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 = '1.000010';
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::Unix->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 $prime_package = $package if lc($prime_package) eq lc($package);
236 if ( $package eq $prime_package ) {
237 if ( exists( $prime{$package} ) ) {
238 croak "Unexpected conflict in '$package'; multiple versions found.\n";
240 $mapped_filename = "$package.pm" if lc("$package.pm") eq lc($mapped_filename);
241 $prime{$package}{file} = $mapped_filename;
242 $prime{$package}{version} = $version if defined( $version );
245 push( @{$alt{$package}}, {
246 file => $mapped_filename,
253 # Then we iterate over all the packages found above, identifying conflicts
254 # and selecting the "best" candidate for recording the file & version
256 foreach my $package ( keys( %alt ) ) {
257 my $result = $resolve_module_versions->( $alt{$package} );
259 if ( exists( $prime{$package} ) ) { # primary package selected
261 if ( $result->{err} ) {
262 # Use the selected primary package, but there are conflicting
263 # errors among multiple alternative packages that need to be
266 "Found conflicting versions for package '$package'\n" .
267 " $prime{$package}{file} ($prime{$package}{version})\n" .
271 } elsif ( defined( $result->{version} ) ) {
272 # There is a primary package selected, and exactly one
273 # alternative package
275 if ( exists( $prime{$package}{version} ) &&
276 defined( $prime{$package}{version} ) ) {
277 # Unless the version of the primary package agrees with the
278 # version of the alternative package, report a conflict
279 if ( $compare_versions->(
280 $prime{$package}{version}, '!=', $result->{version}
285 "Found conflicting versions for package '$package'\n" .
286 " $prime{$package}{file} ($prime{$package}{version})\n" .
287 " $result->{file} ($result->{version})\n"
292 # The prime package selected has no version so, we choose to
293 # use any alternative package that does have a version
294 $prime{$package}{file} = $result->{file};
295 $prime{$package}{version} = $result->{version};
299 # no alt package found with a version, but we have a prime
300 # package so we use it whether it has a version or not
303 } else { # No primary package was selected, use the best alternative
305 if ( $result->{err} ) {
307 "Found conflicting versions for package '$package'\n" .
312 # Despite possible conflicting versions, we choose to record
313 # something rather than nothing
314 $prime{$package}{file} = $result->{file};
315 $prime{$package}{version} = $result->{version}
316 if defined( $result->{version} );
320 # Normalize versions. Can't use exists() here because of bug in YAML::Node.
321 # XXX "bug in YAML::Node" comment seems irrelvant -- dagolden, 2009-05-18
322 for (grep defined $_->{version}, values %prime) {
323 $_->{version} = $normalize_version->( $_->{version} );
334 my $filename = shift;
337 my $handle = delete $props{handle};
338 my( %valid_props, @valid_props );
339 @valid_props = qw( collect_pod inc );
340 @valid_props{@valid_props} = delete( @props{@valid_props} );
341 warn "Unknown properties: @{[keys %props]}\n" if scalar( %props );
345 filename => $filename,
356 my $self = bless(\%data, $class);
359 $self->_parse_fh($handle);
362 $self->_parse_file();
365 unless($self->{module} and length($self->{module})) {
366 my ($v, $d, $f) = File::Spec->splitpath($self->{filename});
369 my @candidates = grep /$f$/, @{$self->{packages}};
370 $self->{module} = shift(@candidates); # punt
373 if(grep /main/, @{$self->{packages}}) {
374 $self->{module} = 'main';
377 $self->{module} = $self->{packages}[0] || '';
382 $self->{version} = $self->{versions}{$self->{module}}
383 if defined( $self->{module} );
389 sub _do_find_module {
391 my $module = shift || croak 'find_module_by_name() requires a package name';
392 my $dirs = shift || \@INC;
394 my $file = File::Spec->catfile(split( /::/, $module));
395 foreach my $dir ( @$dirs ) {
396 my $testfile = File::Spec->catfile($dir, $file);
397 return [ File::Spec->rel2abs( $testfile ), $dir ]
398 if -e $testfile and !-d _; # For stuff like ExtUtils::xsubpp
399 return [ File::Spec->rel2abs( "$testfile.pm" ), $dir ]
400 if -e "$testfile.pm";
406 sub find_module_by_name {
407 my $found = shift()->_do_find_module(@_) or return;
412 sub find_module_dir_by_name {
413 my $found = shift()->_do_find_module(@_) or return;
418 # given a line of perl code, attempt to parse it if it looks like a
419 # $VERSION assignment, returning sigil, full name, & package name
420 sub _parse_version_expression {
424 my( $sig, $var, $pkg );
425 if ( $line =~ /$VERS_REGEXP/o ) {
426 ( $sig, $var, $pkg ) = $2 ? ( $1, $2, $3 ) : ( $4, $5, $6 );
428 $pkg = ($pkg eq '::') ? 'main' : $pkg;
433 return ( $sig, $var, $pkg );
439 my $filename = $self->{filename};
440 my $fh = IO::File->new( $filename )
441 or croak( "Can't open '$filename': $!" );
443 $self->_handle_bom($fh, $filename);
445 $self->_parse_fh($fh);
448 # Look for a UTF-8/UTF-16BE/UTF-16LE BOM at the beginning of the stream.
449 # If there's one, then skip it and set the :encoding layer appropriately.
451 my ($self, $fh, $filename) = @_;
453 my $pos = $fh->getpos;
454 return unless defined $pos;
457 my $count = $fh->read( $buf, length $buf );
458 return unless defined $count and $count >= 2;
461 if ( $buf eq "\x{FE}\x{FF}" ) {
462 $encoding = 'UTF-16BE';
463 } elsif ( $buf eq "\x{FF}\x{FE}" ) {
464 $encoding = 'UTF-16LE';
465 } elsif ( $buf eq "\x{EF}\x{BB}" ) {
467 $count = $fh->read( $buf, length $buf );
468 if ( defined $count and $count >= 1 and $buf eq "\x{BF}" ) {
473 if ( defined $encoding ) {
474 if ( "$]" >= 5.008 ) {
475 # $fh->binmode requires perl 5.10
476 binmode( $fh, ":encoding($encoding)" );
480 or croak( sprintf "Can't reset position to the top of '$filename'" );
487 my ($self, $fh) = @_;
489 my( $in_pod, $seen_end, $need_vers ) = ( 0, 0, 0 );
490 my( @pkgs, %vers, %pod, @pod );
495 while (defined( my $line = <$fh> )) {
499 next if $line =~ /^\s*#/;
502 if ( $line =~ /^=(.{0,3})/ ) {
503 $is_cut = $1 eq 'cut';
507 # Would be nice if we could also check $in_string or something too
508 last if !$in_pod && $line =~ /^__(?:DATA|END)__$/;
510 if ( $in_pod || $is_cut ) {
512 if ( $line =~ /^=head\d\s+(.+)\s*$/ ) {
514 if ( $self->{collect_pod} && length( $pod_data ) ) {
515 $pod{$pod_sect} = $pod_data;
521 } elsif ( $self->{collect_pod} ) {
522 $pod_data .= "$line\n";
531 # parse $line to see if it's a $VERSION declaration
532 my( $vers_sig, $vers_fullname, $vers_pkg ) =
534 ? $self->_parse_version_expression( $line )
537 if ( $line =~ /$PKG_REGEXP/o ) {
539 push( @pkgs, $pkg ) unless grep( $pkg eq $_, @pkgs );
540 $vers{$pkg} = (defined $2 ? $2 : undef) unless exists( $vers{$pkg} );
541 $need_vers = defined $2 ? 0 : 1;
543 # VERSION defined with full package spec, i.e. $Module::VERSION
544 } elsif ( $vers_fullname && $vers_pkg ) {
545 push( @pkgs, $vers_pkg ) unless grep( $vers_pkg eq $_, @pkgs );
546 $need_vers = 0 if $vers_pkg eq $pkg;
548 unless ( defined $vers{$vers_pkg} && length $vers{$vers_pkg} ) {
550 $self->_evaluate_version_line( $vers_sig, $vers_fullname, $line );
552 # Warn unless the user is using the "$VERSION = eval
553 # $VERSION" idiom (though there are probably other idioms
554 # that we should watch out for...)
555 warn <<"EOM" unless $line =~ /=\s*eval/;
556 Package '$vers_pkg' already declared with version '$vers{$vers_pkg}',
557 ignoring subsequent declaration on line $line_num.
561 # first non-comment line in undeclared package main is VERSION
562 } elsif ( !exists($vers{main}) && $pkg eq 'main' && $vers_fullname ) {
565 $self->_evaluate_version_line( $vers_sig, $vers_fullname, $line );
567 push( @pkgs, 'main' );
569 # first non-comment line in undeclared package defines package main
570 } elsif ( !exists($vers{main}) && $pkg eq 'main' && $line =~ /\w+/ ) {
573 push( @pkgs, 'main' );
575 # only keep if this is the first $VERSION seen
576 } elsif ( $vers_fullname && $need_vers ) {
579 $self->_evaluate_version_line( $vers_sig, $vers_fullname, $line );
582 unless ( defined $vers{$pkg} && length $vers{$pkg} ) {
586 Package '$pkg' already declared with version '$vers{$pkg}'
587 ignoring new version '$v' on line $line_num.
597 if ( $self->{collect_pod} && length($pod_data) ) {
598 $pod{$pod_sect} = $pod_data;
601 $self->{versions} = \%vers;
602 $self->{packages} = \@pkgs;
603 $self->{pod} = \%pod;
604 $self->{pod_headings} = \@pod;
609 sub _evaluate_version_line {
611 my( $sigil, $var, $line ) = @_;
613 # Some of this code came from the ExtUtils:: hierarchy.
615 # We compile into $vsub because 'use version' would cause
616 # compiletime/runtime issues with local()
618 $pn++; # everybody gets their own package
619 my $eval = qq{BEGIN { q# Hide from _packages_inside()
620 #; package Module::Metadata::_version::p$pn;
633 # Try to get the $VERSION
635 # some modules say $VERSION = $Foo::Bar::VERSION, but Foo::Bar isn't
636 # installed, so we need to hunt in ./lib for it
637 if ( $@ =~ /Can't locate/ && -d 'lib' ) {
638 local @INC = ('lib',@INC);
641 warn "Error evaling version line '$eval' in $self->{filename}: $@\n"
643 (ref($vsub) eq 'CODE') or
644 croak "failed to build version sub for $self->{filename}";
645 my $result = eval { $vsub->() };
646 croak "Could not get version from $self->{filename} by executing:\n$eval\n\nThe fatal error was: $@\n"
649 # Upgrade it into a version object
650 my $version = eval { _dwim_version($result) };
652 croak "Version '$result' from $self->{filename} does not appear to be valid:\n$eval\n\nThe fatal error was: $@\n"
653 unless defined $version; # "0" is OK!
659 # Try to DWIM when things fail the lax version test in obvious ways
662 # Best case, it just works
663 sub { return shift },
665 # If we still don't have a version, try stripping any
666 # trailing junk that is prohibited by lax rules
669 $v =~ s{([0-9])[a-z-].*$}{$1}i; # 1.23-alpha or 1.23b
673 # Activestate apparently creates custom versions like '1.23_45_01', which
674 # cause version.pm to think it's an invalid alpha. So check for that
678 my $num_dots = () = $v =~ m{(\.)}g;
679 my $num_unders = () = $v =~ m{(_)}g;
680 my $leading_v = substr($v,0,1) eq 'v';
681 if ( ! $leading_v && $num_dots < 2 && $num_unders > 1 ) {
683 $num_unders = () = $v =~ m{(_)}g;
688 # Worst case, try numifying it like we would have before version objects
691 no warnings 'numeric';
698 my ($result) = shift;
700 return $result if ref($result) eq 'version';
702 my ($version, $error);
703 for my $f (@version_prep) {
704 $result = $f->($result);
705 $version = eval { version->new($result) };
706 $error ||= $@ if $@; # capture first failure
707 last if defined $version;
710 croak $error unless defined $version;
716 ############################################################
719 sub name { $_[0]->{module} }
721 sub filename { $_[0]->{filename} }
722 sub packages_inside { @{$_[0]->{packages}} }
723 sub pod_inside { @{$_[0]->{pod_headings}} }
724 sub contains_pod { $#{$_[0]->{pod_headings}} }
728 my $mod = shift || $self->{module};
730 if ( defined( $mod ) && length( $mod ) &&
731 exists( $self->{versions}{$mod} ) ) {
732 return $self->{versions}{$mod};
741 if ( defined( $sect ) && length( $sect ) &&
742 exists( $self->{pod}{$sect} ) ) {
743 return $self->{pod}{$sect};
753 Module::Metadata - Gather package and POD information from perl module files
757 use Module::Metadata;
759 # information about a .pm file
760 my $info = Module::Metadata->new_from_file( $file );
761 my $version = $info->version;
763 # CPAN META 'provides' field for .pm files in a directory
764 my $provides = Module::Metadata->provides(
765 dir => 'lib', version => 2
770 This module provides a standard way to gather metadata about a .pm file
771 without executing unsafe code.
779 =item C<< new_from_file($filename, collect_pod => 1) >>
781 Constructs a C<Module::Metadata> object given the path to a file. Returns
782 undef if the filename does not exist.
784 C<collect_pod> is a optional boolean argument that determines whether POD
785 data is collected and stored for reference. POD data is not collected by
786 default. POD headings are always collected.
788 If the file begins by an UTF-8, UTF-16BE or UTF-16LE byte-order mark, then
789 it is skipped before processing, and the content of the file is also decoded
790 appropriately starting from perl 5.8.
792 =item C<< new_from_handle($handle, $filename, collect_pod => 1) >>
794 This works just like C<new_from_file>, except that a handle can be provided
795 as the first argument.
797 Note that there is no validation to confirm that the handle is a handle or
798 something that can act like one. Passing something that isn't a handle will
799 cause a exception when trying to read from it. The C<filename> argument is
800 mandatory or undef will be returned.
802 You are responsible for setting the decoding layers on C<$handle> if
805 =item C<< new_from_module($module, collect_pod => 1, inc => \@dirs) >>
807 Constructs a C<Module::Metadata> object given a module or package name.
808 Returns undef if the module cannot be found.
810 In addition to accepting the C<collect_pod> argument as described above,
811 this method accepts a C<inc> argument which is a reference to an array of
812 directories to search for the module. If none are given, the default is
815 If the file that contains the module begins by an UTF-8, UTF-16BE or
816 UTF-16LE byte-order mark, then it is skipped before processing, and the
817 content of the file is also decoded appropriately starting from perl 5.8.
819 =item C<< find_module_by_name($module, \@dirs) >>
821 Returns the path to a module given the module or package name. A list
822 of directories can be passed in as an optional parameter, otherwise
825 Can be called as either an object or a class method.
827 =item C<< find_module_dir_by_name($module, \@dirs) >>
829 Returns the entry in C<@dirs> (or C<@INC> by default) that contains
830 the module C<$module>. A list of directories can be passed in as an
831 optional parameter, otherwise @INC is searched.
833 Can be called as either an object or a class method.
835 =item C<< provides( %options ) >>
837 This is a convenience wrapper around C<package_versions_from_directory>
838 to generate a CPAN META C<provides> data structure. It takes key/value
839 pairs. Valid option keys include:
843 =item version B<(required)>
845 Specifies which version of the L<CPAN::Meta::Spec> should be used as
846 the format of the C<provides> output. Currently only '1.4' and '2'
847 are supported (and their format is identical). This may change in
848 the future as the definition of C<provides> changes.
850 The C<version> option is required. If it is omitted or if
851 an unsupported version is given, then C<provides> will throw an error.
855 Directory to search recursively for F<.pm> files. May not be specified with
860 Array reference of files to examine. May not be specified with C<dir>.
864 String to prepend to the C<file> field of the resulting output. This defaults
865 to F<lib>, which is the common case for most CPAN distributions with their
866 F<.pm> files in F<lib>. This option ensures the META information has the
867 correct relative path even when the C<dir> or C<files> arguments are
868 absolute or have relative paths from a location other than the distribution
873 For example, given C<dir> of 'lib' and C<prefix> of 'lib', the return value
874 is a hashref of the form:
879 file => 'lib/Package/Name.pm'
881 'OtherPackage::Name' => ...
884 =item C<< package_versions_from_directory($dir, \@files?) >>
886 Scans C<$dir> for .pm files (unless C<@files> is given, in which case looks
887 for those files in C<$dir> - and reads each file for packages and versions,
888 returning a hashref of the form:
893 file => 'Package/Name.pm'
895 'OtherPackage::Name' => ...
898 The C<DB> and C<main> packages are always omitted, as are any "private"
899 packages that have leading underscores in the namespace (e.g.
902 Note that the file path is relative to C<$dir> if that is specified.
903 This B<must not> be used directly for CPAN META C<provides>. See
904 the C<provides> method instead.
906 =item C<< log_info (internal) >>
908 Used internally to perform logging; imported from Log::Contextual if
909 Log::Contextual has already been loaded, otherwise simply calls warn.
913 =head2 Object methods
919 Returns the name of the package represented by this module. If there
920 are more than one packages, it makes a best guess based on the
921 filename. If it's a script (i.e. not a *.pm) the package name is
924 =item C<< version($package) >>
926 Returns the version as defined by the $VERSION variable for the
927 package as returned by the C<name> method if no arguments are
928 given. If given the name of a package it will attempt to return the
929 version of that package if it is specified in the file.
931 =item C<< filename() >>
933 Returns the absolute path to the file.
935 =item C<< packages_inside() >>
937 Returns a list of packages. Note: this is a raw list of packages
938 discovered (or assumed, in the case of C<main>). It is not
939 filtered for C<DB>, C<main> or private packages the way the
940 C<provides> method does.
942 =item C<< pod_inside() >>
944 Returns a list of POD sections.
946 =item C<< contains_pod() >>
948 Returns true if there is any POD in the file.
950 =item C<< pod($section) >>
952 Returns the POD data in the given section.
958 Original code from Module::Build::ModuleInfo by Ken Williams
959 <kwilliams@cpan.org>, Randy W. Sims <RandyS@ThePierianSpring.org>
961 Released as Module::Metadata by Matt S Trout (mst) <mst@shadowcat.co.uk> with
962 assistance from David Golden (xdg) <dagolden@cpan.org>.
964 =head1 COPYRIGHT & LICENSE
966 Original code Copyright (c) 2001-2011 Ken Williams.
967 Additional code Copyright (c) 2010-2011 Matt Trout and David Golden.
970 This library is free software; you can redistribute it and/or
971 modify it under the same terms as Perl itself.