I/O control

The following parameters concern the files which are read and written.

-h, --help

Show summary of usage and exit.

-o=filename, --outfile=filename

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.]

-st, --standard-output

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.pl
    

This option may only be used if there is just a single input file. The default is -nst or --nostandard-output.

-se, --standard-error-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.

-oext=ext, --output-file-extension=ext

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".

-opath=path, --output-path=path

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 -o=s parameter.

-b, --backup-and-modify-in-place

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.

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 -pbp (--perl-best-practices) flag, then you must put a -nst flag after the -pbp flag because it contains a -st flag as one of its components, which means that output will go to the standard output stream.

-bext=ext, --backup-file-extension=ext

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 errors
    

-w, --warning-output

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.

-q, --quiet

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 -q
    

where "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.

-log, --logfile

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.

-g=n, --logfile-gap=n

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.

-npro --noprofile

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.

-pro=filename or --profile=filename

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=testcfg
    

would 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.

-opt, --show-options

Write a list of all options used to the .LOG file. Please see --dump-options for a simpler way to do this.

-f, --force-read-binary

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.

-ast, --assert-tidy

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.

-asu, --assert-untidy

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.

-sal=s, --sub-alias-list=s

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.

-gal=s, --grep-alias-list=s

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} @list
    

Perltidy 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 reductions
    

Any operators specified with --grep-alias-list are added to this list. The next parameter can be used to remove words from this default list.

-gaxl=s, --grep-alias-exclusion-list=s

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='*'.

Basic Options

--notidy

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.

-i=n, --indent-columns=n

Use n columns per indentation level (default n=4).

-l=n, --maximum-line-length=n

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.

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.

-vmll, --variable-maximum-line-length

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 code
    

When 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).

-wc=n, --whitespace-cycle=n

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 -vmll and -wc=n provides a solution to the problem of displaying arbitrarily deep data structures and code in a finite window, although -wc=n may of course be used without -vmll.

The default is not to use this, which can also be indicated using -wc=0.

Tabs

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 -fws). If you have any tabs in your comments, quotes, or here-documents, they will remain.

-xs, --extended-syntax

A problem with formatting Perl code is that some modules can introduce new syntax. 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 and the braces would not be balanced:

    method deposit( Num $amount) {
        $self->balance( $self->balance + $amount );
    }
    

For one of the extensions, module Switch::Plain, colons are marked as labels. If you use this module, you may want to also use the --nooutdent-labels flag to prevent lines such as 'default:' from being outdented.

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.

For another method of handling extended syntax see the section "Skipping Selected Sections of Code".

-io, --indent-only

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 -noll or -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.

-enc=s, --character-encoding=s

This flag indicates if the input data stream use 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. (Note that perltidy often works on snippets of code rather than complete files so it cannot rely on use utf8 directives).

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.pl
    

or

   perltidy -guess file.pl
    

or simply

   perltidy file.pl
    

since -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.

-eos=s, --encode-output-strings=s

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:

-gcs, --use-unicode-gcstring

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.

-ole=s, --output-line-ending=s

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.

-ple, --preserve-line-endings

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.

-atnl, --add-terminal-newline

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.

-it=n, --iterations=n

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.

-conv, --converge

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.

Code Indentation Control

-ci=n, --continuation-indentation=n

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, -ola, and control 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).

-xci, --extended-continuation-indentation

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
                )
              )
    

-sil=n --starting-indentation-level=n

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.

List indentation using --line-up-parentheses, -lp or --extended--line-up-parentheses , -xlp

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:

-lpil=s, --line-up-parentheses-inclusion-list and -lpxl=s, --line-up-parentheses-exclusion-list

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 -lpil=s to specify the containers to which -lp applies, or

use -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 check on contents is made
  '1' exclude B<-lp> unless the contents is a simple list without sublists
  '2' exclude B<-lp> unless the contents is a simple list without sublists, without
      code blocks, and without ternary operators
    

For example,

  -lpil = 'f(2'
    

means only apply -lp to function call lists which do not contain any sublists, code blocks or ternary expressions.

-cti=n, --closing-token-indentation

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 always
    

The flags -cti=1 and -cti=2 work well with the -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.

-icp, --indent-closing-paren

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.

-icb, --indent-closing-brace

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.

-nib, --non-indenting-braces

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.

-nibp=s, --non-indenting-brace-prefix=s

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 appending pre-pending 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='#\+\+'
    

-olq, --outdent-long-quotes

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 or --nooutdent-long-lines.

-oll, --outdent-long-lines

This command is equivalent to --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.

Outdenting Labels: -ola, --outdent-labels

This command will cause labels to be outdented by 2 spaces (or whatever -ci has been set to), if possible. This is the default. For example:

        my $i;
      LOOP: while ( $i = <FOTOS> ) {
            chomp($i);
            next unless $i;
            fixit($i);
        }
    

Use -nola to not outdent labels. To control line breaks after labels see "bal=n, --break-after-labels=n".

Outdenting Keywords

Whitespace Control

Whitespace refers to the blank space between variables, operators, and other code tokens.

-fws, --freeze-whitespace

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.

Tightness of curly braces, parentheses, and square brackets

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=2
    

When 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=2
    

Curly 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=2
    

And 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=2
    

To 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>.

-tso, --tight-secret-operators

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.

-sts, --space-terminal-semicolon

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)
    

-sfs, --space-for-semicolon

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 ) {    # -nsfs
    

-asc, --add-semicolons

Setting -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.

-dsm, --delete-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).

-aws, --add-whitespace

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 -fws to leave whitespace completely unchanged).

-dws, --delete-old-whitespace

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.

Detailed whitespace controls around tokens

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.

WARNING Be sure to put these tokens in quotes to avoid having them misinterpreted by your command shell.

Note1: Perltidy does always follow whitespace controls

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.

Note2: Perltidy's whitespace rules are not perfect

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.

Space between specific keywords and opening paren

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 and or xor eq ne if else elsif until unless
   while for foreach return switch case given when
    

These 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.

Space between all keywords and opening parens

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.

Space between all function names and opening parens

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 );   # -sfp
    

You will probably also want to use the flag -skp (previous item) too.

The reason this is not recommended is that 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.

-fpva or --function-paren-vertical-alignment

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 );
    

-spp=n or --space-prototype-paren=n

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 paren
    

The 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]
    

-kpit=n or --keyword-paren-inner-tightness=n

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=0
    

In 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 foreach
    

These can be changed with the parameter -kpitl=s described in the next section.

-kpitl=string or --keyword-paren-inner-tightness=string

This command can be used to change the keywords to which the the -kpit=n command 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.

-lop or --logical-padding

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 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;
    

Trimming whitespace around "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.

-sbq=n or --space-backslash-quote=n

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 quote
    

The default is -sbq=1, meaning that a space will be used if there is one in the source code.

Trimming trailing whitespace from lines of POD

-trp or --trim-pod will remove trailing whitespace from lines of POD. The default is not to do this.

Comment Controls

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.

-ibc, --indent-block-comments

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.

-isbc, --indent-spaced-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.

-olc, --outdent-long-comments

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.

-msc=n, --minimum-space-to-comment=n

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.

-fpsc=n, --fixed-position-side-comment=n

This parameter tells perltidy to line up side comments in column number n whenever possible. The default, n=0, will not do this.

-iscl, --ignore-side-comment-lengths

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 well
    

whereas 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 well
    

-hsc, --hanging-side-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 this
    

A 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.

Closing 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 message
    

A 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:

Static Block Comments

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.

    @month_of_year = (    # -sbc (default)
        'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct',
    ##  'Dec', 'Nov'
        'Nov', 'Dec');

    @month_of_year = (   # -nsbc
        'Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct',
        ##  'Dec', 'Nov'
        'Nov', 'Dec'
    );

Static Side Comments

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.

Skipping Selected Sections of Code

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

Additional text may appear on the special comment lines provided that it is separated from the marker by at least one space, 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.

-fs, --format-skipping

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.

-fsb=string, --format-skipping-begin=string

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 #*****
    

-fse=string, --format-skipping-end=string

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.

-cs, --code-skipping

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.

-csb=string, --code-skipping-begin=string

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'.

-cse=string, --code-skipping-end=string

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'.

Line Break Control

The parameters in this section control breaks after non-blank lines of code. Blank lines are controlled separately by parameters in the section "Blank Line Control".

-fnl, --freeze-newlines

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 -noll with this.

Note: If you also want to keep your blank lines exactly as they are, you can use the -fbl flag which is described in the section "Blank Line Control".

-ce, --cuddled-else

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 -cbl, 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 -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 -ce 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.

-cb, --cuddled-blocks

This flag is equivalent to -ce.

-cbl, --cuddled-block-list

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,grep
    

Note 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 -cbl flag by itself merely specifies which blocks are formatted with the cuddled format. It has no effect unless this formatting style is activated with -ce.

-cblx, --cuddled-block-list-exclusive

When cuddled else formatting is selected with -ce, setting this flag causes perltidy to ignore its built-in defaults and rely exclusively on the block types specified on the -cbl flag described in the previous section. For example, to avoid using cuddled catch and finally, which among in the defaults, the following set of parameters could be used:

  perltidy -ce -cbl='else elsif continue' -cblx
    

-cbo=n, --cuddled-break-option=n

Cuddled 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.

-bl, --opening-brace-on-new-line, or --brace-left

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.

-bll=s, --brace-left-list=s

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='*'.

-blxl=s, --brace-left-exclusion-list=s

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.

-sbl, --opening-sub-brace-on-new-line

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'.

For example,

 perltidy -sbl
    

produces this result:

 sub message
 {
    if (!defined($_[0])) {
        print("Hello, World\n");
    }
    else {
        print($_[0], "\n");
    }
 }
    

This flag is negated with -nsbl, which is the default.

-asbl, --opening-anonymous-sub-brace-on-new-line

The flag -asbl is like the -sbl flag except that it applies to anonymous sub's instead of named subs. For example

 perltidy -asbl
    

produces this result:

 $a = sub
 {
     if ( !defined( $_[0] ) ) {
         print("Hello, World\n");
     }
     else {
         print( $_[0], "\n" );
     }
 };
    

This flag is negated with -nasbl, and the default is -nasbl.

-bli, --brace-left-and-indent

The flag -bli is similar to the -bl flag but in addition it causes one unit of continuation indentation ( see -ci ) to be placed before an opening and closing block braces.

For example, perltidy -bli gives

        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.

-blil=s, --brace-left-and-indent-list=s

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'.

-blixl=s, --brace-left-and-indent-exclusion-list=s

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.

-bar, --opening-brace-always-on-right

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 -bl and -bar are specified.

-otr, --opening-token-right and related flags

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-right
    

-bbhb=n, --break-before-hash-brace=n and related flags

When 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 break
    

For example,

    # perltidy -bbhb=3
    $romanNumerals =
      {
        one   => 'I',
        two   => 'II',
        three => 'III',
        four  => 'IV',
      };
    

There are several points to note about this flag:

-bbhbi=n, --break-before-hash-brace-and-indent=n

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 level
    

For 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.

-bbsb=n, --break-before-square-bracket=n

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 break
    

-bbsbi=n, --break-before-square-bracket-and-indent=n

This 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 level
    

-bbp=n, --break-before-paren=n

This 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 break
    

-bbpi=n, --break-before-paren-and-indent=n

This 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 level
    

-wn, --weld-nested-containers

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.

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 -conv 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 -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.

Another limitation is that an opening symbol which delimits quoted text cannot be included in a welded pair. This is because quote delimiters are treated specially in perltidy.

Finally, the stacking of containers defined by this flag have priority over any other container stacking flags. This is because any welding is done first.

-wnxl=s, --weld-nested-exclusion-list

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-bracets, 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 {{  ... }}
    

Vertical tightness of non-block curly braces, parentheses, and square brackets.

These parameters control what shall be called vertical tightness. Here are the main points:

    # 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',
    );

    # 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 ] )
                )
    );

    # perltidy -lp -vt=2 -vtc=2
    $init->add( mysprintf( "(void)find_threadsv(%s);",
                           cstring( $threadsv_names[ $op->targ ] ) ) );

-bbvt=n or --block-brace-vertical-tightness=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 -scbb.

-sot, --stack-opening-tokens and related flags

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-brace
    

The 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.

-sct, --stack-closing-tokens and related flags

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 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-brace
    

The 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 -sot -sct.

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 -sot and -sct flags do.

-dnl, --delete-old-newlines

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.

-anl, --add-newlines

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.

Controlling whether perltidy breaks before or after operators

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 bl 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 xor
    

and 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 -wba and -wbb 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.

bal=n, --break-after-labels=n

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 label
    

For example,

      # perltidy -bal=1
      RETURN:
        return;
      # perltidy -bal=2
      RETURN: return;
    

Controlling List Formatting

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. If we place a side comment anywhere between the opening and closing parens, the original line break points are retained. 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,
    );

The side comment can be a single hash symbol without any text. 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.

For another possibility see the -fs flag in "Skipping Selected Sections of Code".

-boc, --break-at-old-comma-breakpoints

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.

-mft=n, --maximum-fields-per-table=n

If the computed number of fields for any table exceeds n, then it will be reduced to n. The default value for n is a large number, 40. While this value should probably be left unchanged as a general rule, it 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 single comment could be introduced somewhere to freeze the formatting in future applications of perltidy.

    # perltidy -mft=2
    @month_of_year = (
        'Jan', 'Feb',
        'Mar', 'Apr',
        'May', 'Jun',
        'Jul', 'Aug',
        'Sep', 'Oct',
        'Nov', 'Dec'
    );
    

-cab=n, --comma-arrow-breakpoints=n

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
    );
    

Retaining or Ignoring Existing Line Breaks

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.

-bol, --break-at-old-logical-breakpoints

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('|');
    

-bom, --break-at-old-method-breakpoints

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 preformatted 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.

-bos, --break-at-old-semicolon-breakpoints

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.

-bok, --break-at-old-keyword-breakpoints

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.

-bot, --break-at-old-ternary-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.

-boa, --break-at-old-attribute-breakpoints

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.

Keeping old breakpoints at specific token types

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.

-iob, --ignore-old-breakpoints

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.

-kis, --keep-interior-semicolons

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 Line Control

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.

-fbl, --freeze-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).

-bbc, --blanks-before-comments

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.

-blbs=n, --blank-lines-before-subs=n

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.

-blbp=n, --blank-lines-before-packages=n

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.

-bbs, --blanks-before-subs

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.

-bbb, --blanks-before-blocks

A blank line will be introduced before blocks of coding delimited by for, foreach, while, until, and if, unless, in the following circumstances:

-lbl=n --long-block-line-count=n

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.

-blao=i or --blank-lines-after-opening-block=i

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.

-blbc=i or --blank-lines-before-closing-block=i

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).

-blaol=s or --blank-lines-after-opening-block-list=s

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.

-blbcl=s or --blank-lines-before-closing-block-list=s

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.

Note on using the -blao and -blbc options.

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='*' filename
    

Now 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 -kbl=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.

-mbl=n --maximum-consecutive-blank-lines=n

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 -blbp and -blbs parameters. If n=0 then no blank lines will be output (unless all old blank lines are retained with the -kbl=2 flag of the next section).

This flag obviously does not apply to pod sections, here-documents, and quotes.

-kbl=n, --keep-old-blank-lines=n

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> flag
    

The default is n=1.

-sob, --swallow-optional-blank-lines

This is equivalent to kbl=0 and is included for compatibility with previous versions.

-nsob, --noswallow-optional-blank-lines

This is equivalent to kbl=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.

Styles

A style refers to a convenient collection of existing parameters.

-gnu, --gnu-style

-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 -icp
    

To use this style with -xlp instead of -lp use -gnu -xlp.

-pbp, --perl-best-practices

-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 and -se 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 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.

One-line blocks

There are a few points to note regarding one-line blocks. A one-line block is something like this,

    if ( -e $file ) { print "'$file' exists\n" }
    

where the contents within the curly braces is short enough to fit on a single line.

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 one-line block style of the input file.

If an existing one-line block is longer than the maximum line length, however, it will be broken into multiple lines. When this happens, perltidy checks for and adds any optional terminating semicolon (unless the -nasc option is used) if the block is a code block.

The main exception is that perltidy will attempt to form new one-line blocks following the keywords "map", "eval", and "sort", because these code blocks are often small and most clearly displayed in a single line.

One-line block rules can conflict with the cuddled-else option. When the cuddled-else option is used, perltidy retains existing one-line blocks, even if they do not obey cuddled-else formatting.

Occasionally, when one-line blocks get broken because they exceed the available line length, the formatting will violate the requested brace style. If this happens, reformatting the script a second time should correct the problem.

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 >outfile
    

When 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.

-olbs=n, --one-line-block-semicolons=n

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 blocks
    

Note 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.

-olbn=n, --one-line-block-nesting=n

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 intact
    

For 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.

Controlling Vertical Alignment

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'
        );

Completely turning off vertical alignment with -novalign

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.

Controlling code alignment with --valign-code or -vc

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.

Controlling side comment alignment with --valign-side-comments or -vsc

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

Controlling block comment alignment with --valign-block-comments or -vbc

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.

Finer alignment control with --valign-exclusion-list=s or -vxl=s and --valign-inclusion-list=s or -vil=s

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 until
    

These 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';    # unused
    

To 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';    # unused
    

To 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';    # unused
    

Notice 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.

Other Controls

Deleting selected text

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).

Writing selected text to a file

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.

Using a .perltidyrc command file

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. Instructions for Windows 95 and later versions can be found here:

http://www.netmanage.com/000/20021101_005_tcm21-6336.pdf

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 tight
    

The 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 -ss
    

There are several options may be helpful in debugging a .perltidyrc file:

Creating a new abbreviation

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).

Skipping leading non-perl commands with -x or --look-for-hash-bang

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.

Making a file unreadable

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.new
    

This 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 -dac to reduce the file size of a perl script.

Debugging

The following flags are available for debugging:

--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-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 useful during program development, when large numbers of files with varying sizes are processed, because it can reduce virtual memory usage.

--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=20
    

This only applies to files specified by filename on the command line.

--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.

-DEBUG will write a file with extension .DEBUG for each input file showing the tokenization of all lines of code.

Working with MakeMaker, AutoLoader and SelfLoader

The first $VERSION line of a file which might be eval'd by MakeMaker is passed through unchanged except for indentation. 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. 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. Use --nolook-for-selfloader, or -nlsl, to deactivate this feature.

Working around problems with older version of Perl

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.

Specifying Block Types

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='*'

Specifying File Extensions

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~ .