NAME
Win32::CommandLine - Retrieve and reparse the Win32 command line
VERSION
$Win32::CommandLine::VERSION = '0.954'; # ( Win32-CommandLine-0.954 )
SYNOPSIS
@ARGV = Win32::CommandLine::argv() if eval { require Win32::CommandLine; };
or
use Win32::CommandLine qw( command_line parse );
my $commandline = command_line();
my @args = parse( $commandline );
DESCRIPTION
This module provides a simple way for any perl script to reread and
reparse the windows command line, adding improved parsing and more
robust quote mechanics, augmented with powerful bash-like shell
enhancements (including brace and tilde expansion, extended file glob
expansion, and subshell command substitution).
Use of the companion script, "xx.bat" (along with "doskey"), can,
transparently, grant those same features to the command line interface
of *any* windows executable.
Note that bash-compatible globbing and argument expansion are supplied,
including command substitution. Glob patterns may also contain
meta-notations, such as '"a[bc]*"' or '"foo.{bat,pl,exe,o}"'.
Quote mechanics/expansion and subshell command substitution
'...' literal (no escapes and no globbing within quotes)
"..." literal (no escapes and no globbing within quotes) (see *NOTE-1)
$'...' string including all ANSI C string escapes (see *NOTE-2); no globbing within quotes
$"..." literal (no escapes and no globbing within quotes) [same as "..."]
$( ... ) command substitution (see *NOTE-3)
$("...") command substitution (quotes removed; see *NOTE-4)
NOTE-1: DOS character escape sequences (such as "\"") are parsed prior
to being put into the command line and, so, are valid and still
(unavoidably) interpreted within double-quotes.
NOTE-2: ANSI C string escapes are "\a", "\b", "\e", "\f", "\n", "\r",
"\t", "\v", "\\", "\'", "\"", "\[0-9]{1,3}", "\x[0-9a-fA-F]{1,2}",
"\c[@A-Z[\\\]^_`?" ; all other escaped characters are left in place
without transformation ("\<x>" => "\<x>").
NOTE-3: Command substitution replaces the "$(...)" argument with the
standard output of that argument's execution. Command substitution
strings are not, themselves, automatically expanded; use '"$(xx
*COMMAND*)"' to trigger expansion of the subshell command line.
NOTE-4: "$("...")" is present to enable delayed DOS/Windows
interpretation of redirection & continuation characters. This allows
redirection & continuation characters to be used within the subshell
command string.
ref: bash ANSI-C Quoting
<http://www.gnu.org/software/bash/manual/html_node/ANSI_002dC-Quoting.ht
ml>"@" <http://www.webcitation.org/66M8skmP8>
Expansion and "glob" meta-characters
\ Quote the next metacharacter
[] Character class
{} Multiple pattern
* Match any string of characters
? Match any single character
~ Current user home directory
~USERNAME Home directory of USERNAME
~TEXT Environment variable named ~TEXT (aka $ENV{~TEXT}) [overrides ~USERNAME expansion]
The multiple pattern meta-notation '"a{b,c,d}e"' is a shorthand for
'"abe ace ade"'. Left to right order is preserved, with results of
matches being sorted separately at a low level to preserve this order.
"xx.bat" Usage
doskey type=call xx type $*
type [a-c]*.pl
doskey perl=call xx perl $*
perl -e 'print "test"' &@:: would otherwise FAIL
doskey cpan=call xx cpan $*
cpan $(dzil listdeps) &@:: with a CPAN wrapper program
@:: print all files in current directory [appropriately quoted for the CMD shell]
xx -e *
@:: * assumes `ls` is installed
@:: print all directories in current directory
xx echo $(" ls -ALp --quoting-style=c --color=no . | grep --color=no "[\\/]$" ")
@:: print all files (non-directories) in current directory
xx echo $(" ls -ALp --quoting-style=shell --color=no . | grep --color=no -v "[\\/]$" ")
INSTALLATION
To install this module, run the following commands:
perl Build.PL
perl Build
perl Build test
perl Build install
This is minor modification of the usual perl build idiom. This version
is portable across multiple platforms.
Alternatively, the standard make idiom is also available (although it is
deprecated):
perl Makefile.PL
make
make test
make install
On Windows platforms, when using this make idiom, replace ""make"" with
the result of '"perl -MConfig -e "print $Config{make}""' (usually,
either "dmake", "gmake", or "nmake").
Note that the Makefile.PL script is just a pass-through, and
Module::Build is still ultimately required for installation. Makefile.PL
will throw an exception if Module::Build is missing from your current
installation. "cpan" will notify the user of the build prerequisites
(and install them for the build, if it is setup to do so [see the cpan
configuration option "build_requires_install_policy"
<https://metacpan.org/pod/CPAN#Config-Variables>]).
PPM installation bundles should also be available in the standard PPM
repositories (eg, ActiveState <http://code.activestate.com/ppm>, etc.).
Note: for ActivePerl installations, '"perl ./Build install"' will do a
full installation using "ppm" <https://metacpan.org/pod/PPM>. During the
installation, a PPM package is constructed locally and then subsequently
used for the final module install. This allows for uninstalls (by using
'"ppm uninstall Win32::CommandLine"') and also keeps local HTML
documentation current.
INTERFACE
"command_line( ) => $"
* "$" : [return] the original command line for the process (as a
string)
Use the Win32 API to recapture the original command line for the current
process.
my $commandline = command_line();
"argv( [\%options] ) => @ARGS"
* "\%options" : (optional) reference to hash containing function
options
* @ARGS : [return] revised argument array (may replace @ARGV)
Reparse & glob-expand the original command line, returning a new,
revised argument array (which is a drop-in replacement for @ARGV).
@ARGV = argv();
"parse( $s [,\%options ] ) => @ARGS"
* $s : string argument to parse/expand
* "\%options" : (optional) reference to hash containing function
options
* @ARGS : [return] parsed/expanded arguments
my @argv_new = parse( command_line() );
Parse & glob-expand a string argument; returns the results of
parsed/expanded argument as an array.
Function options ( "\%options" )
my %options = (
remove_exe_prefix => 1, # = 0/<true> [default = true] # if true, remove all initial args up to and including the exe name from the @args array
dosquote => 0, # = 0/<true>/'all' [default = 0] # if true, convert all non-globbed ARGS to DOS/Win32 CLI compatible tokens (escaping internal quotes and quoting whitespace and special characters)
dosify => 0, # = 0/<true>/'all' [default = 0] # if true, convert all _globbed_ ARGS to DOS/Win32 CLI compatible tokens (escaping internal quotes and quoting whitespace and special characters); 'all' => do so for for _all_ ARGS which are determined to be files
unixify => 0, # = 0/<true>/'all' [default = 0] # if true, convert all _globbed_ ARGS to UNIX path style; 'all' => do so for for _all_ ARGS which are determined to be files
nullglob => defined($ENV{nullglob}) ? $ENV{nullglob} : 0, # = 0/<true> [default = 0] # if true, patterns which match no files are expanded to a null string (no token), rather than the pattern itself ## $ENV{nullglob} (if it exists) overrides the default
glob => 1, # = 0/<true> [default = true] # when true, globbing is performed
## ToDO: rework this ... need carp/croak on unbalanced quotes/subshells (? carp_ub_quotes, carp_ub_shells, carp = 0/1/warn/carp/die/croak)
croak_unbalanced => 1, # = 0/true/'quotes'/'subshells' [default = true] # if true, croak for unbalanced command line quotes or subshell blocks (takes precedence over carp_unbalanced)
carp_unbalanced => 1, # = 0/true/'quotes'/'subshells' [default = true] # if true, carp for unbalanced command line quotes or subshell blocks
## ToDO: add globstar option
);
RATIONALE
This began as a simple need to work-around the less-than-stellar
"COMMAND.COM"/"CMD.EXE" command line parser, just to accomplish more
"correct" quotation interpretation. It then grew into a small odyssey:
learning XS and how to create a perl module, learning the perl build
process and creating a customized build script/environment, researching
tools and developing methods for revision control and versioning,
learning and creating perl testing processes, and finally learning about
PAUSE and perl publishing practices. And, somewhere in the middle,
adding some of the "bash" shell magic to the CMD shell.
Some initial attempts were made using "Win32::API"
<https://metacpan.org/pod/Win32::API> and "Inline::C"
<https://metacpan.org/pod/Inline::C>. For example, a "Win32::API"
attempt (which caused GPFs):
@rem = '--*-Perl-*--
@echo off
if "%OS%" == "Windows_NT" goto WinNT
perl -x -S "%0" %1 %2 %3 %4 %5 %6 %7 %8 %9
goto endofperl
:WinNT
perl -x -S %0 %*
if NOT "%COMSPEC%" == "%SystemRoot%\system32\cmd.exe" goto endofperl
if %errorlevel% == 9009 echo You do not have Perl in your PATH.
if errorlevel 1 goto script_failed_so_exit_with_non_zero_val 2>nul
goto endofperl
@rem ';
#!/usr/bin/perl -w
#line 15
#
use Win32::API;
#
Win32::API->Import("kernel32", "LPTSTR GetCommandLine()");
my $string = pack("Z*", GetCommandLine());
#
print "string[".length($string)."] = '$string'\n";
# ------ padding --------------------------------------------------------------------------------------
__END__
:endofperl
Unfortunately, "Win32::API" and "Inline::C" were shown to be too fragile
at the time (in 2007). "Win32::API" caused occasional (but reproducible)
GPFs, and "Inline::C" was shown to be very brittle on Win32 systems
(i.e., not compensating for paths with embedded strings). (See
<http://www.perlmonks.org/?node_id=625182> for a more full explanation
of the problem and initial attempts at a solution.)
So, an initial XS solution was implemented. And from that point, the
lure of "bash"-like command line parsing led slowly, but inexorably, to
the full implementation. The parsing logic is unfortunately still
complex, but seems to be holding up well under testing.
IMPLEMENTATION and INTERNALS
This is a list of internal XS functions (brief descriptions will be
added at a later date):
SV * _wrap_GetCommandLine() // [XS] Use C and Win32 API to get the command line
HANDLE _wrap_CreateToolhelp32Snapshot ( dwFlags, th32ProcessID )
bool _wrap_Process32First ( hSnapshot, lppe )
bool _wrap_Process32Next ( hSnapshot, lppe )
bool _wrap_CloseHandle ( hObject )
// Pass useful CONSTANTS back to perl
int _const_MAX_PATH ()
HANDLE _const_INVALID_HANDLE_VALUE ()
DWORD _const_TH32CS_SNAPPROCESS ()
// Pass useful sizes back to Perl (for testing) */
unsigned int _info_SIZEOF_HANDLE ()
unsigned int _info_SIZEOF_DWORD ()
// Pass PROCESSENTRY32 structure info back to Perl
SV * _info_PROCESSENTRY32 ()
CONFIGURATION and ENVIRONMENT
"Win32::CommandLine" requires no configuration files or environment
variables.
OPTIONAL Environment Variables
"NULLGLOB"
Override the default glob expansion behavior for empty matches
$ENV{NULLGLOB} = 1; # undef/0 | <true>
Default glob expansion, as in bash, expands glob patterns which match
nothing into the glob pattern itself. Use $ENV{NULLGLOB} to override
this default behavior.
Analogous to the bash command '"shopt -s nullglob"', when $ENV{NULLGLOB}
is set to a true (non-NULL, non-zero) value, a glob expansion which
matches nothing will expand to the null string (aka, "q{}").
Note: the default glob expansion behavior can also be modified
programmatically via the function option, "nullglob", when passed to the
argv() and parse() functions. This option, when passed to "argv()" or
"parse()", will override both the default behavior *and* the
$ENV{NULLGLOB} setting.
DEPENDENCIES
"Win32::CommandLine" requires "Carp::Assert" for internal error checking
and warnings.
The optional modules "Win32", "Win32::Security::SID", and
"Win32::TieRegistry" are recommended to allow full glob tilde expansions
for user home directories (eg, "~administrator" expands to
"C:\Users\Administrator"). Expansion of the single tilde ("~") has a
backup implementation based on %ENV variables, and therefore will still
work even without the optional modules.
INCOMPATIBILITIES
None reported.
CAVEATS
Operational Notes
IMPORTANT NOTE: Special shell characters (shell redirection, '|', '<',
'>', and continuation, '&') must be DOUBLE-quoted to escape shell
interpretation (eg, "foo | bar"). The shell does initial parsing and
redirection/continuation (stripping away everything after I/O
redirection and continuation characters) before any process can get a
look at the command line. So, the special shell characters can only be
hidden from shell interpretation by quoting them with double-quote
characters.
"%<X>%" is also replaced by the corresponding environment variable by
the shell before handing the command line off to the OS. The caret "^"
escape character can be used to break the interpretation when needed
(eg, "%^COMSPEC^%" instead of "%COMSPEC%").
Brackets ('{' and '}') and braces ('[' and ']') must be quoted (single
or double quotes) to be matched literally. This may be a gotcha for some
users, although if the filename has internal spaces, tab expansion of
filenames for the standard Win32 shell (cmd.exe) or 4NT/TCC/TCMD will
automatically surround the entire path with spaces (which corrects the
issue).
Some programs may expect their arguments to maintain their surrounding
quotes, but "argv()" parsing only quotes arguments which require it to
maintain equivalence for shell parsing (i.e., those containing spaces,
special characters, etc). And, since single quotes have no special
meaning to the shell, all arguments which require quoting for correct
shell interpretation will be quoted with double-quote characters, even
if they were originally quoted with single-quotes. Neither of these
issues should be a problem for programs using "Win32::CommandLine", but
may be an issue for 'legacy' applications which have their command line
expanded with "xx.bat".
Be careful with backslashed quotes within quoted strings. Note that
"foo\" is an unbalanced string containing a double quote. Place the
backslash outside of the quotation ("foo"\) or use a double backslash
within ("foo\\") to include the backslash it in the parsed token.
However, backslashes ONLY need to be doubled when placed prior to a
quotation mark ("foo\bar" will work as expected).
SUPPORT
Bugs / Feature Requests
Please report any issues through the issue tracker at
<https://github.com/rivy/perl.Win32-CommandLine/issues>. The developers
will be notified, and you'll automatically be notified of progress on
your issue.
Documentation
You can find documentation for this module with the perldoc command:
perldoc Win32::CommandLine
Further information
* MetaCPAN / CPAN module documentation
<https://metacpan.org/pod/Win32::CommandLine>
<http://search.cpan.org/~rivy/Win32-CommandLine>
* Issue tracker
<https://github.com/rivy/perl.Win32-CommandLine/issues>
* CPAN Ratings
<https://cpanratings.perl.org/dist/Win32-CommandLine>
* CPANTESTERS: Test results
<https://www.cpantesters.org/distro/W/Win32-CommandLine.html>
<http://matrix.cpantesters.org/?dist=Win32-CommandLine+0.954>
<http://fast-matrix.cpantesters.org/?dist=Win32-CommandLine+0.954>
* CPANTS: CPAN Testing Service module summary
<https://cpants.cpanauthors.org/dist/Win32-CommandLine>
ACKNOWLEDGEMENTS
Thanks to BrowserUK and syphilis (aka SISYPHUS on CPAN) for some helpful
ideas (including an initial XS starting point for the module) during a
discussion on PerlMonks <http://www.perlmonks.org/?node_id=625151>.
AUTHOR
Roy Ivy III <rivy@cpan.org>
COPYRIGHT
Copyright (c) 2007-2018, Roy Ivy III <rivy@cpan.org>. All rights reserved.
LICENSE
This module is free software; you can redistribute it and/or modify it
under the Perl Artistic License v2.0
<http://opensource.org/licenses/artistic-license-2.0.php>.
DISCLAIMER OF WARRANTY
THIS PACKAGE IS PROVIDED BY THE COPYRIGHT HOLDER AND CONTRIBUTORS "AS
IS" AND WITHOUT ANY EXPRESS OR IMPLIED WARRANTIES. THE IMPLIED
WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR
NON-INFRINGEMENT ARE DISCLAIMED TO THE EXTENT PERMITTED BY YOUR LOCAL
LAW. UNLESS REQUIRED BY LAW, NO COPYRIGHT HOLDER OR CONTRIBUTOR WILL BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
ARISING IN ANY WAY OUT OF THE USE OF THE PACKAGE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
[REFER TO THE FULL LICENSE FOR EXPLICIT DEFINITIONS OF ALL TERMS.]