we get back the original line.
# perltidy -aia
return $self->{'commandline'}->{'arg_list'}->[0]->[0]->{'hostgroups'};Selective changes can be made with controls described below.
By default the -add- and -delete- parameters apply to all interbracket arrows.
An optional style can be specified with this parameter string s. In that case the parameters --add-interbracket-arrows and --delete-interbracket-arrows only apply where they would bring the formatting into agreement with the specified style. They may both be used in a single run if a mixed style is specified since there is no conflict.
The style string s gives a graphical description of the desired style. It lists up to four possible pairs of bracket types with an optional arrow between them. For example:
-ias='][ }->{'This means no arrows are wanted between '][' but arrows should be between '}{'. And it means that the unlisted pairs ']{' and '}[' should remain unchanged, either with or without arrows.
In this particular example, if the parameter --delete-interbracket-arrows is used, then only arrows like ']->[' will be deleted, since they are the only ones which disagree with the style.
And likewise, if --add-interbracket-arrows is used, then arrows will only be inserted between brackets like '}{' to bring the formatting into conformity with the style in this example.
Spaces in the string s are optional. They are ignored when the expression is parsed.
The style corresponding to all possible arrows is
-ias=']->[ ]->{ }->[ }->{'For convenience, this may also be requested with -ias=1 or -ias='*'.
The style corresponding to no interbracket arrows is
-ias='] [ ] { } [ } {'which may also be requested with -ias=0.
If this parameter is set, then a message is written to the error file in the following cases:
If an arrow is added or deleted by an add or delete command.
If a style is defined and an arrow would have been added or deleted if requested. So for example, the command
perltidy -wia -ias=']['will show where a file has arrows like ]->[' since they do not match the style, but no changes will be made because the delete command -dia has not been given. And
perltidy -wia -ias=0will warn if any arrows exist, since the flag -ias=0 means that no arrows are wanted.
This parameter makes it possible to skip adding or deleting arrows following a container which is complex in some sense. Three levels of complexity can be specified with the integer n, as follows:
n=0 the contents of the left container must be a single thing (token)
n=1 the left container must not contain other containers [DEFAULT]
n=2 the left container may contain anythingSome examples:
# Container complexity
{'commandline'} 0 single token OK by default
{ $type . $name } 1 multiple tokens OK by default
[ $plot{'x-axis'} - 1 ] 2 contains a container SKIPPED by defaultSo, with the default complexity level of 1, an arrow could be added or deleted following the first two of these containers but not the third.
Some points to consider when working with these parameters are:
There are no known bugs, but this is a relatively new feature. So please carefully check file differences and run tests when interbracket arrows are added or removed.
For some unusual spacing parameters, it could take an extra iteration for the spaces between brackets to reach their final state after arrows are added or deleted.
Any comments between brackets will prevent the adding and deleting of arrows.
A defensive programming technique is to require that every if-elsif- chain be terminated with an else block, even though it is not strictly required. This helps insure that there are no holes in the logic.
For example, consider the following snippet:
my $type = get_cards();
if ( $type = 1 ) { action("hold 'em") }
elsif ( $type = 2 ) { action("fold 'em") }
elsif ( $type = 3 ) { action("walk away") }What if the variable $type is some other value? It might have been obvious that this was okay when the code was first written, but it might not be so clear when the code is reviewed a few years later. A terminal else block with a comment would help clarify things.
The parameters in this section can help by either issuing a warning if an else is missing, or even inserting an empty else block where one is missing, or both.
This flag tells perltidy to issue a warning if a program is missing a terminal else block. The default is not to issue such a warning.
This flag tells perltidy to output an empty else block wherever a program is missing a terminal else block. To get a warning when this is done you should also set -wme. The default is not to add missing else blocks.
This string is a side comment which will be written to highlight a new empty else block. The default is:
-amec='##FIXME - added with perltidy -ame'For example, on the above example we can add a missing else and also get a warning notice with:
# perltidy -ame -wme
my $type = get_cards();
if ( $type == 1 ) { action("hold 'em") }
elsif ( $type == 2 ) { action("fold 'em") }
elsif ( $type == 3 ) { action("walk away") }
else {
##FIXME - added with perltidy -ame
}Any ##FIXME comments created in this way should be reviewed and changed appropriately. For example, one might decide that the code fine as is, and just change the comment to indicate that nothing has been overlooked:
my $type = get_cards();
if ( $type == 1 ) { action("hold 'em") }
elsif ( $type == 2 ) { action("fold 'em") }
elsif ( $type == 3 ) { action("walk away") }
else {
# ok - no worries
}Or maybe a deeper analysis reveals that something was missed:
my $type = get_cards();
if ( $type == 1 ) { action("hold 'em") }
elsif ( $type == 2 ) { action("fold 'em") }
elsif ( $type == 3 ) { action("walk away") }
else { action("run") }Sometimes it turns out that the else block should not reachable, in which case an error exit might be appropriate. In any case, having the else block can improve code maintainability.
Several additional parameters are available for controlling the extent to which line breaks in the input script influence the output script. In most cases, the default parameter values are set so that, if a choice is possible, the output style follows the input style. For example, if a short logical container is broken in the input script, then the default behavior is for it to remain broken in the output script.
Most of the parameters in this section would only be required for a one-time conversion of a script from short container lengths to longer container lengths. The opposite effect, of converting long container lengths to shorter lengths, can be obtained by temporarily using a short maximum line length.
By default, if a logical expression is broken at a &&, ||, and, or or, then the container will remain broken. Also, breaks at internal keywords if and unless will normally be retained. To prevent this, and thus form longer lines, use -nbol.
Please note that this flag does not duplicate old logical breakpoints. They are merely used as a hint with this flag that a statement should remain broken. Without this flag, perltidy will normally try to combine relatively short expressions into a single line.
For example, given this snippet:
return unless $cmd = $cmd || ($dot
&& $Last_Shell) || &prompt('|');
# perltidy -bol [default]
return
unless $cmd = $cmd
|| ( $dot
&& $Last_Shell )
|| &prompt('|');
# perltidy -nbol
return unless $cmd = $cmd || ( $dot && $Last_Shell ) || &prompt('|');By default, a method call arrow -> is considered a candidate for a breakpoint, but method chains will fill to the line width before a break is considered. With -bom, breaks before the arrow are preserved, so if you have pre-formatted a method chain:
my $q = $rs
->related_resultset('CDs')
->related_resultset('Tracks')
->search({
'track.id' => {-ident => 'none_search.id'},
})->as_query;It will keep these breaks, rather than become this:
my $q = $rs->related_resultset('CDs')->related_resultset('Tracks')->search({
'track.id' => {-ident => 'none_search.id'},
})->as_query;This flag will also look for and keep a 'cuddled' style of calls, in which lines begin with a closing paren followed by a call arrow, as in this example:
# perltidy -bom -wn
my $q = $rs->related_resultset(
'CDs'
)->related_resultset(
'Tracks'
)->search( {
'track.id' => { -ident => 'none_search.id' },
} )->as_query;You may want to include the -weld-nested-containers flag in this case to keep nested braces and parens together, as in the last line.
Semicolons are normally placed at the end of a statement. This means that formatted lines do not normally begin with semicolons. If the input stream has some lines which begin with semicolons, these can be retained by setting this flag. For example, consider the following two-line input snippet:
$z = sqrt($x**2 + $y**2)
;The default formatting will be:
$z = sqrt( $x**2 + $y**2 );The result using perltidy -bos keeps the isolated semicolon:
$z = sqrt( $x**2 + $y**2 )
;The default is not to do this, -nbos.
By default, perltidy will retain a breakpoint before keywords which may return lists, such as sort and map. This allows chains of these operators to be displayed one per line. Use -nbok to prevent retaining these breakpoints.
By default, if a conditional (ternary) operator is broken at a :, then it will remain broken. To prevent this, and thereby form longer lines, use -nbot.
By default, if an attribute list is broken at a : in the source file, then it will remain broken. For example, given the following code, the line breaks at the ':'s will be retained:
my @field
: field
: Default(1)
: Get('Name' => 'foo') : Set('Name');If the attributes are on a single line in the source code then they will remain on a single line if possible.
To prevent this, and thereby always form longer lines, use -nboa.
It is possible to override the choice of line breaks made by perltidy, and force it to follow certain line breaks in the input stream, with these two parameters:
-kbb=s or --keep-old-breakpoints-before=s, and
-kba=s or --keep-old-breakpoints-after=s
These parameters are each followed by a quoted string, s, containing a list of token types (separated only by spaces). No more than one of each of these parameters should be specified, because repeating a command-line parameter always overwrites the previous one before perltidy ever sees it.
For example, -kbb='=>' means that if an input line begins with a '=>' then the output script should also have a line break before that token.
For example, given the script:
method 'foo'
=> [ Int, Int ]
=> sub {
my ( $self, $x, $y ) = ( shift, @_ );
...;
};
# perltidy [default]
method 'foo' => [ Int, Int ] => sub {
my ( $self, $x, $y ) = ( shift, @_ );
...;
};
# perltidy -kbb='=>'
method 'foo'
=> [ Int, Int ]
=> sub {
my ( $self, $x, $y ) = ( shift, @_ );
...;
};For the container tokens '{', '[' and '(' and, their closing counterparts, use the token symbol. Thus, the command to keep a break after all opening parens is:
perltidy -kba='('It is possible to be more specific in matching parentheses by preceding them with a letter. The possible letters are 'k', 'K', 'f', 'F', 'w', and 'W', with these meanings (these are the same as used in the --weld-nested-exclusion-list and --line-up-parentheses-exclusion-list parameters):
'k' matches if the previous nonblank token is a perl built-in keyword
(such as 'if', 'while'),
'K' matches if 'k' does not, meaning that the previous token is not a keyword.
'f' matches if the previous token is a function other than a keyword.
'F' matches if 'f' does not.
'w' matches if either 'k' or 'f' match.
'W' matches if 'w' does not.So for example the the following parameter will keep breaks after opening function call parens:
perltidy -kba='f('NOTE: A request to break before an opening container, such as -kbb='(', will be silently ignored because it can lead to formatting instability. Likewise, a request to break after a closing container, such as -kba=')', will also be silently ignored.
Use this flag to tell perltidy to ignore existing line breaks to the maximum extent possible. This will tend to produce the longest possible containers, regardless of type, which do not exceed the line length limit. But please note that this parameter has priority over all other parameters requesting that certain old breakpoints be kept.
To illustrate, consider the following input text:
has subcmds => (
is => 'ro',
default => sub { [] },
);The default formatting will keep the container broken, giving
# perltidy [default]
has subcmds => (
is => 'ro',
default => sub { [] },
);If old breakpoints are ignored, the list will be flattened:
# perltidy -iob
has subcmds => ( is => 'ro', default => sub { [] }, );Besides flattening lists, this parameter also applies to lines broken at certain logical breakpoints such as 'if' and 'or'.
Even if this is parameter is not used globally, it provides a convenient way to flatten selected lists from within an editor.
Use the -kis flag to prevent breaking at a semicolon if there was no break there in the input file. Normally perltidy places a newline after each semicolon which terminates a statement unless several statements are contained within a one-line brace block. To illustrate, consider the following input lines:
dbmclose(%verb_delim); undef %verb_delim;
dbmclose(%expanded); undef %expanded;The default is to break after each statement, giving
dbmclose(%verb_delim);
undef %verb_delim;
dbmclose(%expanded);
undef %expanded;With perltidy -kis the multiple statements are retained:
dbmclose(%verb_delim); undef %verb_delim;
dbmclose(%expanded); undef %expanded;The statements are still subject to the specified value of maximum-line-length and will be broken if this maximum is exceeded.
Blank lines can improve the readability of a script if they are carefully placed. Perltidy has several commands for controlling the insertion, retention, and removal of blank lines.
Set -fbl if you want to the blank lines in your script to remain exactly as they are. The rest of the parameters in this section may then be ignored. (Note: setting the -fbl flag is equivalent to setting -mbl=0 and -kbl=2).
A blank line will be introduced before a full-line comment. This is the default. Use -nbbc or --noblanks-before-comments to prevent such blank lines from being introduced.
The parameter -blbs=n requests that least n blank lines precede a sub definition which does not follow a comment and which is more than one-line long. The default is <-blbs=1>. BEGIN and END blocks are included.
The requested number of blanks statement will be inserted regardless of the value of --maximum-consecutive-blank-lines=n (-mbl=n) with the exception that if -mbl=0 then no blanks will be output.
This parameter interacts with the value k of the parameter --maximum-consecutive-blank-lines=k (-mbl=k) as follows:
1. If -mbl=0 then no blanks will be output. This allows all blanks to be suppressed with a single parameter. Otherwise,
2. If the number of old blank lines in the script is less than n then additional blanks will be inserted to make the total n regardless of the value of -mbl=k.
3. If the number of old blank lines in the script equals or exceeds n then this parameter has no effect, however the total will not exceed value specified on the -mbl=k flag.
The parameter -blbp=n requests that least n blank lines precede a package which does not follow a comment. The default is -blbp=1.
This parameter interacts with the value k of the parameter --maximum-consecutive-blank-lines=k (-mbl=k) in the same way as described for the previous item -blbs=n.
For compatibility with previous versions, -bbs or --blanks-before-subs is equivalent to -blbp=1 and -blbs=1.
Likewise, -nbbs or --noblanks-before-subs is equivalent to -blbp=0 and -blbs=0.
A blank line will be introduced before blocks of coding delimited by for, foreach, while, until, and if, unless, in the following circumstances:
The block is not preceded by a comment.
The block is not a one-line block.
The number of consecutive non-blank lines at the current indentation depth is at least -lbl (see next section).
This is the default. The intention of this option is to introduce some space within dense coding. This is negated with -nbbb or --noblanks-before-blocks.
This controls how often perltidy is allowed to add blank lines before certain block types (see previous section). The default is 8. Entering a value of 0 is equivalent to entering a very large number.
This control places a minimum of i blank lines after a line which ends with an opening block brace of a specified type. By default, this only applies to the block of a named sub, but this can be changed (see -blaol below). The default is not to do this (i=0).
Please see the note below on using the -blao and -blbc options.
This control places a minimum of i blank lines before a line which begins with a closing block brace of a specified type. By default, this only applies to the block of a named sub, but this can be changed (see -blbcl below). The default is not to do this (i=0).
The parameter s is a list of block type keywords to which the flag -blao should apply. The section "Specifying Block Types" explains how to list block types.
This parameter is a list of block type keywords to which the flag -blbc should apply. The section "Specifying Block Types" explains how to list block types.
These blank line controls introduce a certain minimum number of blank lines in the text, but the final number of blank lines may be greater, depending on values of the other blank line controls and the number of old blank lines. A consequence is that introducing blank lines with these and other controls cannot be exactly undone, so some experimentation with these controls is recommended before using them.
For example, suppose that for some reason we decide to introduce one blank space at the beginning and ending of all blocks. We could do this using
perltidy -blao=2 -blbc=2 -blaol='*' -blbcl='*' filenameNow suppose the script continues to be developed, but at some later date we decide we don't want these spaces after all. We might expect that running with the flags -blao=0 and -blbc=0 will undo them. However, by default perltidy retains single blank lines, so the blank lines remain.
We can easily fix this by telling perltidy to ignore old blank lines by including the added parameter --keep-old-blank-lines=0 and rerunning. Then the unwanted blank lines will be gone. However, this will cause all old blank lines to be ignored, perhaps even some that were added by hand to improve formatting. So please be cautious when using these parameters.
This parameter specifies the maximum number of consecutive blank lines which will be output within code sections of a script. The default is n=1. If the input file has more than n consecutive blank lines, the number will be reduced to n except as noted above for the --blank-lines-before-subs and --blank-lines-before-subs parameters. If n=0 then no blank lines will be output (unless all old blank lines are retained with the --keep-old-blank-lines=2 flag of the next section).
This flag obviously does not apply to pod sections, here-documents, and quotes.
The -kbl=n flag gives you control over how your existing blank lines are treated.
The possible values of n are:
n=0 ignore all old blank lines
n=1 stable: keep old blanks, but limited by the value of the B<-mbl=n> flag
n=2 keep all old blank lines, regardless of the value of the B<-mbl=n> flagThe default is n=1.
This is equivalent to --keep-old-blank-lines=0 and is included for compatibility with previous versions.
This is equivalent to --keep-old-blank-lines=1 and is included for compatibility with previous versions.
Controls for blank lines around lines of consecutive keywords
The parameters in this section provide some control over the placement of blank lines within and around groups of statements beginning with selected keywords. These blank lines are called here keyword group blanks, and all of the parameters begin with --keyword-group-blanks*, or -kgb* for short. The default settings do not employ these controls but they can be enabled with the following parameters:
-kgbl=s or --keyword-group-blanks-list=s; s is a quoted string of keywords
-kgbs=s or --keyword-group-blanks-size=s; s gives the number of keywords required to form a group.
-kgbb=n or --keyword-group-blanks-before=n; n = (0, 1, or 2) controls a leading blank
-kgba=n or --keyword-group-blanks-after=n; n = (0, 1, or 2) controls a trailing blank
-kgbi or --keyword-group-blanks-inside is a switch for adding blanks between subgroups
-kgbd or --keyword-group-blanks-delete is a switch for removing initial blank lines between keywords
-kgbr=n or --keyword-group-blanks-repeat-count=n can limit the number of times this logic is applied
In addition, the following abbreviations are available to for simplified usage:
-kgb or --keyword-group-blanks is short for -kgbb=2 -kgba=2 kgbi
-nkgb or --nokeyword-group-blanks, is short for -kgbb=1 -kgba=1 nkgbi
Before describing the meaning of the parameters in detail let us look at an example which is formatted with default parameter settings.
print "Entering test 2\n";
use Test;
use Encode qw(from_to encode decode
encode_utf8 decode_utf8
find_encoding is_utf8);
use charnames qw(greek);
my @encodings = grep( /iso-?8859/, Encode::encodings() );
my @character_set = ( '0' .. '9', 'A' .. 'Z', 'a' .. 'z' );
my @source = qw(ascii iso8859-1 cp1250);
my @destiny = qw(cp1047 cp37 posix-bc);
my @ebcdic_sets = qw(cp1047 cp37 posix-bc);
my $str = join( '', map( chr($_), 0x20 .. 0x7E ) );
return unless ($str);using perltidy -kgb gives:
print "Entering test 2\n";
<----------this blank controlled by -kgbb
use Test;
use Encode qw(from_to encode decode
encode_utf8 decode_utf8
find_encoding is_utf8);
use charnames qw(greek);
<---------this blank controlled by -kgbi
my @encodings = grep( /iso-?8859/, Encode::encodings() );
my @character_set = ( '0' .. '9', 'A' .. 'Z', 'a' .. 'z' );
my @source = qw(ascii iso8859-1 cp1250);
my @destiny = qw(cp1047 cp37 posix-bc);
my @ebcdic_sets = qw(cp1047 cp37 posix-bc);
my $str = join( '', map( chr($_), 0x20 .. 0x7E ) );
<----------this blank controlled by -kgba
return unless ($str);Blank lines have been introduced around the my and use sequences. What happened is that the default keyword list includes my and use but not print and return. So a continuous sequence of nine my and use statements was located. This number exceeds the default threshold of five, so blanks were placed before and after the entire group. Then, since there was also a subsequence of six my lines, a blank line was introduced to separate them.
Finer control over blank placement can be achieved by using the individual parameters rather than the -kgb flag. The individual controls are as follows.
-kgbl=s or --keyword-group-blanks-list=s, where s is a quoted string, defines the set of keywords which will be formed into groups. The string is a space separated list of keywords. The default set is s="use require local our my", but any list of keywords may be used. Comment lines may also be included in a keyword group, even though they are not keywords. To include ordinary block comments, include the symbol BC. To include static block comments (which normally begin with '##'), include the symbol SBC.
-kgbs=s or --keyword-group-blanks-size=s, where s is a string describing the number of consecutive keyword statements forming a group (Note: statements separated by blank lines in the input file are considered consecutive for purposes of this count). If s is an integer then it is the minimum number required for a group. A maximum value may also be given with the format s=min.max, where min is the minimum number and max is the maximum number, and the min and max values are separated by one or more dots. No groups will be found if the maximum is less than the minimum. The maximum is unlimited if not given. The default is s=5. Some examples:
s min max number for group
3 3 unlimited 3 or more
1.1 1 1 1
1..3 1 3 1 to 3
1.0 1 0 (no match)There is no really good default value for this parameter. If it is set too small, then an excessive number of blank lines may be generated. However, some users may prefer reducing the value somewhat below the default, perhaps to s=3.
-kgbb=n or --keyword-group-blanks-before=n specifies whether a blank should appear before the first line of the group, as follows:
n=0 => (delete) an existing blank line will be removed
n=1 => (stable) no change to the input file is made [DEFAULT]
n=2 => (insert) a blank line is introduced if possible-kgba=n or --keyword-group-blanks-after=n likewise specifies whether a blank should appear after the last line of the group, using the same scheme (0=delete, 1=stable, 2=insert).
-kgbi or --keyword-group-blanks-inside controls the insertion of blank lines between the first and last statement of the entire group. If there is a continuous run of a single statement type with more than the minimum threshold number (as specified with -kgbs=s) then this switch causes a blank line be inserted between this subgroup and the others. In the example above this happened between the use and my statements.
-kgbd or --keyword-group-blanks-delete controls the deletion of any blank lines that exist in the the group when it is first scanned. When statements are initially scanned, any existing blank lines are included in the collection. Any such original blank lines will be deleted before any other insertions are made when the parameter -kgbd is set. The default is not to do this, -nkgbd.
-kgbr=n or --keyword-group-blanks-repeat-count=n specifies n, the maximum number of times this logic will be applied to any file. The special value n=0 is the same as n=infinity which means it will be applied to an entire script [Default]. A value n=1 could be used to make it apply just one time for example. This might be useful for adjusting just the use statements in the top part of a module for example.
-kgb or --keyword-group-blanks is an abbreviation equivalent to setting -kgbb=1 -kgba=1 -kgbi. This turns on keyword group formatting with a set of default values.
-nkgb or --nokeyword-group-blanks is equivalent to -kgbb=0 -kgba nkgbi. This flag turns off keyword group blank lines and is the default setting.
Here are a few notes about the functioning of this technique.
These parameters are probably more useful as part of a major code reformatting operation rather than as a routine formatting operation.
In particular, note that deleting old blank lines with -kgbd is an irreversible operation so it should be applied with care. Existing blank lines may be serving an important role in controlling vertical alignment.
Conflicts which arise among these kgb* parameters and other blank line controls are generally resolved by producing the maximum number of blank lines implied by any parameter.
For example, if the flags --freeze-blank-lines, or --keep-old-blank-lines=2, are set, then they have priority over any blank line deletion implied by the -kgb flags of this section, so no blank lines will be deleted.
For another example, if a keyword group ends at a sub and the flag kgba=0 requests no blank line there, but we also have --blank-lines-before-subs=2, then two blank lines will still be introduced before the sub.
The introduction of blank lines does not occur if it would conflict with other input controls or code validity. For example, a blank line will not be placed within a here-doc or within a section of code marked with format skipping comments. And in general, a blank line will only be introduced at the end of a group if the next statement is a line of code.
The count which is used to determine the group size is not the number of lines but rather the total number of keywords which are found. Individual statements with a certain leading keyword may continue on multiple lines, but if any of these lines is nested more than one level deep then that group will be ended.
The search for groups of lines with similar leading keywords is based on the input source, not the final formatted source. Consequently, if the source code is badly formatted, it would be best to make a first formatting pass without these options.
A style refers to a convenient collection of existing parameters.
-gnu gives an approximation to the GNU Coding Standards (which do not apply to perl) as they are sometimes implemented. At present, this style overrides the default style with the following parameters:
-lp -bl -noll -pt=2 -bt=2 -sbt=2 -icpTo use this style with -xlp (--extended-line-up-parentheses) instead of -lp (--line-up-parentheses) use -gnu -xlp.
-pbp is an abbreviation for the parameters in the book Perl Best Practices by Damian Conway:
-l=78 -i=4 -ci=4 -st -se -vt=2 -cti=0 -pt=1 -bt=1 -sbt=1 -bbt=1 -nsfs -nolq
-wbb="% + - * / x != == >= <= =~ !~ < > | & =
**= += *= &= <<= &&= -= /= |= >>= ||= //= .= %= ^= x="Please note that this parameter set includes -st (--standard-output) and -se (--standard-error-output) flags, which make perltidy act as a filter on one file only. These can be overridden by placing -nst and/or -nse after the -pbp parameter.
Also note that the value of continuation indentation, -ci=4, is equal to the value of the full indentation, -i=4. It is recommended that the either (1) the parameter -ci=2 be used instead, or (2) the flag -xci be set. This will help show structure, particularly when there are ternary statements. The following snippet illustrates these options.
# perltidy -pbp
$self->{_text} = (
!$section ? ''
: $type eq 'item' ? "the $section entry"
: "the section on $section"
)
. (
$page
? ( $section ? ' in ' : '' ) . "the $page$page_ext manpage"
: ' elsewhere in this document'
);
# perltidy -pbp -ci=2
$self->{_text} = (
!$section ? ''
: $type eq 'item' ? "the $section entry"
: "the section on $section"
)
. (
$page
? ( $section ? ' in ' : '' ) . "the $page$page_ext manpage"
: ' elsewhere in this document'
);
# perltidy -pbp -xci
$self->{_text} = (
!$section ? ''
: $type eq 'item' ? "the $section entry"
: "the section on $section"
)
. ( $page
? ( $section ? ' in ' : '' ) . "the $page$page_ext manpage"
: ' elsewhere in this document'
);The -xci flag was developed after the -pbp parameters were published so you need to include it separately.
The goal of perltidy is to improve the readability of files, but there are two commands which have the opposite effect, --mangle and --extrude. They are actually merely aliases for combinations of other parameters. Both of these strip all possible whitespace, but leave comments and pod documents, so that they are essentially reversible. The difference between these is that --mangle puts the fewest possible line breaks in a script while --extrude puts the maximum possible. Note that these options do not provided any meaningful obfuscation, because perltidy can be used to reformat the files. They were originally developed to help test the tokenization logic of perltidy, but they have other uses. One use for --mangle is the following:
perltidy --mangle myfile.pl -st | perltidy -o myfile.pl.newThis will form the maximum possible number of one-line blocks (see next section), and can sometimes help clean up a badly formatted script.
A similar technique can be used with --extrude instead of --mangle to make the minimum number of one-line blocks.
Another use for --mangle is to combine it with --delete-all-comments (-dac) to reduce the file size of a perl script.
A one-line block is a block of code where the contents within the curly braces is short enough to fit on a single line. For example,
if ( -e $file ) { print "'$file' exists\n" }The alternative, a block which spans multiple lines, is said to be a broken block. With few exceptions, perltidy retains existing one-line blocks, if it is possible within the line-length constraint, but it does not attempt to form new ones. In other words, perltidy will try to follow the input file regarding broken and unbroken blocks.
The main exception to this rule is that perltidy will attempt to form new one-line blocks following the keywords map, eval, and sort, eval, because these code blocks are often small and most clearly displayed in a single line. This behavior can be controlled with the flag --one-line-block-exclusion-list described below.
When the cuddled-else style is used, the default treatment of one-line blocks may interfere with the cuddled style. In this case, the default behavior may be changed with the flag --cuddled-break-option=n described elsewhere.
When an existing one-line block is longer than the maximum line length, and must therefore be broken into multiple lines, perltidy checks for and adds any optional terminating semicolon (unless the -nasc option is used) if the block is a code block.
As noted above, perltidy will, by default, attempt to create new one-line blocks for certain block types. This flag allows the user to prevent this behavior for the block types listed in the string s. The list s may include any of the words sort, map, grep, eval, or it may be * to indicate all of these.
So for example to prevent multi-line eval blocks from becoming one-line blocks, the command would be -olbxl='eval'. In this case, existing one-line eval blocks will remain on one-line if possible, and existing multi-line eval blocks will remain multi-line blocks.
Nested one-line blocks are lines with code blocks which themselves contain code blocks. For example, the following line is a nested one-line block.
foreach (@list) { if ($_ eq $asked_for) { last } ++$found }The default behavior is to break such lines into multiple lines, but this behavior can be controlled with this flag. The values of n are:
n=0 break nested one-line blocks into multiple lines [DEFAULT]
n=1 stable: keep existing nested-one line blocks intactFor the above example, the default formatting (-olbn=0) is
foreach (@list) {
if ( $_ eq $asked_for ) { last }
++$found;
}If the parameter -olbn=1 is given, then the line will be left intact if it is a single line in the source, or it will be broken into multiple lines if it is broken in multiple lines in the source.
This flag controls the placement of semicolons at the end of one-line blocks. Semicolons are optional before a closing block brace, and frequently they are omitted at the end of a one-line block containing just a single statement. By default, perltidy follows the input file regarding these semicolons, but this behavior can be controlled by this flag. The values of n are:
n=0 remove terminal semicolons in one-line blocks having a single statement
n=1 stable; keep input file placement of terminal semicolons [DEFAULT ]
n=2 add terminal semicolons in all one-line blocksNote that the n=2 option has no effect if adding semicolons is prohibited with the -nasc flag. Also not that while n=2 adds missing semicolons to all one-line blocks, regardless of complexity, the n=0 option only removes ending semicolons which terminate one-line blocks containing just one semicolon. So these two options are not exact inverses.
Sometimes it might be desirable to convert a script to have one-line blocks whenever possible. Although there is currently no flag for this, a simple workaround is to execute perltidy twice, once with the flag -noadd-newlines and then once again with normal parameters, like this:
cat infile | perltidy -nanl | perltidy >outfileWhen executed on this snippet
if ( $? == -1 ) {
die "failed to execute: $!\n";
}
if ( $? == -1 ) {
print "Had enough.\n";
die "failed to execute: $!\n";
}the result is
if ( $? == -1 ) { die "failed to execute: $!\n"; }
if ( $? == -1 ) {
print "Had enough.\n";
die "failed to execute: $!\n";
}This shows that blocks with a single statement become one-line blocks.
There is no automatic way to break existing long one-line blocks into multiple lines, but this can be accomplished by processing a script, or section of a script, with a short value of the parameter maximum-line-length=n. Then, when the script is reformatted again with the normal parameters, the blocks which were broken will remain broken (with the exceptions noted above).
Another trick for doing this for certain block types is to format one time with the -cuddled-else flag and --cuddled-break-option=2. Then format again with the normal parameters. This will break any one-line blocks which are involved in a cuddled-else style.
Vertical alignment refers to lining up certain symbols in a list of consecutive similar lines to improve readability. For example, the "fat commas" are aligned in the following statement:
$data = $pkg->new(
PeerAddr => join( ".", @port[ 0 .. 3 ] ),
PeerPort => $port[4] * 256 + $port[5],
Proto => 'tcp'
);Vertical alignment can be completely turned off using the -novalign flag mentioned below. However, vertical alignment can be forced to stop and restart by selectively introducing blank lines. For example, a blank has been inserted in the following code to keep somewhat similar things aligned.
%option_range = (
'format' => [ 'tidy', 'html', 'user' ],
'output-line-ending' => [ 'dos', 'win', 'mac', 'unix' ],
'character-encoding' => [ 'none', 'utf8' ],
'block-brace-tightness' => [ 0, 2 ],
'brace-tightness' => [ 0, 2 ],
'paren-tightness' => [ 0, 2 ],
'square-bracket-tightness' => [ 0, 2 ],
);Vertical alignment is implemented by locally increasing an existing blank space to produce alignment with an adjacent line. It cannot occur if there is no blank space to increase. So if a particular space is removed by one of the existing controls then vertical alignment cannot occur. Likewise, if a space is added with one of the controls, then vertical alignment might occur.
For example,
# perltidy -nwls='=>'
$data = $pkg->new(
PeerAddr=> join( ".", @port[ 0 .. 3 ] ),
PeerPort=> $port[4] * 256 + $port[5],
Proto=> 'tcp'
);The default is to use vertical alignment, but vertical alignment can be completely turned of with the -novalign flag.
A lower level of control of vertical alignment is possible with three parameters -vc, -vsc, and -vbc. These independently control alignment of code, side comments and block comments. They are described in the next section.
The parameter -valign is in fact an alias for -vc -vsc -vbc, and its negative -novalign is an alias for -nvc -nvsc -nvbc.
The -vc flag enables alignment of code symbols such as =. The default is -vc. For detailed control of which symbols to align, see the -valign-exclude-list parameter below.
The -vsc flag enables alignment of side comments and is enabled by default. If side comment alignment is disabled with -nvsc they will appear at a fixed space from the preceding code token. The default is -vsc
When -vbc is enabled, block comments can become aligned for example if one comment of a consecutive sequence of comments becomes outdented due a length in excess of the maximum line length. If this occurs, the entire group of comments will remain aligned and be outdented by the same amount. This coordinated alignment will not occur if -nvbc is set. The default is -vbc.
More detailed control of alignment types is available with these two parameters. Most of the vertical alignments in typical programs occur at one of the tokens ',', '=', and '=>', but many other alignments are possible and are given in the following list:
= **= += *= &= <<= &&= -= /= |= >>= ||= //= .= %= ^= x=
{ ( ? : , ; => && || ~~ !~~ =~ !~ // <=> -> q
if unless and or err for foreach while untilThese alignment types correspond to perl symbols, operators and keywords except for 'q', which refers to the special case of alignment in a 'use' statement of qw quotes and empty parens.
They are all enabled by default, but they can be selectively disabled by including one or more of these tokens in the space-separated list valign-exclusion-list=s. For example, the following would prevent alignment at = and if:
--valign-exclusion-list='= if'If it is simpler to specify only the token types which are to be aligned, then include the types which are to be aligned in the list of --valign-inclusion-list. In that case you may leave the valign-exclusion-list undefined, or use the special symbol * for the exclusion list. For example, the following parameters enable alignment only at commas and 'fat commas':
--valign-inclusion-list=', =>'
--valign-exclusion-list='*' ( this is optional and may be omitted )These parameter lists should consist of space-separated tokens from the above list of possible alignment tokens, or a '*'. If an unrecognized token appears, it is simply ignored. And if a specific token is entered in both lists by mistake then the exclusion list has priority.
The default values of these parameters enable all alignments and are equivalent to
--valign-exclusion-list=' '
--valign-inclusion-list='*'To illustrate, consider the following snippet with default formatting
# perltidy
$co_description = ($color) ? 'bold cyan' : ''; # description
$co_prompt = ($color) ? 'bold green' : ''; # prompt
$co_unused = ($color) ? 'on_green' : 'reverse'; # unusedTo exclude all alignments except the equals (i.e., include only equals) we could use:
# perltidy -vil='='
$co_description = ($color) ? 'bold cyan' : ''; # description
$co_prompt = ($color) ? 'bold green' : ''; # prompt
$co_unused = ($color) ? 'on_green' : 'reverse'; # unusedTo exclude only the equals we could use:
# perltidy -vxl='='
$co_description = ($color) ? 'bold cyan' : ''; # description
$co_prompt = ($color) ? 'bold green' : ''; # prompt
$co_unused = ($color) ? 'on_green' : 'reverse'; # unusedNotice in this last example that although only the equals alignment was excluded, the ternary alignments were also lost. This happens because the vertical aligner sweeps from left-to-right and usually stops if an important alignment cannot be made for some reason.
But also notice that side comments remain aligned because their alignment is controlled separately with the parameter --valign-side-comments described above.
By default, postfix if terms align and postfix unless terms align, but separately. For example,
# perltidy [DEFAULT]
print "Tried to add: @ResolveRPM\n" if ( @ResolveRPM and !$Quiet );
print "Would need: @DepList\n" if ( @DepList and !$Quiet );
print "RPM Output:\n" unless $Quiet;
print join( "\n", @RPMOutput ) . "\n" unless $Quiet;The -viu flag causes a postfix unless to be treated as if it were a postfix if for purposes of alignment. Thus
# perltidy -viu
print "Tried to add: @ResolveRPM\n" if ( @ResolveRPM and !$Quiet );
print "Would need: @DepList\n" if ( @DepList and !$Quiet );
print "RPM Output:\n" unless $Quiet;
print join( "\n", @RPMOutput ) . "\n" unless $Quiet;Setting -vsn causes columns of numbers containing both signed and unsigned values to have leading signs placed in their own column. For example:
# perltidy -vsn
my @correct = (
[ 123456.79, 86753090000.868, 11 ],
[ -123456.79, -86753090000.868, -11 ],
[ 123456.001, 80.080, 10 ],
[ -123456.001, -80.080, 0 ],
[ 10.9, 10.9, 11 ],
);The default is -vsn. This can be turned off to get is strict left justification:
# perltidy -nvsn
my @correct = (
[ 123456.79, 86753090000.868, 11 ],
[ -123456.79, -86753090000.868, -11 ],
[ 123456.001, 80.080, 10 ],
[ -123456.001, -80.080, 0 ],
[ 10.9, 10.9, 11 ],
);Some points regarding -vsn are:
This option works by inserting a single space ahead of unsigned numbers when possible. This is not done if it would require increasing the maximum width of a column.
This option is mainly limited to lists of comma-separated numbers. For multiline lists of numbers, having trailing commas can sometimes improve the results. If missing, perltidy can add them for example with parameters -wtc=b -atc. See "Adding and Deleting Commas".
This option has a control parameter --valign-signed-number-limit=N, or -vsnl=N. This value controls formatting of very long columns of numbers and should not normally need to be changed. To see its purpose, consider a very long column of just unsigned numbers, say 1000 lines. If we add a single negative number, it is undesirable to move all of the other numbers over by one space. This could create many lines of file differences but not really improve the appearance when a local section of the table was viewed. The number N avoids this problem by not adding extra indentation to a run of more than N lines of unsigned numbers. The default value, N=20, is set to be a number of lines for which the ends of a long column of unsigned numbers are not normally both in view.
The following assignment operators are aligned independently by default:
= **= += *= &= <<= &&= -= /= |= >>= ||= //= .= %= ^= x=Setting --valign-wide-equals, or -vwe, causes them to be vertically aligned together with the trailing = all aligned. For example, here is the default formatting for some code with several of these operators:
$str .= SPACE x $total_pad_count;
$str_len += $total_pad_count;
$total_pad_count = 0;
$str .= $rfields->[$j];
$str_len += $rfield_lengths->[$j];And here is the same code formatted with -vme:
# perltidy -vme
$str .= SPACE x $total_pad_count;
$str_len += $total_pad_count;
$total_pad_count = 0;
$str .= $rfields->[$j];
$str_len += $rfield_lengths->[$j];This option was added for issue git #135 and can improve readability, but it is off by default to avoid changing existing formatting.
This section describes some parameters for dealing with extended syntax.
For another method of handling extended syntax see the section "Skipping Selected Sections of Code".
Also note that the module Perl::Tidy supplies a pre-filter and post-filter capability. This requires calling the module from a separate program rather than through the binary perltidy.
This flag allows perltidy to handle certain common extensions to the standard syntax without complaint.
For example, without this flag a structure such as the following would generate a syntax error:
Method deposit( Num $amount) {
$self->balance( $self->balance + $amount );
}This flag is enabled by default but it can be deactivated with -nxs. Probably the only reason to deactivate this flag is to generate more diagnostic messages when debugging a script.
This flag causes one or more words to be treated the same as if they were the keyword sub. The string s contains one or more alias words, separated by spaces or commas.
For example,
perltidy -sal='method fun _sub M4'will cause the perltidy to treat the words 'method', 'fun', '_sub' and 'M4' the same as if they were 'sub'. Note that if the alias words are separated by spaces then the string of words should be placed in quotes.
Note that several other parameters accept a list of keywords, including 'sub' (see "Specifying Block Types"). You do not need to include any sub aliases in these lists. Just include keyword 'sub' if you wish, and all aliases are automatically included.
This flag allows a code block following an external 'list operator' function to be formatted as if it followed one of the built-in keywords grep, map or sort. The string s contains the names of one or more such list operators, separated by spaces or commas.
By 'list operator' is meant a function which is invoked in the form
word {BLOCK} @listPerltidy tries to keep code blocks for these functions intact, since they are usually short, and does not automatically break after the closing brace since a list may follow. It also does some special handling of continuation indentation.
For example, the code block arguments to functions 'My_grep' and 'My_map' can be given formatting like 'grep' with
perltidy -gal='My_grep My_map'By default, the following list operators in List::Util are automatically included:
all any first none notall reduce reductionsAny operators specified with --grep-alias-list are added to this list. The next parameter can be used to remove words from this default list.
The -gaxl=s flag provides a method for removing any of the default list operators given above by listing them in the string s. To remove all of the default operators use -gaxl='*'.
This flag tells perltidy to allow or disallow the syntax associated a pragma in string s. The current possible settings are:
--use-feature='class'. This tells perltidy to recognized the special words class, method, field, and ADJUST as defined for this feature.
--use-feature='noclass'. This tells perltidy not to treat words class, method, field, ADJUST specially.
Neither of these (--use-feature not defined). This is the DEFAULT and recommended setting. In this case perltidy will try to automatically handle both the newer --use-feature 'class' syntax as well as some conflicting uses of some of these special words by existing modules.
Note that this parameter is independent of any use feature control lines within a script. Perltidy does not look for or read such control lines. This is because perltidy must be able to work on small chunks of code sent from an editor, so it cannot assume that such lines will be within the lines being formatted.
Perltidy contains a number of rules which help avoid known subtleties and problems with older versions of perl, and these rules always take priority over whatever formatting flags have been set. For example, perltidy will usually avoid starting a new line with a bareword, because this might cause problems if use strict is active.
There is no way to override these rules.
Perltidy can selectively delete comments and/or pod documentation. The command -dac or --delete-all-comments will delete all comments and all pod documentation, leaving just code and any leading system control lines.
The command -dp or --delete-pod will remove all pod documentation (but not comments).
Two commands which remove comments (but not pod) are: -dbc or --delete-block-comments and -dsc or --delete-side-comments. (Hanging side comments will be deleted with side comments here.)
When side comments are deleted, any special control side comments for non-indenting braces will be retained unless they are deactivated with a -nnib flag.
The negatives of these commands also work, and are the defaults. When block comments are deleted, any leading 'hash-bang' will be retained. Also, if the -x flag is used, any system commands before a leading hash-bang will be retained (even if they are in the form of comments).
When perltidy writes a formatted text file, it has the ability to also send selected text to a file with a .TEE extension. This text can include comments and pod documentation.
The command -tac or --tee-all-comments will write all comments and all pod documentation.
The command -tp or --tee-pod will write all pod documentation (but not comments).
The commands which write comments (but not pod) are: -tbc or --tee-block-comments and -tsc or --tee-side-comments. (Hanging side comments will be written with side comments here.)
The negatives of these commands also work, and are the defaults.
If you use perltidy frequently, you probably won't be happy until you create a .perltidyrc file to avoid typing commonly-used parameters. Perltidy will first look in your current directory for a command file named .perltidyrc. If it does not find one, it will continue looking for one in other standard locations.
These other locations are system-dependent, and may be displayed with the command perltidy -dpro. Under Unix systems, it will first look for an environment variable PERLTIDY. Then it will look for a .perltidyrc file in the home directory, and then for a system-wide file /usr/local/etc/perltidyrc, and then it will look for /etc/perltidyrc. Note that these last two system-wide files do not have a leading dot. Further system-dependent information will be found in the INSTALL file distributed with perltidy.
Under Windows, perltidy will also search for a configuration file named perltidy.ini since Windows does not allow files with a leading period (.). Use perltidy -dpro to see the possible locations for your system. An example might be C:\Documents and Settings\All Users\perltidy.ini.
Another option is the use of the PERLTIDY environment variable. The method for setting environment variables depends upon the version of Windows that you are using.
Under Windows NT / 2000 / XP the PERLTIDY environment variable can be placed in either the user section or the system section. The later makes the configuration file common to all users on the machine. Be sure to enter the full path of the configuration file in the value of the environment variable. Ex. PERLTIDY=C:\Documents and Settings\perltidy.ini
The configuration file is free format, and simply a list of parameters, just as they would be entered on a command line. Any number of lines may be used, with any number of parameters per line, although it may be easiest to read with one parameter per line. Comment text begins with a #, and there must also be a space before the # for side comments. It is a good idea to put complex parameters in either single or double quotes.
Here is an example of a .perltidyrc file:
# This is a simple of a .perltidyrc configuration file
# This implements a highly spaced style
-se # errors to standard error output
-w # show all warnings
-bl # braces on new lines
-pt=0 # parens not tight at all
-bt=0 # braces not tight
-sbt=0 # square brackets not tightThe parameters in the .perltidyrc file are installed first, so any parameters given on the command line will have priority over them.
To avoid confusion, perltidy ignores any command in the .perltidyrc file which would cause some kind of dump and an exit. These are:
-h -v -ddf -dln -dop -dsn -dtt -dwls -dwrs -ssThere are several options may be helpful in debugging a .perltidyrc file:
A very helpful command is --dump-profile or -dpro. It writes a list of all configuration filenames tested to standard output, and if a file is found, it dumps the content to standard output before exiting. So, to find out where perltidy looks for its configuration files, and which one if any it selects, just enter
perltidy -dproIt may be simplest to develop and test configuration files with alternative names, and invoke them with -pro=filename on the command line. Then rename the desired file to .perltidyrc when finished.
The parameters in the .perltidyrc file can be switched off with the -npro option.
Any parameter in the .perltidyrc file can be overridden with a replacement value on the command line. This is because the command line is processed after the .perltidyrc file.
The commands --dump-options, --dump-defaults, --dump-long-names, and --dump-short-names, all described below, may all be helpful.
Versions of perltidy greater than 20240511 have an option to filter unrecognized parameters from a perltidy command file. If a line in the file begins with three dashes followed by a parameter name (rather than one or two), then the line will be removed if the parameter is unknown. Otherwise, a dash will be removed to make the line valid. This option was added to allow a single command file to be used during the transition to a new version of perltidy.
A special notation is available for use in a .perltidyrc file for creating an abbreviation for a group of options. This can be used to create a shorthand for one or more styles which are frequently, but not always, used. The notation is to group the options within curly braces which are preceded by the name of the alias (without leading dashes), like this:
newword {
-opt1
-opt2
}where newword is the abbreviation, and opt1, etc, are existing parameters or other abbreviations. The main syntax requirement is that the new abbreviation along with its opening curly brace must begin on a new line. Space before and after the curly braces is optional.
For a specific example, the following line
oneliner { --maximum-line-length=0 --noadd-newlines --noadd-terminal-newline}or equivalently with abbreviations
oneliner { -l=0 -nanl -natnl }could be placed in a .perltidyrc file to temporarily override the maximum line length with a large value, to temporarily prevent new line breaks from being added, and to prevent an extra newline character from being added the file. All other settings in the .perltidyrc file still apply. Thus it provides a way to format a long 'one liner' when perltidy is invoked with
perltidy --oneliner ...(Either -oneliner or --oneliner may be used).
If your script has leading lines of system commands or other text which are not valid perl code, and which are separated from the start of the perl code by a "hash-bang" line, ( a line of the form #!...perl ), you must use the -x flag to tell perltidy not to parse and format any lines before the "hash-bang" line. This option also invokes perl with a -x flag when checking the syntax. This option was originally added to allow perltidy to parse interactive VMS scripts, but it should be used for any script which is normally invoked with perl -x.
Please note: do not use this flag unless you are sure your script needs it. Parsing errors can occur if it does not have a hash-bang, or, for example, if the actual first hash-bang is in a here-doc. In that case a parsing error will occur because the tokenization will begin in the middle of the here-doc.
The following flags are available for debugging. Note that all commands named --dump-... will simply write some requested information to standard output and then immediately exit.
--dump-cuddled-block-list or -dcbl will dump to standard output the internal hash of cuddled block types created by a -cuddled-block-list input string.
--dump-defaults or -ddf will write the default option set to standard output and quit
--dump-integer-option-range or -dior will write a list of comma-separated values. Each line contains the name of an integer option along with its minimum, maximum, and default values.
--dump-profile or -dpro will write the name of the current configuration file and its contents to standard output and quit.
--dump-options or -dop will write current option set to standard output and quit.
--dump-long-names or -dln will write all command line long names (passed to Get_options) to standard output and quit.
--dump-short-names or -dsn will write all command line short names to standard output and quit.
--dump-token-types or -dtt will write a list of all token types to standard output and quit.
--dump-want-left-space or -dwls will write the hash %want_left_space to standard output and quit. See the section on controlling whitespace around tokens.
--dump-want-right-space or -dwrs will write the hash %want_right_space to standard output and quit. See the section on controlling whitespace around tokens.
--no-memoize or -nmem will turn of memoizing. Memoization can reduce run time when running perltidy repeatedly in a single process. It is on by default but can be deactivated for testing with -nmem.
--no-timestamp or -nts will eliminate any time stamps in output files to prevent differences in dates from causing test installation scripts to fail. There are just a couple of places where timestamps normally occur. One is in the headers of html files, and another is when the -cscw option is selected. The default is to allow timestamps (--timestamp or -ts).
--file-size-order or -fso will cause files to be processed in order of increasing size, when multiple files are being processed. This is particularly useful during program development, when large numbers of files with varying sizes are processed, because it can reduce virtual memory usage. This is the default and can be deactivated with -nfso.
--maximum-file-size-mb=n or -maxfs=n specifies the maximum file size in megabytes that perltidy will attempt to format. This parameter is provided to avoid causing system problems by accidentally attempting to format an extremely large data file. Most perl scripts are less than about 2 MB in size. The integer n has a default value of 10, so perltidy will skip formatting files which have a size greater than 10 MB. The command to increase the limit to 20 MB for example would be
perltidy -maxfs=20This length test is applied to named files before they are read into memory. It is applied to files arriving from standard input after they are read into memory. It is not applied to character strings arriving by a call to the Perl::Tidy module.
--maximum-level-errors=n or -maxle=n specifies the maximum number of indentation level errors are allowed before perltidy skips formatting and just outputs a file verbatim. The default is n=1. This means that if the final indentation of a script differs from the starting indentation by more than 1 levels, the file will be output verbatim. To avoid formatting if there are any indentation level errors use -maxle=0. To skip this check you can either set n equal to a large number, such as n=100, or set n=-1.
For example, the following script has level error of 3 and will be output verbatim
Input and default output:
{{{
perltidy -maxle=100
{
{
{--maximum-unexpected-errors=n or -maxue=n specifies the maximum number of unexpected tokenization errors are allowed before formatting is skipped and a script is output verbatim. The intention is to avoid accidentally formatting a non-perl script, such as an html file for example. This check can be turned off by setting n=0.
A recommended value is n=3. However, the default is n=0 (skip this check) to avoid causing problems with scripts which have extended syntaxes.
Many of the input parameters take integer values. Before processing begins, a check is made to see if any of these integer parameters exceed their valid ranges. The default behavior when a range is exceeded is to write a warning message and reset the value to its default setting. This default behavior can be changed with the parameter --integer-range-check=n, or -irc=n, as follows:
n=0 skip check completely (for stress-testing perltidy only)
n=1 reset bad values to defaults but do not issue a warning
n=2 reset bad values to defaults and issue warning [DEFAULT]
n=3 stop if any values are out of boundsThe values n=0 and n=1 are mainly useful for testing purposes.
-DEBUG, -D will write a file with extension .DEBUG for each input file showing the tokenization of all lines of code. This can produce a lot of output and is mainly useful for debugging tokenization issues during perltidy development.
Perltidy reports any obvious issues that are found during formatting, such as unbalanced braces. But several parameters are available for making certain additional checks for issues which might be of interest to a programmer. These parameters fall into two categories as indicated by their prefix, --dump- or --warn-:
The --dump- parameters read a file, write information to the standard output, and then exit without doing any formatting.
The --warn- parameters, on the other hand, cause perltidy to function normally but issue warnings to the error output when certain conditions are encountered.
Some of these have associated control parameters.
A table listing information about the blocks of code in a file can be made with --dump-block-summary, or -dbs. This causes perltidy to read and parse the file, write a table of comma-separated values for selected code blocks to the standard output, and then exit. This parameter must be on the command line, not in a .perlticyrc file, and it requires a single file name on the command line. For example
perltidy -dbs somefile.pl >blocks.csvproduces an output file blocks.csv whose lines hold these parameters:
filename - the name of the file
line - the line number of the opening brace of this block
line_count - the number of lines between opening and closing braces
code_lines - the number of lines excluding blanks, comments, and pod
type - the block type (sub, for, foreach, ...)
name - the block name if applicable (sub name, label, asub name)
depth - the nesting depth of the opening block brace
max_change - the change in depth to the most deeply nested code block
block_count - the total number of code blocks nested in this block
mccabe_count - the McCabe complexity measure of this code blockThis feature was developed to help identify complex sections of code as an aid in refactoring. The McCabe complexity measure follows the definition used by Perl::Critic. By default the table contains these values for subroutines, but the user may request them for any or all blocks of code or packages. For blocks which are loops nested within loops, a postfix '+' to the type is added to indicate possible code complexity. Although the table does not otherwise indicate which blocks are nested in other blocks, this can be determined by computing and comparing the block ending line numbers.
For subroutines, the number of call arguments (args) is listed in parentheses in the type column. For example, sub(9) indicates a sub with 9 args. Subroutines whose arg count cannot easily be determined are indicated as sub(*). The count includes any leading object passed in a method call (such as $self).
By default the table lists subroutines with more than 20 code_lines, but this can be changed with the following two parameters:
--dump-block-minimum-lines=n, or -dbl=n, where n is the minimum number of code_lines to be included. The default is -n=20. Note that code_lines is the number of lines excluding comments, blanks and pod.
--dump-block-types=s, or -dbt=s, where string s is a list of block types to be included. The type of a block is either the name of the perl builtin keyword for that block (such as sub if elsif else for foreach ..) or the word immediately before the opening brace. In addition, there are a few symbols for special block types, as follows:
if elsif else for foreach ... any keyword introducing a block
sub - any sub or anonymous sub
asub - any anonymous sub
* - any block except nameless blocks
+ - any nested inner block loop
package - any package or class
closure - any nameless block
elsif3 - an if-elsif-..-else chain with 3 or more elsif's (3 is arbitrary, see below)A chain of if-elsif-... blocks may be reported as a single line item by entering the word elsif with an appended integer, as indicated by the last item in this list. The integer indicates the number of elsif blocks required for a chain to be reported. If you use this, you may want to also use -dbl=n, with a smaller number of lines n than the default.
In addition, specific block loop types which are nested in other loops can be selected by adding a + after the block name. (Nested loops are sometimes good candidates for restructuring).
The default is -dbt='sub'.
In the following examples a table block.csv is created for a file somefile.pl:
This selects both subs and packages which have 20 or more lines of code. This can be useful in code which contains multiple packages.
perltidy -dbs -dbt='sub package' somefile.pl >blocks.csvThis selects block types sub for foreach while with 10 or more code lines.
perltidy -dbs -dbl=10 -dbt='sub for foreach while' somefile.pl >blocks.csvThis selects blocks with 2 or more code lines which are type sub or which are inner loops.
perltidy -dbs -dbl=2 -dbt='sub +' somefile.pl >blocks.csvThis selects every block and package.
perltidy -dbs -dbl=1 -dbt='* closure' somefile.pl >blocks.csvThis selects every if-chain which contains 2 or more elsif blocks:
perltidy -dbs -dbl=1 -dbt='elsif2' somefile.pl >blocks.csvVariables with certain properties of interest to a programmer can be listed with --dump-unusual-variables or -duv. This parameter must be on the command line, along with a single file name. It causes perltidy to scan the file for certain variable types, write any found to the standard output, and then immediately exit without doing any formatting. For example
perltidy -duv somefile.pl >vars.txtproduces a file with lines which look something like
1778:u: my $input_file
6089:r: my $j: reused - see line 6076The values on the line are separated by colons and have the following meaning:
line number - the number of the line of the input file
issue - a single letter indicating the issue, see below
variable name - the name of the variable, preceded by a keyword
note - an optional note referring to another lineIf there are a large number of issues it can be convenient to read the file into a spreadsheet.
The checks are made for variables introduced by the keywords my, state, and our, along with variables defined with use vars and use constant. It is a good idea to also set use strict in a script so that Perl itself can find issues with variables which appear in a script without one of these methods.
The types of checks which are made are identified in the output with one of the letters r, s, p, u, and c as follows:
These are variables which are re-declared in the scope of a variable with the identical name. This can be confusing, perhaps not when the code is first written, but possibly later during maintenance work. For example, this can make it difficult to locate the correct variable with an editor when changes are being made. This issue can be avoided by renaming one of the conflicting variables. Note that this is similar to the Perl::Critic policy Variables::ProhibitReusedNames.
These are variables which have the same bareword name but a different sigil ($, @, or %) as another variable in the same scope. For example, this occurs if variables $data and %data share the same scope. This can also be confusing for the reasons mentioned above and can be avoided by renaming one of the variables.
These are lexical variables which are declared in one package and still visible in subroutines of a different package in the same file. This can be confusing, and it might cause the program to run differently, or fail, if the the packages were ever split into separate files. This issue can usually be avoided by placing code in block braces of some type. For example, this issue is often found in test code and can sometimes be fixed by using the structure
main();
sub main { #<<<
## old main code goes here
}The non-indenting-braces side comment #<<< is not required but will keep the indentation of the old code unchanged.
This check is only applied to package statements which are not enclosed in block braces in order avoid warnings at temporary package changes.
These are lexical variables declared with my or state (but not our) and not referenced again within their scope. Calling them unused is convenient but not really accurate; this is a "gray area" for a program. There are some good reasons for having such variables. For example, they might occur in a list of values provided by another routine or data structure, and therefore must be listed, even though they might not be referenced again. Having such variables can make them immediately available for future development and debugging, and can be beneficial for program clarity.
But sometimes they can occur due to being orphaned by a coding change, due to a misspelling, or by having an unintentional preceding my. So it is worth reviewing them, especially for new code. Here is an example of an unused variable in a script located with this method:
BEGIN { my $string = "" }
...
$string .= "ok";This looks nice at first glance, but the scope of the my declaration is limited to the surrounding braces, so it is not the same variable as the other $string and must therefore be reported as unused. This particular problem would have also been caught by perl if the author had used strict.
These are names which are declared with a use constant and a reference was not seen again within their package. They might be needed by an external package, or a set of standard definitions, or available for future development. And in some unusual cases a reference may have been missed by perltidy. But they might also be unused remnants from code development, or due to a misspelling, so it can be worthwhile reviewing them.
Exception: The following our variables are exempt from warnings: $VERSION, @EXPORT, @EXPORT_OK, %EXPORT_TAGS, @ISA, $AUTOLOAD.
The flag --warn-variable-types=string, or -wvt=string, is the --warn counterpart to --dump-unusual-variables, and can be used to produce a warning message if certain of the above variable types are encountered during formatting. All possible variable warnings may be requested with -wvt='*' or -wvt=1.
For example,
perltidy -wvt='*' somefile.plThe default is not to do any of these checks, and it can also be indicated with -wvt=0.
To restrict the check to a specific set warnings, set the input string to be a space-separated or comma-separated list of the letters associated with the types of variables to be checked. For example:
perltidy -wvt='s r' somefile.plwill process somefile.pl normally but issue a warning if either of the issues s or r, described above, are encountered.
A companion flag, --warn-variable-exclusion-list=string, or -wvxl=string, can be used to skip warning checks for a list of variable names. A leading and/or trailing '*' may be placed on any of these variable names to allow a partial match.
For example,
perltidy -wvt=1 -wvxl='$self $class *_unused' somefile.plwill do all possible checks but not report any warnings for variables $self, $class, and for example $value_unused.
This partial match option provides a way to trigger a warning message when a new unused variable is detected in a script. This can be accomplished by adding a unique suffix to the names of existing unused variables, such as _unused. This suffix is then added to the exclusion list.
As a specific example, consider the following line which is part of some debug code which only references the latter three variables (but might someday need to reference the package variable too).
my ( $package_uu, $filename, $line, $subroutine ) = caller();The unused variable, $package_uu, has been specially marked with suffix _uu. No type u (unused variable) warning will be produced provided that this wildcard suffix is in the exclusion list:
-wvxl='*_uu'The parameter --dump-mixed-call-parens, or -dmcp, provides information on the use of call parens within a program. It produces a list of keywords and sub names which occur both both with and without parens. In other words, with a mixed style. This might be useful if one is working to standardize the call style for some particular keyword or function. For example,
perltidy -dmcp somefile.pl >output.txtwill analyze the text of somefile.pl, write the results to output.txt, and then immediately exit (like all dump- parameters).
The output shows a list of operators and the number of times they were used with parens and the number of times without parens. For example, here is a small section of the output from one file in a past Perl distribution:
k:length:17:9
k:open:30:9
k:pop:3:4The first line shows that the length function occurs 17 times with parens and 9 times without parens. The 'k' indicates that length is a Perl builtin keyword ('U' would mean user-defined sub, and 'w' would mean unknown bareword). So from this partial output we see that the author had a preference for parens around the args of length and open, whereas pop was about equally likely to have parens as not.
More detailed information can be obtained with the parameters described in the next section.
The parameter --want-call-parens=s, or -wcp=s, can be used to to produce a warning message if call parens are missing from selected functions. Likewise, --nowant-call-parens=s, or -nwcp=s, can warn if call parens exist for selected functions. When either of these parameters are set, perltidy will report any discrepancies from the requested style in its error output.
Before using either of these parameters, it may be helpful to first use --dump-mixed-call-parens=s, described in the previous section, to get an overview of the existing paren usage in a file.
The string arguments s are space-separated lists of the names of the functions to be checked. The function names may be builtin keywords or user-defined subs. They may not include a package prefix or sigil. To illustrate,
perltidy -wcp='length open' -nwcp='pop' somefile.plmeans that the builtin functions length and open should have parens around their call args but pop should not. The error output might contain lines such as:
2314:open FD_TO_CLOSE: no call parens
3652:pop (: has call parens
3783:length $DB: no call parens
...For builtin keywords which have both a block form and a trailing modifier form, such as if, only the trailing modifier form will be checked since parens are mandatory for the block form.
The symbol & may entered instead of a function name to mean all user-defined subs not explicitly listed. So the compact expression
perltidy -wcp='&' somefile.plmeans that calls to all user-defined subs in the file being processed should have their call arguments enclosed in parens.
Perltidy does not have the ability to add or delete call parens because it is difficult to automate, so changes must be made manually. When adding or removing parentheses, it is essential to pay attention to operator precedence issues. For example, if the parens in the following statement are removed, then || must be changed to or:
open( IN, "<", $infile ) || die("cannot open $infile:$!\n");Otherwise, the || will operate on $infile rather than the return value of open.
The parameter --dump-mismatched-args, or -dma, causes perltidy to examine the definitions of subroutines in a file, and calls to those subs, and report certain differences. Like all --dump commands, it writes its report to standard output and exits immediately. For example
perltidy -dma somefile.pl >results.txtFour types of issues are reported, types a, o, u, and i:
For example the following two lines would be reported as a mismatch:
Fault();and
$self->Fault();This may or may not be an error, but it is worth checking. It might become an error in the future if sub Fault starts to access $self.
For example
sub gnab_gib {
my $self=shift;
my ($v1,$v2)=@_;
...
}
$self->gnab_gib(42);In this case, the sub is expecting a total of three args ($self, $v1, and $v2) but only receives two ($self and 42), so an undercount is reported. This is not necessarily an error because the sub may allow for this possibility, but it is worth checking. The simple static processing done by perltidy cannot determine which sub args are optional.
Notes and Limitations:
This option works best for subs which unpack call args in an orderly manner near the beginning of the sub from @_ and/or with shift operations. It will also work for direct access to specific elements of the @_ array. However if the coding for arg extraction is complex then the number of sub args will be considered indeterminate and a count cannot be checked.
Sub calls made without parentheses around the args are not checked.
Anonymous subs and lexical subs (introduced with my) are not checked.
Only calls which appear to be to subs defined within the file being processed are checked. But note that a file may contain multiple packages.
This is similar to the -dump parameter described above except that any mismatches are reported in the error file and otherwise formatting continues normally. The basic usage is
perltidy -wma somefile.plSeveral companion controls are available to avoid unwanted error messages:
--warn-mismatched-arg-types=s, or -wmat=s, can be used to select specific tests, type a (arrow test) or o (overcounts) or u (undercounts). All checks may be requested with -wmat='*' or -wmat=1. This is the default.
To restrict the check to a specific warning type, set the string equal to the letter of that warning, any a, o, or u. For example
perltidy -wmat='a o' somefile.plwill format somefile.pl and report any arrow-type mismatches and overcount mismatches, but will skip undercount mismatches.
--warn-mismatched-arg-exclusion-list, or -wmaxl=string, can be given to skip the warning checks for a list of subroutine names, entered as a quoted string of space- or comma-separated names, without a package prefix. All subs with those names will be skipped, regardless of package. A leading and/or trailing * on a name may be used to indicate a partial string match.
--warn-mismatched-arg-undercount-cutoff=n, or -wmauc=n, can be used to avoid undercount warnings when the expected number of args is less than n. Please note that this number n is the number of args from the point of view of the sub definition, so an object like $self passed with an arrow operator counts as one arg.
The default value is n=4. This has been found to allow most programs to pass without warnings, but it should be reduced if possible for better error checking. The minimum possible value of n needed to avoid triggering an error for a program can be determined by running with -wma -wmauc=0. If there are undercount errors, a note at the bottom of the error output indicates the value of n required to avoid reporting them.
--warn-mismatched-arg-overcount-cutoff=n, or -wmaoc=n, can be used to avoid overcount warnings when the expected number of args is less than n. The default value is n=1. This avoids warning messages for subroutines which are dummy placeholders for possible expansion.
To illustrate these controls,
perltidy -wma -wmat='o u' -wmaxl='new old' -wmauc=2 somefile.plmeans format somefile.pl as usual and check for mismatched overcounts and undercounts but not arrows. Skip checking for any sub named new or old, and only warn of undercounts for subs expecting more than 2 args.
The parameter --dump-mismatched-returns, or -dmr, examines the return side of sub call statements. Like all --dump commands, it writes its report to standard output and exits immediately. For example
perltidy -dmr somefile.pl >results.txtThe following types of issues are reported:
These issue types are illustrated with the following code
sub macho {
...
( $name, $flags ); # 2 values but no 'return' statement
}
( $name, $flags ) = macho(); # 'x' (want array but no return stmt)
$name = macho(); # 'y' (want scalar but no return stmt)
sub wimp {
...;
return ( $name, $flags ); # 2 values with 'return' statement
}
( $name, $flags, $access) = wimp(); # 'o' (want array 3 > 2)
($name) = wimp(); # 'u' (want array 1 < 2)
$name = wimp(); # 's' (want scalar but 2 values returned)This analysis works by scanning all call statements and all sub return statements, and comparing the the number of items wanted with the possible number of items returned. If a specific value for either of these numbers cannot be determined for a call then it cannot be checked.
Since only return statements are scanned for return values, this analysis will not be useful for programming which relies on the default return mechanism, as in the first sub above. Note that the Perl::Critic policy RequireFinalReturn can be used to check for code in this situation.
Reported issues are not necessarily errors, but they might be, or they might indicate potentially confusing code.
This is similar to the -dump parameter described above except that any mismatches are reported in the error file and otherwise formatting continues normally. The basic usage is
perltidy -wmr somefile.plThe following companion controls are available to avoid unwanted error messages:
--warn-mismatched-return-types=string, or -wmrt=string, can be used to limit checks.
To restrict the checking, set the string equal to the letter(s) of that warning, any x, y, o, u, or s. For example
perltidy -wmrt='x o s' somefile.plwill format somefile.pl and report issue types x, o, and s but not types u and y. All checks may be requested with -wmrt='*' or -wmrt=1. This is the default.
--warn-mismatched-return-exclusion-list, or -wmrxl=string, can be given to skip the warning checks for a list of subroutine names, entered as a quoted string of space- or comma-separated names, without a package prefix. All subs with those names will be skipped, regardless of package. A leading and/or trailing * on a name may be used to indicate a partial string match.
The first $VERSION line of a file which might be eval'd by MakeMaker is passed through unchanged except for indentation. The default --pass-version-line, or -pvl, is to do this. Use --nopass-version-line, or -npvl, to deactivate this feature.
If the AutoLoader module is used, perltidy will continue formatting code after seeing an __END__ line. The default --look-for-autoloader, or -lal, is to do this. Use --nolook-for-autoloader, or -nlal, to deactivate this feature.
Likewise, if the SelfLoader module is used, perltidy will continue formatting code after seeing a __DATA__ line. The default --look-for-selfloader, or -lsl, is to do this. Use --nolook-for-selfloader, or -nlsl, to deactivate this feature.
The flag -html causes perltidy to write an html file with extension .html. So, for example, the following command
perltidy -html somefile.plwill produce a syntax-colored html file named somefile.pl.html which may be viewed with a browser.
Please Note: In this case, perltidy does not do any formatting to the input file, and it does not write a formatted file with extension .tdy. This means that two perltidy runs are required to create a fully reformatted, html copy of a script.
When the -pre flag is given, only the pre-formatted section, within the <PRE> and </PRE> tags, will be output. This simplifies inclusion of the output in other files. The default is to output a complete web page.
When the -nnn flag is given, the output lines will be numbered.
By default, a table of contents to packages and subroutines will be written at the start of html output. Use -ntoc to prevent this. This might be useful, for example, for a pod document which contains a number of unrelated code snippets. This flag only influences the code table of contents; it has no effect on any table of contents produced by pod2html (see next item).
There are two options for formatting pod documentation. The default is to pass the pod through the Pod::Html module (which forms the basis of the pod2html utility). Any code sections are formatted by perltidy, and the results then merged. Note: perltidy creates a temporary file when Pod::Html is used; see "FILES". Also, Pod::Html creates temporary files for its cache.
NOTE: Perltidy counts the number of =cut lines, and either moves the pod text to the top of the html file if there is one =cut, or leaves the pod text in its original order (interleaved with code) otherwise.
Most of the flags accepted by pod2html may be included in the perltidy command line, and they will be passed to pod2html. In some cases, the flags have a prefix pod to emphasize that they are for the pod2html, and this prefix will be removed before they are passed to pod2html. The flags which have the additional pod prefix are:
--[no]podheader --[no]podindex --[no]podrecurse --[no]podquiet
--[no]podverbose --podflushThe flags which are unchanged from their use in pod2html are:
--backlink=s --cachedir=s --htmlroot=s --libpods=s --title=s
--podpath=s --podroot=swhere 's' is an appropriate character string. Not all of these flags are available in older versions of Pod::Html. See your Pod::Html documentation for more information.
The alternative, indicated with -npod, is not to use Pod::Html, but rather to format pod text in italics (or whatever the stylesheet indicates), without special html markup. This is useful, for example, if pod is being used as an alternative way to write comments.
By default, a single html output file is produced. This can be changed with the -frm option, which creates a frame holding a table of contents in the left panel and the source code in the right side. This simplifies code browsing. Assume, for example, that the input file is MyModule.pm. Then, for default file extension choices, these three files will be created:
MyModule.pm.html - the frame
MyModule.pm.toc.html - the table of contents
MyModule.pm.src.html - the formatted source codeObviously this file naming scheme requires that output be directed to a real file (as opposed to, say, standard output). If this is not the case, or if the file extension is unknown, the -frm option will be ignored.
Use this flag to specify the extra file extension of the table of contents file when html frames are used. The default is "toc". See "Specifying File Extensions".
Use this flag to specify the extra file extension of the content file when html frames are used. The default is "src". See "Specifying File Extensions".
This flag controls the use of Html::Entities for html formatting. By default, the module Html::Entities is used to encode special symbols. This may not be the right thing for some browser/language combinations. Use --nohtml-entities or -nhent to prevent this.
Style sheets make it very convenient to control and adjust the appearance of html pages. The default behavior is to write a page of html with an embedded style sheet.
An alternative to an embedded style sheet is to create a page with a link to an external style sheet. This is indicated with the -css=filename, where the external style sheet is filename. The external style sheet filename will be created if and only if it does not exist. This option is useful for controlling multiple pages from a single style sheet.
To cause perltidy to write a style sheet to standard output and exit, use the -ss, or --stylesheet, flag. This is useful if the style sheet could not be written for some reason, such as if the -pre flag was used. Thus, for example,
perltidy -html -ss >mystyle.csswill write a style sheet with the default properties to file mystyle.css.
The use of style sheets is encouraged, but a web page without a style sheets can be created with the flag -nss. Use this option if you must to be sure that older browsers (roughly speaking, versions prior to 4.0 of Netscape Navigator and Internet Explorer) can display the syntax-coloring of the html files.
Note: It is usually more convenient to accept the default properties and then edit the stylesheet which is produced. However, this section shows how to control the properties with flags to perltidy.
Syntax colors may be changed from their default values by flags of the either the long form, -html-color-xxxxxx=n, or more conveniently the short form, -hcx=n, where xxxxxx is one of the following words, and x is the corresponding abbreviation:
Token Type xxxxxx x
---------- -------- --
comment comment c
number numeric n
identifier identifier i
bareword, function bareword w
keyword keyword k
quote, pattern quote q
here doc text here-doc-text h
here doc target here-doc-target hh
punctuation punctuation pu
parentheses paren p
structural braces structure s
semicolon semicolon sc
colon colon co
comma comma cm
label label j
sub definition name subroutine m
pod text pod-text pdA default set of colors has been defined, but they may be changed by providing values to any of the following parameters, where n is either a 6 digit hex RGB color value or an ascii name for a color, such as 'red'.
To illustrate, the following command will produce an html file somefile.pl.html with "aqua" keywords:
perltidy -html -hck=00ffff somefile.pland this should be equivalent for most browsers:
perltidy -html -hck=aqua somefile.plPerltidy merely writes any non-hex names that it sees in the html file. The following 16 color names are defined in the HTML 3.2 standard:
black => 000000,
silver => c0c0c0,
gray => 808080,
white => ffffff,
maroon => 800000,
red => ff0000,
purple => 800080,
fuchsia => ff00ff,
green => 008000,
lime => 00ff00,
olive => 808000,
yellow => ffff00
navy => 000080,
blue => 0000ff,
teal => 008080,
aqua => 00ffff,Many more names are supported in specific browsers, but it is safest to use the hex codes for other colors. Helpful color tables can be located with an internet search for "HTML color tables".
Besides color, two other character attributes may be set: bold, and italics. To set a token type to use bold, use the flag --html-bold-xxxxxx or -hbx, where xxxxxx or x are the long or short names from the above table. Conversely, to set a token type to NOT use bold, use --nohtml-bold-xxxxxx or -nhbx.
Likewise, to set a token type to use an italic font, use the flag --html-italic-xxxxxx or -hix, where again xxxxxx or x are the long or short names from the above table. And to set a token type to NOT use italics, use --nohtml-italic-xxxxxx or -nhix.
For example, to use bold braces and lime color, non-bold, italics keywords the following command would be used:
perltidy -html -hbs -hck=00FF00 -nhbk -hik somefile.plThe background color can be specified with --html-color-background=n, or -hcbg=n for short, where n is a 6 character hex RGB value. The default color of text is the value given to punctuation, which is black as a default.
Here are some notes and hints:
1. If you find a preferred set of these parameters, you may want to create a .perltidyrc file containing them. See the perltidy man page for an explanation.
2. Rather than specifying values for these parameters, it is probably easier to accept the defaults and then edit a style sheet. The style sheet contains comments which should make this easy.
3. The syntax-colored html files can be very large, so it may be best to split large files into smaller pieces to improve download times.
Several parameters which refer to code block types may be customized by also specifying an associated list of block types. The type of a block is the name of the keyword which introduces that block, such as if, else, or sub. An exception is a labeled block, which has no keyword, and should be specified with just a colon. To specify all blocks use '*'.
The keyword sub indicates a named sub. For anonymous subs, use the special keyword asub.
For example, the following parameter specifies sub, labels, BEGIN, and END blocks:
-cscl="sub : BEGIN END"(the meaning of the -cscl parameter is described above.) Note that quotes are required around the list of block types because of the spaces. For another example, the following list specifies all block types for vertical tightness:
-bbvtl='*'Several parameters allow default file extensions to be overridden. For example, a backup file extension may be specified with -bext=ext, where ext is some new extension. In order to provides the user some flexibility, the following convention is used in all cases to decide if a leading '.' should be used. If the extension ext begins with A-Z, a-z, or 0-9, then it will be appended to the filename with an intermediate '.' (or perhaps a '_' on VMS systems). Otherwise, it will be appended directly.
For example, suppose the file is somefile.pl. For -bext=old, a '.' is added to give somefile.pl.old. For -bext=.old, no additional '.' is added, so again the backup file is somefile.pl.old. For -bext=~, then no dot is added, and the backup file will be somefile.pl~ .
The following list shows all short parameter names which allow a prefix 'n' to produce the negated form:
D aia altc ame anl asbl asc ast asu atc
atnl aws b baa baao bar bbao bbb bbc bbs
bl bli boa boc bok bol bom bos bot cblx
ce conv cpb cs csc cscb cscw dac dbc dbs
dcbl dcsc ddf dia dior dln dltc dma dmcp dmr
dnl dop dp dpro drc dsc dsm dsn dtc dtco
dtt duv dwic dwls dwrs dws eos f fpva frm
fs fso gcs hbc hbcm hbco hbh hbhh hbi hbj
hbk hbm hbn hbp hbpd hbpu hbq hbs hbsc hbv
hbw hent hic hicm hico hih hihh hii hij hik
him hin hip hipd hipu hiq his hisc hiv hiw
hsc html ibc icb icp iob ipc isbc iscl kgb
kgbd kgbi kis lal log lop lp lsl mci mem
nib ohbr okw ola olc oll olq opr opt osbc
osbr otr ple pod pvl q sac sbc sbl scbb
schb scp scsb sct se sfp sfs skp sob sobb
sohb sop sosb sot ssc st sts t tac tbc
toc tp tqw trp ts tsc tso vbc vc viu
vmll vsc vsn vwe w wfc wia wma wme wmr
wn x xbt xci xlp xsEquivalently, the prefix 'no' or 'no-' on the corresponding long names may be used.
Perltidy should work properly on most perl scripts. It does a lot of self-checking, but still, it is possible that an error could be introduced and go undetected. Therefore, it is essential to make careful backups and to test reformatted scripts.
The main current limitation is that perltidy does not scan modules included with 'use' statements. This makes it necessary to guess the context of any bare words introduced by such modules. Perltidy has good guessing algorithms, but they are not infallible. When it must guess, it leaves a message in the log file.
If you encounter a bug, please report it.
Perltidy indents but does not reformat comments and qw quotes. Perltidy does not in any way modify the contents of here documents or quoted text, even if they contain source code. (You could, however, reformat them separately). Perltidy does not format 'format' sections in any way. And, of course, it does not modify pod documents.
Under the -html option with the default --pod2html flag, a temporary file is required to pass text to Pod::Html. Unix systems will try to use the POSIX tmpnam() function. Otherwise the file perltidy.TMP will be temporarily created in the current working directory.
When standard input is used, the log file, if saved, is perltidy.LOG, and any errors are written to perltidy.ERR unless the -se flag is set. These are saved in the current working directory.
The following file extensions are used by perltidy, and files with these extensions may be overwritten or deleted: .ERR, .LOG, .TEE, and/or .tdy, .html, and .bak, depending on the run type and settings.
Perltidy does not operate on files for which the run could produce a file with a duplicated file extension. These extensions include .LOG, .ERR, .TEE, and perhaps .tdy and .bak, depending on the run type. The purpose of this rule is to prevent generating confusing filenames such as somefile.tdy.tdy.tdy.
An exit value of 0, 1, or 2 is returned by perltidy to indicate the status of the result.
A exit value of 0 indicates that perltidy ran to completion with no error messages.
A non-zero exit value indicates some kind of problem was detected.
An exit value of 1 indicates that perltidy terminated prematurely, usually due to some kind of errors in the input parameters. This can happen for example if a parameter is misspelled or given an invalid value. Error messages in the standard error output will indicate the cause of any problem. If perltidy terminates prematurely then no output files will be produced.
An exit value of 2 indicates that perltidy was able to run to completion but there there are (1) warning messages in the standard error output related to parameter errors or problems and/or (2) warning messages in the perltidy error file(s) relating to possible syntax errors in one or more of the source script(s) being tidied. When multiple files are being processed, an error detected in any single file will produce this type of exit condition.
perlstyle(1), Perl::Tidy(3)
The perltidy binary uses the Perl::Tidy module and is installed when that module is installed. The module name is case-sensitive. For example, the basic command for installing with cpanm is 'cpanm Perl::Tidy'.
This man page documents perltidy version 20240903
The source code repository is at https://github.com/perltidy/perltidy.
To report a new bug or problem, use the "issues" link on this page.
Copyright (c) 2000-2024 by Steve Hancock
This package is free software; you can redistribute it and/or modify it under the terms of the "GNU General Public License".
Please refer to the file "COPYING" for details.
This package is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the "GNU General Public License" for more details.