Getopt::ArgvFile - interpolates script options from files into @ARGV or another array

This manual describes version 1.06.

  # load the module
  use Getopt::ArgvFile qw(argvFile);

  # load another module to evaluate the options, e.g.:
  use Getopt::Long;
  ...

  # solve option files
  argvFile;

  # evaluate options, e.g. this common way:
  GetOptions(%options, 'any');

If options should be processed into another array, this can be done this way:

  # prepare target array
  my @options=('@options1', '@options2', '@options3');

  ...

  # replace file hints by the options stored in the files
  argvFile(array=>\@options);

This module simply interpolates option file hints in CW@ARGV by the contents of the pointed files. This enables option reading from files instead of or additional to the usual reading from the command line.

Alternatively, you can process any array instead of CW@ARGV which is used by default and mentioned mostly in this manual.

The interpolated CW@ARGV could be subsequently processed by the usual option handling, e.g. by a Getopt::xxx module. Getopt::ArgvFile does not perform any option handling itself, it only prepares the array CW@ARGV.

Option files can significantly simplify the call of a script. Imagine the following:

Breaking command line limits

A script may offer a lot of options, with possibly a few of them even taking parameters. If these options and their parameters are passed onto the program call directly, the number of characters accepted by your shells command line may be exceeded. Perl itself does not limit the number of characters passed to a script by parameters, but the shell or command interpreter often sets a limit here. The same problem may occur if you want to store a long call in a system file like crontab. If such a limit restricts you, options and parameters may be moved into option files, which will result in a shorter command line call.

Script calls prepared by scripts

Sometimes a script calls another script. The options passed onto the nested script could depend on variable situations, such as a users input or the detected environment. In such a case, it can be easier to generate an intermediate option file which is then passed to the nested script. Or imagine two cron jobs one preparing the other: the first may generate an option file which is then used by the second.

Simple access to typical calling scenarios

If several options need to be set, but in certain circumstances are always the same, it could become sligthly nerveracking to type them in again and again. With an option file, they can be stored once and recalled easily as often as necessary. Further more, option files may be used to group options. Several settings may set up one certain behaviour of the program, while others influence another. Or a certain set of options may be useful in one typical situation, while another one should be used elsewhere. Or there is a common set of options which has to be used in every call, while other options are added depending on the current needs. Or there are a few user groups with different but typical ways to call your script. In all these cases, option files may collect options belonging together, and may be combined by the script users to set up a certain call. In conjunction with the possiblity to nest such collections, this is perhaps the most powerful feature provided by this method.

Individual and installationwide default options

The module allows the programmer to enable user setups of default options; for both individual users or generally all callers of a script. This is especially useful for administrators who can configure the default behaviour of a script by setting up its installationwide startup option file. All script users are free then to completely forget every already configured setup option. And if one of them regularly adds certain options to every call, he could store them in his individual startup option file. For example, I use this feature to make my scripts both flexible and usable. I have several scripts accessing a database via DBI. The database account parameters as well as the DBI startup settings should not be coded inside the scripts because this is not very flexible, so I implemented them by options. But on the other hand, there should be no need for a normal user to pass all these settings to every script call. My solution for this is to use default option files set up and maintained by an administrator. This is very transparent, most of the users know nothing of these (documented ;-) configuration settings ... and if anything changes, only the option files have to be adapted.

No symbol is exported by default, but you may explicitly import the argvFile() function.

Example:

  use Getopt::ArgvFile qw(argvFile);

Scans the command line parameters (stored in CW@ARGV or an alternatively passed array) for option file hints (see Basics below), reads the pointed files and makes their contents part of the source array (@ARGV by default) replacing the hints.

Because the function was intentionally designed to work on CW@ARGV and this is still the default behaviour, this manual mostly speaks about CW@ARGV. Please note that it is possible to process any other array as well.

Basics

An option file hint is simply the filename preceeded by (at least) one @ character:

  > script -optA argA -optB @optionFile -optC argC

This will cause argvFile() to scan optionFile for options. The element @optionFile will be removed from the CW@ARGV array and will be replaced by the options found.

Note: you can choose another prefix by using the prefix parameter, see below.

An option file which cannot be found is quietly skipped.

Well, what is within an option file? It is intended to store command line arguments which should be passed to the called script. They can be stored exactly as they would be written in the command line, but may be spread to multiple lines. To make the file more readable, space and comment lines (starting with a #) are allowed additionally. POD comments are supported as well. For example, the call

  > script -optA argA -optB -optC cArg par1 par2

could be transformed into

  > script @scriptOptions par1 par2

where the file scriptOptions may look like this:

  # option a
  -optA argA

  =pod
  option b
  =cut
  -optB

  # option c
  -optC cArg

Nested option files

Option files can be nested. Recursion is avoided globally, that means that every file will be opened only once (the first time argvFile() finds a hint pointing to it). This is the simplest implementation, indeed, but should be suitable. (Unfortunately, there are LIMITS.)

By using this feature, you may combine groups of typical options into a top level option file, e.g.:

  File ab:

  # option a
  -optA argA
  # option b
  -optB

  File c:

  # option c
  -optC cArg

  File abc:

  # combine ab and c
  @ab @c

If anyone provides these files, a user can use a very short call:

  > script @abc

and argvFile() will recursively move all the filed program parameters into CW@ARGV.

Startup support

By setting several named parameters, you can enable automatic processing of startup option files. There are three of them:

The default option file is searched in the installation path of the calling script, the home option file is searched in the users home (evaluated via environment variable HOME), and the current option script is searched in the current directory.

By default, all startup option files are expected to be named like the script, preceeded by a dot, but this can be adapted to individual needs if preferred, see below.

 Examples:
  If a script located in "/path/script" is invoked in directory
  /the/current/dir by a user "user" whoms "HOME" variable points
  to "/homes/user", the following happens:

  argvFile()                    # ignores all startup option files;
  argvFile(default=>1)          # searches and expands "/path/.script",
                                # if available (the "default" settings);
  argvFile(home=>1)             # searches and expands "/homes/user/.script",
                                # if available (the "home" settings);
  argvFile(current=>1)          # searches and expands "/the/current/dir/.script",
                                # if available (the "current" settings);
  argvFile(
           default => 1,
           home    => 1,
           current => 1
          )                     # tries to handle all startups.

Any true value will activate the setting it is assigned to.

In case the .script name rule does not meet your needs or does not fit into a certain policy, the expected startup filenames can be set up by an option CWstartupFilename. The option value may be a scalar used as the expected filename, or a reference to code returning the name. Such code will be called once and will receive the name of the script.

  # use ".config"
  argvFile(startupFilename => '.config');

  # emulate the default behaviour,
  # but use an extra dot postfix
  my $nameBuilder=sub {join('', '.', basename($_[0]), '.');}
  argvFile(startupFilename => $nameBuilder);

The contents found in a startup file is placed before all explicitly set command line arguments. This enables to overwrite a default setting by an explicit option. If all startup files are read, current startup files can overwrite home files which have preceedence over default ones, so that the default startups are most common. In other words, if the module would not support startup files, you could get the same result with script @/path/.script @/homes/user/.script @/the/current/dir/.script.

Note: There is one certain case when overwriting will not work completely because duplicates are sorted out: if all three types of startup files are used and the script is started in the installation directory, the default file will be identical to the current file. The default file is processed, but the current file is skipped as a duplicate later on and will not overwrite settings made caused by the intermediately processed home file. If started in another directory, it will overwrite the home settings. But the alternative seems to be even more confusing: the script would behave differently if just started in its installation path. Because a user might be more aware of configuration editing then of the current path, I choose the current implementation, but this preceedence might become configurable in a future version.

If there is no HOME environment variable, the home setting takes no effect to avoid trouble accessing the root directory.

Cascades

The function supports multi-level (or so called cascaded) option files. If a filename in an option file hint starts with a @ again, this complete name is the resolution written back to CW@ARGV - assuming there will be another utility reading option files.

 Examples:
  @rfile          rfile will be opened, its contents is
                  made part of @ARGV.
  @@rfile         cascade: "@rfile" is written back to
                  @ARGV assuming that there is a subsequent
                  tool called by the script to which this
                  hint will be passed to solve it by an own
                  call of argvFile().

The number of cascaded hints is unlimited.

Processing an alternative array

However the function was designed to process CW@ARGV, it is possible to process another array as well if you prefer. To do this, simply pass a reference to this array by parameter array.

 Examples:
  argvFile()                    # processes @ARGV;
  argvFile(array=>\@options);   # processes @options;

Choosing an alternative hint prefix

By default, @ is the prefix used to mark an option file. This can be changed by using the optional parameter prefix:

 Examples:
  argvFile();                   # use "@";
  argvFile(prefix=>'~');        # use "~";

Note that the strings #, =, - and + are reserved and cannot be chosen here because they are used to start plain or POD comments or are typically option prefixes.

If a script calling CWargvFile() with the CWdefault switch is invoked using a relative path, it is strongly recommended to perform the call of CWargvFile() in the startup directory because CWargvFile() then uses the relative script path as well.

If an option file does not exist, argvFile() simply ignores it. No message will be displayed, no special return code will be set.

Jochen Stenzel <mailto://perl@jochen-stenzel.de>

Copyright (c) 1993-2002 Jochen Stenzel. All rights reserved.

This program is free software, you can redistribute it and/or modify it under the terms of the Artistic License distributed with Perl version 5.003 or (at your option) any later version. Please refer to the Artistic License that came with your Perl distribution for more details.