1 # -*- mode: cperl; tab-width: 8; indent-tabs-mode: nil; basic-offset: 2 -*-
2 # vim:ts=8:sw=2:et:sta:sts=2
3 package Module::Metadata;
5 # Adapted from Perl-licensed code originally distributed with
6 # Module-Build by Ken Williams
8 # This module provides routines to gather information about
9 # perl modules (assuming this may be expanded in the distant
10 # parrot future to look at other types of modules).
13 use vars qw($VERSION);
14 $VERSION = '1.000009';
15 $VERSION = eval $VERSION;
22 if ($INC{'Log/Contextual.pm'}) {
23 Log::Contextual->import('log_info');
25 *log_info = sub (&) { warn $_[0]->() };
28 use File::Find qw(find);
30 my $V_NUM_REGEXP = qr{v?[0-9._]+}; # crudely, a v-string or decimal
32 my $PKG_REGEXP = qr{ # match a package declaration
33 ^[\s\{;]* # intro chars on a line
34 package # the word 'package'
36 ([\w:]+) # a package name
37 \s* # optional whitespace
38 ($V_NUM_REGEXP)? # optional version number
39 \s* # optional whitesapce
40 [;\{] # semicolon line terminator or block start (since 5.16)
43 my $VARNAME_REGEXP = qr{ # match fully-qualified VERSION name
44 ([\$*]) # sigil - $ or *
46 ( # optional leading package name
47 (?:::|\')? # possibly starting like just :: (Ì la $::VERSION)
48 (?:\w+(?:::|\'))* # Foo::Bar:: ...
54 my $VERS_REGEXP = qr{ # match a VERSION definition
56 \(\s*$VARNAME_REGEXP\s*\) # with parens
58 $VARNAME_REGEXP # without parens
61 =[^=~] # = but not ==, nor =~
67 my $filename = File::Spec->rel2abs( shift );
69 return undef unless defined( $filename ) && -f $filename;
70 return $class->_init(undef, $filename, @_);
77 return undef unless defined($handle) && defined($filename);
78 $filename = File::Spec->rel2abs( $filename );
80 return $class->_init(undef, $filename, @_, handle => $handle);
90 $props{inc} ||= \@INC;
91 my $filename = $class->find_module_by_name( $module, $props{inc} );
92 return undef unless defined( $filename ) && -f $filename;
93 return $class->_init($module, $filename, %props);
98 my $compare_versions = sub {
99 my ($v1, $op, $v2) = @_;
100 $v1 = version->new($v1)
101 unless UNIVERSAL::isa($v1,'version');
103 my $eval_str = "\$v1 $op \$v2";
104 my $result = eval $eval_str;
105 log_info { "error comparing versions: '$eval_str' $@" } if $@;
110 my $normalize_version = sub {
112 if ( $version =~ /[=<>!,]/ ) { # logic, not just version
113 # take as is without modification
115 elsif ( ref $version eq 'version' ) { # version objects
116 $version = $version->is_qv ? $version->normal : $version->stringify;
118 elsif ( $version =~ /^[^v][^.]*\.[^.]+\./ ) { # no leading v, multiple dots
119 # normalize string tuples without "v": "1.2.3" -> "v1.2.3"
120 $version = "v$version";
128 # separate out some of the conflict resolution logic
130 my $resolve_module_versions = sub {
131 my $packages = shift;
133 my( $file, $version );
135 foreach my $p ( @$packages ) {
136 if ( defined( $p->{version} ) ) {
137 if ( defined( $version ) ) {
138 if ( $compare_versions->( $version, '!=', $p->{version} ) ) {
139 $err .= " $p->{file} ($p->{version})\n";
141 # same version declared multiple times, ignore
145 $version = $p->{version};
148 $file ||= $p->{file} if defined( $p->{file} );
152 $err = " $file ($version)\n" . $err;
167 croak "provides() requires key/value pairs \n" if @_ % 2;
170 croak "provides() takes only one of 'dir' or 'files'\n"
171 if $args{dir} && $args{files};
173 croak "provides() requires a 'version' argument"
174 unless defined $args{version};
176 croak "provides() does not support version '$args{version}' metadata"
177 unless grep { $args{version} eq $_ } qw/1.4 2/;
179 $args{prefix} = 'lib' unless defined $args{prefix};
183 $p = $class->package_versions_from_directory($args{dir});
186 croak "provides() requires 'files' to be an array reference\n"
187 unless ref $args{files} eq 'ARRAY';
188 $p = $class->package_versions_from_directory($args{files});
191 # Now, fix up files with prefix
192 if ( length $args{prefix} ) { # check in case disabled with q{}
193 $args{prefix} =~ s{/$}{};
194 for my $v ( values %$p ) {
195 $v->{file} = "$args{prefix}/$v->{file}";
202 sub package_versions_from_directory {
203 my ( $class, $dir, $files ) = @_;
212 push @files, $_ if -f $_ && /\.pm$/;
218 # First, we enumerate all packages & versions,
219 # separating into primary & alternative candidates
221 foreach my $file (@files) {
222 my $mapped_filename = File::Spec->abs2rel( $file, $dir );
223 my @path = split( /\//, $mapped_filename );
224 (my $prime_package = join( '::', @path )) =~ s/\.pm$//;
226 my $pm_info = $class->new_from_file( $file );
228 foreach my $package ( $pm_info->packages_inside ) {
229 next if $package eq 'main'; # main can appear numerous times, ignore
230 next if $package eq 'DB'; # special debugging package, ignore
231 next if grep /^_/, split( /::/, $package ); # private package, ignore
233 my $version = $pm_info->version( $package );
235 if ( $package eq $prime_package ) {
236 if ( exists( $prime{$package} ) ) {
237 croak "Unexpected conflict in '$package'; multiple versions found.\n";
239 $prime{$package}{file} = $mapped_filename;
240 $prime{$package}{version} = $version if defined( $version );
243 push( @{$alt{$package}}, {
244 file => $mapped_filename,
251 # Then we iterate over all the packages found above, identifying conflicts
252 # and selecting the "best" candidate for recording the file & version
254 foreach my $package ( keys( %alt ) ) {
255 my $result = $resolve_module_versions->( $alt{$package} );
257 if ( exists( $prime{$package} ) ) { # primary package selected
259 if ( $result->{err} ) {
260 # Use the selected primary package, but there are conflicting
261 # errors among multiple alternative packages that need to be
264 "Found conflicting versions for package '$package'\n" .
265 " $prime{$package}{file} ($prime{$package}{version})\n" .
269 } elsif ( defined( $result->{version} ) ) {
270 # There is a primary package selected, and exactly one
271 # alternative package
273 if ( exists( $prime{$package}{version} ) &&
274 defined( $prime{$package}{version} ) ) {
275 # Unless the version of the primary package agrees with the
276 # version of the alternative package, report a conflict
277 if ( $compare_versions->(
278 $prime{$package}{version}, '!=', $result->{version}
283 "Found conflicting versions for package '$package'\n" .
284 " $prime{$package}{file} ($prime{$package}{version})\n" .
285 " $result->{file} ($result->{version})\n"
290 # The prime package selected has no version so, we choose to
291 # use any alternative package that does have a version
292 $prime{$package}{file} = $result->{file};
293 $prime{$package}{version} = $result->{version};
297 # no alt package found with a version, but we have a prime
298 # package so we use it whether it has a version or not
301 } else { # No primary package was selected, use the best alternative
303 if ( $result->{err} ) {
305 "Found conflicting versions for package '$package'\n" .
310 # Despite possible conflicting versions, we choose to record
311 # something rather than nothing
312 $prime{$package}{file} = $result->{file};
313 $prime{$package}{version} = $result->{version}
314 if defined( $result->{version} );
318 # Normalize versions. Can't use exists() here because of bug in YAML::Node.
319 # XXX "bug in YAML::Node" comment seems irrelvant -- dagolden, 2009-05-18
320 for (grep defined $_->{version}, values %prime) {
321 $_->{version} = $normalize_version->( $_->{version} );
332 my $filename = shift;
335 my $handle = delete $props{handle};
336 my( %valid_props, @valid_props );
337 @valid_props = qw( collect_pod inc );
338 @valid_props{@valid_props} = delete( @props{@valid_props} );
339 warn "Unknown properties: @{[keys %props]}\n" if scalar( %props );
343 filename => $filename,
354 my $self = bless(\%data, $class);
357 $self->_parse_fh($handle);
360 $self->_parse_file();
363 unless($self->{module} and length($self->{module})) {
364 my ($v, $d, $f) = File::Spec->splitpath($self->{filename});
367 my @candidates = grep /$f$/, @{$self->{packages}};
368 $self->{module} = shift(@candidates); # punt
371 if(grep /main/, @{$self->{packages}}) {
372 $self->{module} = 'main';
375 $self->{module} = $self->{packages}[0] || '';
380 $self->{version} = $self->{versions}{$self->{module}}
381 if defined( $self->{module} );
387 sub _do_find_module {
389 my $module = shift || croak 'find_module_by_name() requires a package name';
390 my $dirs = shift || \@INC;
392 my $file = File::Spec->catfile(split( /::/, $module));
393 foreach my $dir ( @$dirs ) {
394 my $testfile = File::Spec->catfile($dir, $file);
395 return [ File::Spec->rel2abs( $testfile ), $dir ]
396 if -e $testfile and !-d _; # For stuff like ExtUtils::xsubpp
397 return [ File::Spec->rel2abs( "$testfile.pm" ), $dir ]
398 if -e "$testfile.pm";
404 sub find_module_by_name {
405 my $found = shift()->_do_find_module(@_) or return;
410 sub find_module_dir_by_name {
411 my $found = shift()->_do_find_module(@_) or return;
416 # given a line of perl code, attempt to parse it if it looks like a
417 # $VERSION assignment, returning sigil, full name, & package name
418 sub _parse_version_expression {
422 my( $sig, $var, $pkg );
423 if ( $line =~ $VERS_REGEXP ) {
424 ( $sig, $var, $pkg ) = $2 ? ( $1, $2, $3 ) : ( $4, $5, $6 );
426 $pkg = ($pkg eq '::') ? 'main' : $pkg;
431 return ( $sig, $var, $pkg );
437 my $filename = $self->{filename};
438 my $fh = IO::File->new( $filename )
439 or croak( "Can't open '$filename': $!" );
441 $self->_parse_fh($fh);
445 my ($self, $fh) = @_;
447 my( $in_pod, $seen_end, $need_vers ) = ( 0, 0, 0 );
448 my( @pkgs, %vers, %pod, @pod );
453 while (defined( my $line = <$fh> )) {
457 next if $line =~ /^\s*#/;
460 if ( $line =~ /^=(.{0,3})/ ) {
461 $is_cut = $1 eq 'cut';
465 # Would be nice if we could also check $in_string or something too
466 last if !$in_pod && $line =~ /^__(?:DATA|END)__$/;
468 if ( $in_pod || $is_cut ) {
470 if ( $line =~ /^=head\d\s+(.+)\s*$/ ) {
472 if ( $self->{collect_pod} && length( $pod_data ) ) {
473 $pod{$pod_sect} = $pod_data;
479 } elsif ( $self->{collect_pod} ) {
480 $pod_data .= "$line\n";
489 # parse $line to see if it's a $VERSION declaration
490 my( $vers_sig, $vers_fullname, $vers_pkg ) =
491 $self->_parse_version_expression( $line );
493 if ( $line =~ $PKG_REGEXP ) {
495 push( @pkgs, $pkg ) unless grep( $pkg eq $_, @pkgs );
496 $vers{$pkg} = (defined $2 ? $2 : undef) unless exists( $vers{$pkg} );
497 $need_vers = defined $2 ? 0 : 1;
499 # VERSION defined with full package spec, i.e. $Module::VERSION
500 } elsif ( $vers_fullname && $vers_pkg ) {
501 push( @pkgs, $vers_pkg ) unless grep( $vers_pkg eq $_, @pkgs );
502 $need_vers = 0 if $vers_pkg eq $pkg;
504 unless ( defined $vers{$vers_pkg} && length $vers{$vers_pkg} ) {
506 $self->_evaluate_version_line( $vers_sig, $vers_fullname, $line );
508 # Warn unless the user is using the "$VERSION = eval
509 # $VERSION" idiom (though there are probably other idioms
510 # that we should watch out for...)
511 warn <<"EOM" unless $line =~ /=\s*eval/;
512 Package '$vers_pkg' already declared with version '$vers{$vers_pkg}',
513 ignoring subsequent declaration on line $line_num.
517 # first non-comment line in undeclared package main is VERSION
518 } elsif ( !exists($vers{main}) && $pkg eq 'main' && $vers_fullname ) {
521 $self->_evaluate_version_line( $vers_sig, $vers_fullname, $line );
523 push( @pkgs, 'main' );
525 # first non-comment line in undeclared package defines package main
526 } elsif ( !exists($vers{main}) && $pkg eq 'main' && $line =~ /\w+/ ) {
529 push( @pkgs, 'main' );
531 # only keep if this is the first $VERSION seen
532 } elsif ( $vers_fullname && $need_vers ) {
535 $self->_evaluate_version_line( $vers_sig, $vers_fullname, $line );
538 unless ( defined $vers{$pkg} && length $vers{$pkg} ) {
542 Package '$pkg' already declared with version '$vers{$pkg}'
543 ignoring new version '$v' on line $line_num.
553 if ( $self->{collect_pod} && length($pod_data) ) {
554 $pod{$pod_sect} = $pod_data;
557 $self->{versions} = \%vers;
558 $self->{packages} = \@pkgs;
559 $self->{pod} = \%pod;
560 $self->{pod_headings} = \@pod;
565 sub _evaluate_version_line {
567 my( $sigil, $var, $line ) = @_;
569 # Some of this code came from the ExtUtils:: hierarchy.
571 # We compile into $vsub because 'use version' would cause
572 # compiletime/runtime issues with local()
574 $pn++; # everybody gets their own package
575 my $eval = qq{BEGIN { q# Hide from _packages_inside()
576 #; package Module::Metadata::_version::p$pn;
589 # Try to get the $VERSION
591 # some modules say $VERSION = $Foo::Bar::VERSION, but Foo::Bar isn't
592 # installed, so we need to hunt in ./lib for it
593 if ( $@ =~ /Can't locate/ && -d 'lib' ) {
594 local @INC = ('lib',@INC);
597 warn "Error evaling version line '$eval' in $self->{filename}: $@\n"
599 (ref($vsub) eq 'CODE') or
600 croak "failed to build version sub for $self->{filename}";
601 my $result = eval { $vsub->() };
602 croak "Could not get version from $self->{filename} by executing:\n$eval\n\nThe fatal error was: $@\n"
605 # Upgrade it into a version object
606 my $version = eval { _dwim_version($result) };
608 croak "Version '$result' from $self->{filename} does not appear to be valid:\n$eval\n\nThe fatal error was: $@\n"
609 unless defined $version; # "0" is OK!
615 # Try to DWIM when things fail the lax version test in obvious ways
618 # Best case, it just works
619 sub { return shift },
621 # If we still don't have a version, try stripping any
622 # trailing junk that is prohibited by lax rules
625 $v =~ s{([0-9])[a-z-].*$}{$1}i; # 1.23-alpha or 1.23b
629 # Activestate apparently creates custom versions like '1.23_45_01', which
630 # cause version.pm to think it's an invalid alpha. So check for that
634 my $num_dots = () = $v =~ m{(\.)}g;
635 my $num_unders = () = $v =~ m{(_)}g;
636 my $leading_v = substr($v,0,1) eq 'v';
637 if ( ! $leading_v && $num_dots < 2 && $num_unders > 1 ) {
639 $num_unders = () = $v =~ m{(_)}g;
644 # Worst case, try numifying it like we would have before version objects
647 no warnings 'numeric';
654 my ($result) = shift;
656 return $result if ref($result) eq 'version';
658 my ($version, $error);
659 for my $f (@version_prep) {
660 $result = $f->($result);
661 $version = eval { version->new($result) };
662 $error ||= $@ if $@; # capture first failure
663 last if defined $version;
666 croak $error unless defined $version;
672 ############################################################
675 sub name { $_[0]->{module} }
677 sub filename { $_[0]->{filename} }
678 sub packages_inside { @{$_[0]->{packages}} }
679 sub pod_inside { @{$_[0]->{pod_headings}} }
680 sub contains_pod { $#{$_[0]->{pod_headings}} }
684 my $mod = shift || $self->{module};
686 if ( defined( $mod ) && length( $mod ) &&
687 exists( $self->{versions}{$mod} ) ) {
688 return $self->{versions}{$mod};
697 if ( defined( $sect ) && length( $sect ) &&
698 exists( $self->{pod}{$sect} ) ) {
699 return $self->{pod}{$sect};
709 Module::Metadata - Gather package and POD information from perl module files
713 use Module::Metadata;
715 # information about a .pm file
716 my $info = Module::Metadata->new_from_file( $file );
717 my $version = $info->version;
719 # CPAN META 'provides' field for .pm files in a directory
720 my $provides = Module::Metadata->provides(
721 dir => 'lib', version => 2
726 This module provides a standard way to gather metadata about a .pm file
727 without executing unsafe code.
735 =item C<< new_from_file($filename, collect_pod => 1) >>
737 Construct a C<Module::Metadata> object given the path to a file. Takes an
738 optional argument C<collect_pod> which is a boolean that determines whether POD
739 data is collected and stored for reference. POD data is not collected by
740 default. POD headings are always collected. Returns undef if the filename
743 =item C<< new_from_handle($handle, $filename, collect_pod => 1) >>
745 This works just like C<new_from_file>, except that a handle can be provided
746 as the first argument. Note that there is no validation to confirm that the
747 handle is a handle or something that can act like one. Passing something that
748 isn't a handle will cause a exception when trying to read from it. The
749 C<filename> argument is mandatory or undef will be returned.
751 =item C<< new_from_module($module, collect_pod => 1, inc => \@dirs) >>
753 Construct a C<Module::Metadata> object given a module or package name. In addition
754 to accepting the C<collect_pod> argument as described above, this
755 method accepts a C<inc> argument which is a reference to an array of
756 of directories to search for the module. If none are given, the
757 default is @INC. Returns undef if the module cannot be found.
759 =item C<< find_module_by_name($module, \@dirs) >>
761 Returns the path to a module given the module or package name. A list
762 of directories can be passed in as an optional parameter, otherwise
765 Can be called as either an object or a class method.
767 =item C<< find_module_dir_by_name($module, \@dirs) >>
769 Returns the entry in C<@dirs> (or C<@INC> by default) that contains
770 the module C<$module>. A list of directories can be passed in as an
771 optional parameter, otherwise @INC is searched.
773 Can be called as either an object or a class method.
775 =item C<< provides( %options ) >>
777 This is a convenience wrapper around C<package_versions_from_directory>
778 to generate a CPAN META C<provides> data structure. It takes key/value
779 pairs. Valid option keys include:
783 =item version B<(required)>
785 Specifies which version of the L<CPAN::Meta::Spec> should be used as
786 the format of the C<provides> output. Currently only '1.4' and '2'
787 are supported (and their format is identical). This may change in
788 the future as the definition of C<provides> changes.
790 The C<version> option is required. If it is omitted or if
791 an unsupported version is given, then C<provides> will throw an error.
795 Directory to search recursively for F<.pm> files. May not be specified with
800 Array reference of files to examine. May not be specified with C<dir>.
804 String to prepend to the C<file> field of the resulting output. This defaults
805 to F<lib>, which is the common case for most CPAN distributions with their
806 F<.pm> files in F<lib>. This option ensures the META information has the
807 correct relative path even when the C<dir> or C<files> arguments are
808 absolute or have relative paths from a location other than the distribution
813 For example, given C<dir> of 'lib' and C<prefix> of 'lib', the return value
814 is a hashref of the form:
819 file => 'lib/Package/Name.pm'
821 'OtherPackage::Name' => ...
824 =item C<< package_versions_from_directory($dir, \@files?) >>
826 Scans C<$dir> for .pm files (unless C<@files> is given, in which case looks
827 for those files in C<$dir> - and reads each file for packages and versions,
828 returning a hashref of the form:
833 file => 'Package/Name.pm'
835 'OtherPackage::Name' => ...
838 The C<DB> and C<main> packages are always omitted, as are any "private"
839 packages that have leading underscores in the namespace (e.g.
842 Note that the file path is relative to C<$dir> if that is specified.
843 This B<must not> be used directly for CPAN META C<provides>. See
844 the C<provides> method instead.
846 =item C<< log_info (internal) >>
848 Used internally to perform logging; imported from Log::Contextual if
849 Log::Contextual has already been loaded, otherwise simply calls warn.
853 =head2 Object methods
859 Returns the name of the package represented by this module. If there
860 are more than one packages, it makes a best guess based on the
861 filename. If it's a script (i.e. not a *.pm) the package name is
864 =item C<< version($package) >>
866 Returns the version as defined by the $VERSION variable for the
867 package as returned by the C<name> method if no arguments are
868 given. If given the name of a package it will attempt to return the
869 version of that package if it is specified in the file.
871 =item C<< filename() >>
873 Returns the absolute path to the file.
875 =item C<< packages_inside() >>
877 Returns a list of packages. Note: this is a raw list of packages
878 discovered (or assumed, in the case of C<main>). It is not
879 filtered for C<DB>, C<main> or private packages the way the
880 C<provides> method does.
882 =item C<< pod_inside() >>
884 Returns a list of POD sections.
886 =item C<< contains_pod() >>
888 Returns true if there is any POD in the file.
890 =item C<< pod($section) >>
892 Returns the POD data in the given section.
898 Original code from Module::Build::ModuleInfo by Ken Williams
899 <kwilliams@cpan.org>, Randy W. Sims <RandyS@ThePierianSpring.org>
901 Released as Module::Metadata by Matt S Trout (mst) <mst@shadowcat.co.uk> with
902 assistance from David Golden (xdg) <dagolden@cpan.org>.
906 Original code Copyright (c) 2001-2011 Ken Williams.
907 Additional code Copyright (c) 2010-2011 Matt Trout and David Golden.
910 This library is free software; you can redistribute it and/or
911 modify it under the same terms as Perl itself.