File Coverage

blib/lib/PPI/Xref.pm

Criterion	Covered	Total	%
statement	755	896	84.2
branch	267	386	69.1
condition	92	150	61.3
subroutine	111	131	84.7
pod	24	34	70.5
total	1249	1597	78.2

package PPI::Xref;

require v5.14;  # 5.10: defined-or; 5.14: package Sub { ... }

our $VERSION = '0.008';

use strict;
use warnings;

use PPI;

# We will use file ids (integers) instead of filenames, mainly for
# space savings, but also speed, and only convert back to filenames
# on leaving the API boundary.
my $FILE_ID = 0;
my %FILE_BY_ID;

my %CTOR_OPTS =
    map { $_ => 1} qw/process_verbose cache_verbose recurse_verbose
                      recurse INC
                      cache_directory
                      __allow_relative/;

my $HASHALGO = 'sha1';

package Sub {  # For getting the current sub name.
    sub TIESCALAR { bless \$_[1], $_[0] }
    sub FETCH { (caller(1))[3] }
}
tie my $Sub, 'Sub';

sub __is_readwrite_directory {
    my ($self, $dir) = @_;
    return -d $dir && -r $dir && -w $dir;
}

# Our constructor.
sub new {
    my ($class, $opt) = @_;
    $opt //= {};
    # In the opt you can specify:
    # - process_verbose: for process progress
    # - cache_verbose: for cache activity
    # - recurse_verbose: for recurse, show revisits
    # - INC: an aref for a custom @INC
    # - recurse: or not (default: yes)
    # - cache_directory: directory where to cache the results

    $opt->{recurse} //= 1;

    my %unexpopt = %$opt;
    delete @unexpopt{keys %CTOR_OPTS};

    for my $k (sort keys %unexpopt) {
        warn "$Sub: unexpected option: $k\n";
    }

    my $self = { opt => $opt };

    my $cache_directory = $opt->{cache_directory};
    if (defined $cache_directory) {
        unless (PPI::Xref->__is_readwrite_directory($cache_directory)) {
            warn "$Sub: cache_directory '$cache_directory': not a read-write directory\n";
        }

        $self->{__cache_prefix_length} = length($cache_directory) + 1;

        use Sereal::Encoder;
        use Sereal::Decoder;
        $self->{encoder} = Sereal::Encoder->new;
        $self->{decoder} = Sereal::Decoder->new;
    }

    bless $self, $class;
}

# Unless $self->{inc_dirs} is set, set it from either opt INC, if set,
# or from the system @INC.
sub __inc_dirs {
    my $self = shift;
    unless (defined $self->{inc_dirs}) {
        for my $d ($self->{opt}{INC} ? @{ $self->{opt}{INC}} : @INC) {
            next if ref $d;
            next unless -d $d && -x $d;
            push @{ $self->{inc_dirs} }, $d;
        }
    }
}

sub PPI::Xref::INC {
    my $self = shift;
    $self->__inc_dirs;
    return $self->{inc_dirs};
}

# Given a file, look for it in @INC.
sub __find_file {
    my ($self, $file) = @_;
    $self->__inc_dirs;
    unless (exists $self->{inc_file}{$file}) {
        for my $d (@{ $self->{inc_dirs}}) {
            unless ($self->{opt}{__allow_relative}) {  # For testing.
                use File::Spec;
                $d = File::Spec->rel2abs($d) unless
                    File::Spec->file_name_is_absolute($d);
            }
            my $f = "$d/$file";
            if (-f $f) {
                $self->{inc_file}{$file} = $f;
                last;
            }
        }
        unless (exists $self->{inc_file}{$file}) {
            $self->{inc_file}{$file} = undef;
        }
    }
    return $self->{inc_file}{$file};
}

# Given a module name, look for its file in @INC.
sub __find_module {
    my ($self, $module_name) = @_;
    unless (exists $self->{module_file}{$module_name}) {
        my $m = $module_name;
        $m =~ s{::}{/}g;
        $m .= '.pm';
        $self->{module_file}{$module_name} = $self->__find_file($m);
    }
    return $self->{module_file}{$module_name};
}

# Remove comments and tokens, and squeeze
# multiple whitespaces into one.
sub __normalize_whitespace {
    my @n;
    my $prev_ws;
    my $curr_ws;
    for my $n (@_) {
        next if $n->isa('PPI::Token::Comment');
        next if $n->isa('PPI::Token::Pod');
        $curr_ws = $n->isa('PPI::Token::Whitespace');
        next if $prev_ws && $curr_ws;
        push @n, $n;
        $prev_ws = $curr_ws;
    }
    return @n;
}

# For a filename, assign it a file id (an integer) if it does not
# have one, and in any case return its file id.
sub __assign_file_id {
    my ($self, $filename) = @_;
    my $file_id = $self->{file_id}{$filename};
    unless (defined $file_id) {
        $file_id = $self->{file_id}{$filename} = $FILE_ID++;
        $FILE_BY_ID{$file_id} = $filename;
    }
    return $file_id;
}

# Close the current package, if any.
sub __close_package {
    my ($self, $file_id, $package, $elem) = @_;
    if (exists $self->{file_packages} &&
        ref $self->{file_packages}{$file_id} &&
        @{ $self->{file_packages}{$file_id} }) {
        push @{ $self->{file_packages}{$file_id}[-1] },
             $elem->line_number,
             $elem->column_number;
    }
}

# Open a new package.
sub __open_package {
    my ($self, $file_id, $package, $elem) = @_;
    push @{ $self->{file_packages}{$file_id} },
         [
          $package,             # 0
          $elem->line_number,   # 1
          $elem->column_number, # 2
         ];
}

# Close the current package, if any, and open a new one.
sub __close_open_package {
    my ($self, $file_id, $old_package, $old_elem,
        $new_package, $new_elem, $fileloc) = @_;
    if (defined $old_package && $old_package ne 'main') {
        $self->__close_package($file_id, $old_package, $old_elem, $fileloc);
    }
    $self->__open_package($file_id, $new_package, $new_elem);
}

# Function for portably turning the directory portion of a pathname
# into a directory name.  If the $flatten_volume is true, loses
# information in platforms that have a volume name in pathnames, but
# the main idea is to safely split the argument into a new directory
# name (possibly modified by prepending the volume name as a
# directory), and the filename.  E.g.  '/a/b/c' -> ('/a/b', 'c')
# 'c:/d/e' -> ('/c/d', 'e')
sub __safe_vol_dir_file {
    my ($self, $path, $flatten_volume) = @_;
    my ($vol, $dirs, $file) =  File::Spec->splitpath($path);
    if ($flatten_volume && $^O eq 'MSWin32') {
      $vol =~ s/:$//;  # splitpath() leaves the $vol as e.g. "c:"
    }
    return (File::Spec->catpath($vol, $dirs), $file);
}

# Returns the directory part and the file part.  Note that this will
# convert the volume name (if any) in the $path into a directory name,
# e.g. 'c:/d/e' -> ('/c/d', 'e').  This is useful for re-rooting
# a pathname under a new directory.
sub __safe_dir_and_file_flatten_volume {
  my ($self, $path) = @_;
  return $self->__safe_vol_dir_file($path, 1);
}

# Returns the directory part (with the possible volume part prepended)
# and the file part.  Kind of like a safe dirname().
sub __safe_dir_and_file_same_volume {
  my ($self, $path) = @_;
  return $self->__safe_vol_dir_file($path, 0);
}

my $CACHE_EXT = '.cache';

# "shadow file" is a filename rooted into a new, "shadow", directory.
sub __shadow_cache_filename {
    my ($self, $shadowdir, $filename) = @_;

    # Paranoia check.  (Either absolute or relative is fine, though.)
    if ($filename =~ m{\.\.}) {
        warn "$Sub: Skipping unexpected file: '$filename'\n";
        return;
    }

    use File::Spec;
    my $absfile =
        File::Spec->file_name_is_absolute($filename) ?
        $filename :
        File::Spec->rel2abs($filename);
    my ($redir, $file) = $self->__safe_dir_and_file_flatten_volume($absfile);

    # For portable filenames, we cannot just keep on
    # appending filename extensions with dots, and we
    # are going to append the cache filename extension.
    # So we mangle the .pm or .pl as _pm and _pl.
    $file =~ s{\.(p[ml])$}{_$1};

    return File::Spec->catfile($shadowdir, $redir, $file . $CACHE_EXT);
}

# The hash checksum for the file, and the mtime timestamp.
sub __current_filehash_and_mtime {
    my ($self, $origfilename) = @_;
    return unless -f $origfilename;
    my $origfilefh;
    unless (open($origfilefh, $origfilename)) {
        warn qq[$Sub: Failed to open "$origfilename": $!\n];
        return;
    }
    use Digest::SHA;
    my $sha = Digest::SHA->new($HASHALGO);
    $sha->addfile($origfilefh);
    return (
        "$HASHALGO:". $sha->hexdigest,
        (stat($origfilefh))[9], # mtime
        );
}

# Create the directory of the filename.
sub __make_path_file {
    my ($self, $base) = @_;
    use File::Path qw[make_path];
    my ($dir, $file) = $self->__safe_dir_and_file_same_volume($base);
    return eval { make_path($dir) unless -d $dir; 1; };
}

# The attributes that are written to and read from cache.
my @CACHE_FIELDS =
    qw[
       file_incs
       file_lines
       file_modules
       file_packages
       file_subs
       file_missing_modules
       file_parse_errors
      ];

# Error fields are cache fields but they should not be cleared
# since they accumulate and are hrefs as opposed to arefs.
my %CACHE_FIELDS_KEEP =
    map { $_ => 1 }
    qw[
       file_missing_modules
       file_parse_errors
      ];

my @CACHE_FIELDS_CLEAR =
    grep { ! exists $CACHE_FIELDS_KEEP{$_} } @CACHE_FIELDS;

# Given the href, serialize it to the file.
sub __encode_to_file {
    my ($self, $file, $cached) = @_;

    my $success = 0;

    my $temp = "$file.$$";  # For atomic renaming.

    # If anything goes wrong, abort the commit.
  COMMIT: {
      my $blob = $self->{encoder}->encode($cached);
      unless (defined $blob) {
          warn "$Sub: Failed to encode into '$temp'\n";
          last COMMIT;
      }

      unless ($self->__make_path_file($temp)) {
          warn "$Sub: Failed to create path for '$temp'\n";
          last COMMIT;
      }

      my $fh;
      use Fcntl qw[O_CREAT O_WRONLY];
      unless (sysopen($fh, $temp, O_CREAT|O_WRONLY, 0644)) {
          warn "$Sub: Failed to open '$temp' for writing: $!\n";
          last COMMIT;
      }

      my $size = length($blob);
      my $wrote = syswrite($fh, $blob);

      unless (defined $wrote && $wrote == $size) {
          warn "$Sub: Failed to write $size bytes to '$temp': $!\n";
          last COMMIT;
      }

      unless (close($fh)) {
          warn "$Sub: Failed to close '$temp': $!\n";
          last COMMIT;
      }

      unless (rename($temp, $file)) {
          warn "$Sub: Failed to rename '$temp' as '$file': !$\n";
          last COMMIT;
      }

      # Finally we are happy.
      $success = 1;

    }  # COMMIT

    if (-f $temp) {
        warn "$Sub: Cleaning temporary file '$temp'\n";
    }
    unlink $temp;   # In any case.

    return $success;
}

# Write the results to the file.
sub __write_cachefile {
    my ($self, $cache_filename, $hash_current, $file_id, $file_mtime) = @_;

    if ($self->{opt}{cache_verbose}) {
        print "$Sub: writing $cache_filename\n";
    }

    my $cached;  # Re-root the data we care about.
    for my $k (@CACHE_FIELDS) {
        if (defined $self->{$k}{$file_id}) {
            $cached->{$k} = $self->{$k}{$file_id};
        }
    }
    $cached->{file_hash} = $hash_current;

    # The mtime is in UTC, and should only be used for
    # maintenance / statistics.  In other words, it should
    # NOT be used for uptodateness.
    $cached->{file_mtime} = $file_mtime;

    # Mark also in the object that we have processed this one.
    $self->{file_hash}{$file_id} = $hash_current;

    return $self->__encode_to_file($cache_filename, $cached);
}

# Compose a cache filename, given an original filename.
# The filenames are re-rooted in the cache_directory.
sub __cache_filename {
    my ($self, $path) = @_;
    return if $path eq '-';

    my $cache_directory = $self->{opt}{cache_directory};
    return unless defined $cache_directory;

    unless ($self->__is_readwrite_directory($cache_directory)) {
        warn "$Sub: Not a read-write directory '$cache_directory'\n";
        return;
    }

    if ($path !~ /\.p[ml]$/) {
        warn "$Sub: Unexpected filename: '$path'\n";
        return;
    }

    return $self->__shadow_cache_filename($cache_directory, $path);
}

# Deserialize from the file.
sub __decode_from_file {
    my ($self, $file) = @_;

    my $fh;
    use Fcntl qw[O_RDONLY];
    unless (sysopen($fh, $file, O_RDONLY)) {
        # warn "$Sub: Failed to open '$file' for reading: $!\n";
        return;
    }

    my $size = -s $fh;
    my $read = sysread($fh, my $blob, $size);
    unless ($read == $size) {
        warn "$Sub: Failed to read $size bytes from '$file': $!\n";
        return;
    }

    return $self->{decoder}->decode($blob);
}

# Check if we have the results for this file cached.
sub __check_cached {
    my ($self, $origfile) = @_;
    return if $origfile eq '-';

    my ($hash_current, $file_mtime) =
        $self->__current_filehash_and_mtime($origfile);
    my $cache_directory = $self->{opt}{cache_directory};
    my $cache_filename;
    my $cached;
    my $hash_previous;
    my $hash_match;

    if (defined $cache_directory) {
        $cache_filename = $self->__cache_filename($origfile);
        if (defined $cache_filename) {
            if ($self->{opt}{cache_verbose}) {
                print "$Sub: reading $cache_filename\n";
            }
            $cached = $self->__decode_from_file($cache_filename);
            if (defined $cached) {
                if ($self->{opt}{cache_verbose}) {
                    print "$Sub: reading $cache_filename SUCCESS\n";
                }
                $hash_previous = $cached->{file_hash};
                $hash_match =
                    defined $hash_previous &&
                    defined $hash_current &&
                    $hash_previous eq $hash_current;
            } else {
                if ($self->{opt}{cache_verbose}) {
                    print "$Sub: reading $cache_filename FAILURE\n";
                }
            }
        }
    }

    return ($cache_filename,
            $cached,
            $hash_current,
            $hash_match,
            $file_mtime);
}

# Write to the cache and tick various counters.
sub __to_cache {
    my ($self, $cache_filename, $hash_current, $file_id, $file_mtime) = @_;

    my $had_cache = -f $cache_filename;
    if ($self->__write_cachefile($cache_filename, $hash_current,
                                 $file_id, $file_mtime)) {
        if ($self->{opt}{cache_verbose}) {
            print "$Sub: writing $cache_filename SUCCESS\n";
        }
        $self->{__cachewrites}++;
        unless ($had_cache) {
            $self->{__cachecreates}++;
        }
    } else {
        if ($self->{opt}{cache_verbose}) {
            print "$Sub: writing $cache_filename FAILURE\n";
        }
    }
}

# Import the fields we care about from the cached data.
sub __import_cached {
    my ($self, $file_id, $cached) = @_;

    for my $k (@CACHE_FIELDS) {
        $self->{$k}{$file_id} = $cached->{$k};
    }

    return 1;
}

# Clear the cached fields.  Used especially in preparation of import.
sub __clear_cached {
    my ($self, $file_id) = @_;

    for my $k (@CACHE_FIELDS_CLEAR) {
        delete $self->{$k}{$file_id};
    }

    return 1;
}

sub __parse_error {
    my ($self, $file_id, $file, $fileloc, $error) = @_;
    if (defined $fileloc) {
       warn qq[$Sub: $error in $fileloc\n];
    } else {
       warn qq[$Sub: $error\n];
    }
    $self->{file_parse_errors}{$file_id}{$fileloc // $file} = $error;
}

sub __doc_create {
    my ($self, $arg, $file, $file_id) = @_;
    my $doc;
    eval { $doc = PPI::Document->new($arg) };
    unless (defined $doc) {
        $self->__parse_error($file_id, $file, $file,
                             "PPI::Document creation failed");
    } else {
        my $complete;
        eval { $complete = $doc->complete };
        unless ($complete) {
            my $pseudo = $file eq '-';
            if (!$pseudo && ! -f $file) {
                $self->__parse_error($file_id, $file, undef,
                                     "Missing file");
            } elsif (!$pseudo && ! -s $file) {
                $self->__parse_error($file_id, $file, undef,
                                     "Empty file");
            } else {
                $self->__parse_error($file_id, $file, $file,
                                     "PPI::Document incomplete");
            }
        }
    }
    return $doc;
}

# Process a given filename.
sub __process_file {
    my ($self, $arg, $file, $process_depth) = @_;
    $file //= $arg;
    $self->{file_counts}{$file}++;
    if ($file eq '-') {  # Pseudofile.
        if ($self->{opt}{process_verbose}) {
            printf "$Sub: %*s%s\n", $process_depth + 1, ' ', $file;
        }
        my $file_id = $self->__assign_file_id($file);
        my $doc = $self->__doc_create($arg, $file, $file_id);
        return unless defined $doc;
        $self->{__docscreated}++;
        $self->__process_id($doc, $file_id, $process_depth);
    } elsif ($self->{seen_file}{$file}) {
        if ($self->{opt}{process_verbose} && $self->{opt}{recurse_verbose}) {
            printf "$Sub: %*s%s [seen]\n", $process_depth + 1, ' ', $file;
        }
    } elsif (! $self->{seen_file}{$file}++) {
        if ($self->{opt}{process_verbose}) {
            printf "$Sub: %*s%s\n", $process_depth + 1, ' ', $file;
        }
        my $file_id = $self->__assign_file_id($file);
        my ($cache_filename, $cached, $hash_current,
            $hash_match, $file_mtime) =
            $self->__check_cached($file);
        if ($hash_match) {
            $self->__clear_cached($file_id);
            $self->__import_cached($file_id, $cached);
            $self->__process_cached_incs($file_id, $process_depth);
            $self->{__cachereads}++;
        } else {
            my $doc = $self->__doc_create($arg, $file, $file_id);
            return unless defined $doc;
            $self->__clear_cached($file_id);
            $self->__process_id($doc, $file_id, $process_depth);
            $self->{__docscreated}++;
        }
        if (defined $cache_filename &&
            defined $hash_current &&
            !$hash_match) {
            if ($self->__to_cache($cache_filename, $hash_current,
                                  $file_id, $file_mtime)) {
                if (!$hash_match && defined $cached) {
                    $self->{__cacheupdates}++;
                }
            }
        }
    }
    return $self->{file_id}{$file};
}

# Counter getters.

sub docs_created {
    my ($self) = @_;
    return $self->{__docscreated} // 0;
}

sub cache_reads {
    my ($self) = @_;
    return $self->{__cachereads} // 0;
}

sub cache_writes {
    my ($self) = @_;
    return $self->{__cachewrites} // 0;
}

sub cache_creates {
    my ($self) = @_;
    return $self->{__cachecreates} // 0;
}

sub cache_updates {
    my ($self) = @_;
    return $self->{__cacheupdates} // 0;
}

sub cache_deletes {
    my ($self) = @_;
    return $self->{__cachedeletes} // 0;
}

# For results imported from cache, process any cached inclusions.
# The [6] is the include file, the [7] will become its (new) file id.
sub __process_cached_incs {
    my ($self, $file_id, $process_depth) = @_;
    for my $inc (@{ $self->{file_incs}{$file_id} }) {
        my $include_file = $inc->[6];
        $self->__process_file($include_file, undef,
                              $process_depth + 1);
        $inc->[7] = $self->{file_id}{$include_file};
    }
}

# For freshly computed results, process any cached inclusions.
# The [6] is the include file, the [7] will become its (new) file id.
sub __process_pending_incs {
    my ($self, $file_id, $process_depth) = @_;
    if ($self->{__incs_flush}{$file_id}) {
        $self->{file_incs}{$file_id} =
            delete $self->{__incs_pending}{$file_id};
        for my $inc (@{ $self->{file_incs}{$file_id} }) {
            my $include_file = $inc->[6];
            $self->__process_file($include_file, undef,
                                  $process_depth + 1);
            $inc->[7] = $self->{file_id}{$include_file};
        }
        delete $self->{__incs_flush}{$file_id};  # Defuse the trigger.
    }
}

# Process a given PPI document, that has a given file id.
sub __process_id {
    my ($self, $doc, $file_id, $process_depth) = @_;
    my @queue = @{$doc->{children}};
    my $scope_depth = 0;
    my %package = ( 0 => 'main' );
    my $package = $package{$scope_depth};
    my $prev_package;
    my $elem;
    my $prev_elem;
    my $filename = $FILE_BY_ID{$file_id};
    my $fileloc;
    while (@queue) {
        $elem = shift @queue;
        my $linenumber = $elem->line_number;
        $fileloc = "$filename:$linenumber";
        if (0) {
            printf("$fileloc elem = %s[%s]\n",
                   $filename, ref $elem, $elem->content);
        }
        my @children = exists $elem->{children} ?
            __normalize_whitespace(@{$elem->{children}}) : ();
        if ($elem->isa('PPI::Token::Structure')) {
            # { ... }
            if ($elem->content eq '{') {
                $scope_depth++;
            } elsif ($elem->content eq '}') {
                if ($scope_depth <= 0) {
                    $self->__parse_error($file_id, $filename, $fileloc,
                                         "scope pop underflow");
                } else {
                    $scope_depth--;
                    delete @package{ grep { $_ > $scope_depth } keys %package };
                    use List::Util qw[first];
                    $package =
                        $package{$scope_depth} //
                        first { defined } @package{reverse 0..$scope_depth};
                    if (defined $prev_package && $package ne $prev_package) {
                        if (0) {
                            print "$fileloc: package change: $prev_package -> $package\n";
                        }
                        $self->__close_open_package(
                            $file_id, $prev_package, $prev_elem,
                            $package, $elem);
                    }
                }
            }
        }
        if (@children) {
            if ($elem->isa('PPI::Statement::Package') && @children >= 2) {
                # package ...
                #
                # Remember to test 'use mro' and look for next::can().
                $package = $children[2]->content;
                $package{$scope_depth} = $package;
                if (defined $package && length $package) {
                    # Okay, keep going.
                } else {
                    $self->__parse_error($file_id,
                                         $filename,
                                         $fileloc, "missing package");
                }
                if (defined $prev_package) {
                    if ($package ne $prev_package) {
                        if (0) {
                            print "$fileloc: package change: $prev_package -> $package\n";
                        }
                        $self->__close_open_package(
                            $file_id, $prev_package, $prev_elem,
                            $package, $elem);
                    }
                } else {
                    if (0) {
                        print "$fileloc: package first: $package\n";
                    }
                    $self->__open_package($file_id, $package, $elem);
                }
            } elsif ($elem->isa('PPI::Statement::Sub') &&
                     defined $elem->name &&
                     !$elem->forward  # Not a forward declaration.
                ) {
                # sub ...
                my $sub = $elem->name;
                unless ($sub =~ /::/) {  # sub x::y::z { ... }
                    $package //= 'main';
                    $sub = $package . '::' . $sub;
                }
                my $finish = $elem->block->finish;
                unless (defined $finish) {
                    # E.g. Devel::Peek:debug_flags() fails to have a finish.
                   $finish = $elem;  # Fake it.
                   $self->__parse_error($file_id,
                                        $filename,
                                        $fileloc, "missing finish");
                }
                push @{ $self->{file_subs}{$file_id} },
                     [
                      $sub,                   # 0
                      $elem->line_number,     # 1
                      $elem->column_number,   # 2
                      $finish->line_number,   # 3
                      $finish->column_number, # 4
                     ];
            } elsif ($elem->isa('PPI::Statement')) {
                # use, no, require
                my $stmt_content = $children[0]->content;
                my $include = $children[2];
                next unless defined $include;
                my $include_content = $include->content;
                my $including_module;
                my $including_file;
                if ($elem->isa('PPI::Statement::Include') &&
                    # use/no/require Module
                    $stmt_content =~ /^(?:use|no|require)$/ &&
                    $include->isa('PPI::Token::Word') &&
                    $include_content !~ /^v?5/) {
                    $including_module = 1;
                } elsif ($stmt_content =~ /^(?:require|do)$/ &&
                         $include->isa('PPI::Token::Quote')) {
                    # require/do "file"
                    $including_file = 1;
                } else {
                    # Not a use/no/require/do, quietly exit stage left.
                    next;
                }
                unless (defined $include) {
                    $self->__parse_error($file_id, $fileloc, "missing include");
                    next;
                }
                my $last = $children[-1];
                my $include_file;
                my $include_string;
                if ($including_module) {
                    $include_string = $include_content;
                    $include_file = $self->__find_module($include_content);
                    $self->{file_modules}{$file_id}{$include_content}++;
                    unless (defined $include_file) {
                        $self->{file_missing_modules}{$file_id}{$include_content}{$fileloc}++;
                        warn "$Sub: warning: Failed to find module '$include_string' in $fileloc\n";
                    }
                } elsif ($including_file) {
                    $include_string = $include->string;
                    $include_file = $self->__find_file($include_string);
                    unless (defined $include_file) {
                        warn "$Sub: warning: Failed to find file '$include_string' in $fileloc\n";
                    }
                }
                if (defined $include_file) {
                    if ($self->{opt}{recurse}) {
                        push @{ $self->{__incs_pending}{$file_id} },
                             [
                              $stmt_content,         # 0
                              $elem->line_number,    # 1
                              $elem->column_number,  # 2
                              $last->line_number,    # 3
                              $last->column_number,  # 4
                              $include_string,       # 5
                              $include_file,         # 6
                              # 7 will be the file_id of include_file
                             ];
                    } else {
                        $self->__assign_file_id($include_file);
                    }
                }
            }
        }
        if ($elem->isa('PPI::Structure')) {
            # { ... }
            unshift @queue, $elem->finish if $elem->finish;
            unshift @queue, @children;
            unshift @queue, $elem->start  if $elem->start;
        } else {
            unshift @queue, @children;
        }
        $prev_elem = $elem;
        $prev_package = $package;
    }
    if (defined $elem) {
        $self->{file_lines}{$file_id} = $elem->line_number;
        $self->__close_package($file_id, $package, $elem);
    } elsif (@{ $doc->{children} }) {
        $self->__parse_error($file_id, $filename,
                             "Undefined token when leaving");
    }

    # Mark the __incs_pending as ready to be recursed into.
    $self->{__incs_flush}{$file_id}++;

    $self->__process_pending_incs($file_id, $process_depth);
}

sub __trash_cache {
    my $self = shift;
    delete $self->{result_cache};
    delete $self->{seen_file};
}

# Parse the given filenames (or if a scalar ref, a string of code,
# in which case filename is assumed to be '-').
sub process {
    my $self = shift;
    $self->__trash_cache;
    my $success = 1;
    for my $arg (@_) {
        my $file;
        my $ref = ref $arg;
        if ($ref eq '') {
            $file = $arg;
        } elsif ($ref eq 'SCALAR') {
            $file = '-';
        } else {
            warn "$Sub: Unexpected argument '$arg' (ref: $ref)\n";
            $success = 0;
            next;
        }
        my $file_id = $self->__process_file($arg, $file, 0);
        unless (defined $file_id) {
            $success = 0;
            next;
        }
        $self->{__process}{ $file_id }++;
    }
    return $success;
}

sub process_files_from_cache {
    my $self = shift;
    my %files;
    $self->find_cache_files(\%files);
    if ($self->{opt}{process_verbose} || $self->{opt}{cache_verbose}) {
        my $cache_directory = $self->{opt}{cache_directory};
        printf("$Sub: found %d cache files from %s\n",
               scalar keys %files, $cache_directory);
    }
    $self->process(sort keys %files);
}

sub process_files_from_system {
    my $self = shift;
    my %files;
    $self->find_system_files(\%files);
    if ($self->{opt}{process_verbose}) {
        my $cache_directory = $self->{opt}{cache_directory};
        printf("$Sub: found %d system files from @INC\n",
               scalar keys %files);
    }
    $self->process(sort keys %files);
}

# Returns the seen filenames.
sub files {
    my $self = shift;
    unless (defined $self->{result_cache}{files}) {
        return unless $self->{file_id};
        $self->{result_cache}{files} = [ sort keys %{ $self->{file_id} } ];
    }
    return @{ $self->{result_cache}{files} };
}

# Computes the total number of lines.
sub total_lines {
    my $self = shift;
    unless (defined $self->{result_cache}{total_lines}) {
        return unless $self->{file_lines};
        use List::Util qw[sum];
        $self->{result_cache}{total_lines} = sum grep { defined } values %{ $self->{file_lines} };
    }
    return $self->{result_cache}{total_lines};
}

# Lines in a file.
sub file_lines {
    my ($self, $file) = @_;
    return $self->{file_lines}{$self->{file_id}{$file}};
}

# Returns the known file ids.
sub __file_ids {
    my $self = shift;
    unless (defined $self->{result_cache}{__file_ids}) {
        return unless $FILE_ID;
        $self->{result_cache}{__file_ids} = [ 0..$FILE_ID-1 ];
    }
    return @{ $self->{result_cache}{__file_ids} };
}

# Returns the reference count of a filename.
sub file_count {
    my ($self, $file) = @_;
    return unless $self->{file_counts};
    return $self->{file_counts}->{$file};
}

# Computes the seen modules.
sub __modules {
    my $self = shift;
    unless (defined $self->{result_cache}{modules}) {
        return unless $self->{file_modules};
        delete $self->{modules};
        for my $f ($self->__file_ids) {
            for my $m (keys %{ $self->{file_modules}{$f} }) {
                $self->{modules}{$m} += $self->{file_modules}{$f}{$m};
            }
        }
        $self->{result_cache}{modules} = [ sort keys %{ $self->{modules} } ];
    }
}

# Returns the seen modules.
sub modules {
    my $self = shift;
    $self->__modules;
    return @{ $self->{result_cache}{modules} };
}

# Computes the missing modules.
sub __missing_modules {
    my $self = shift;
    unless (defined $self->{result_cache}{missing_modules}) {
        $self->{result_cache}{missing_modules} //= [];
        $self->{result_cache}{missing_modules_files} //= {};
        $self->{result_cache}{missing_modules_count} //= {};
        return unless $self->{file_missing_modules};
        delete $self->{missing_modules};
        delete $self->{missing_modules_files};
        delete $self->{missing_modules_lines};
        delete $self->{missing_modules_count};
        for my $f ($self->__file_ids) {
            my $file = $FILE_BY_ID{$f};
            for my $m (keys %{ $self->{file_missing_modules}{$f} }) {
                for my $l (keys %{ $self->{file_missing_modules}{$f}{$m} }) {
                    my $c = $self->{file_missing_modules}{$f}{$m}{$l};
                    $self->{missing_modules_files}{$m}{$file} += $c;
                    $self->{missing_modules_lines}{$m}{$l} += $c;
                    $self->{missing_modules_count}{$m} += $c;
                }
            }
        }
        $self->{result_cache}{missing_modules} =
            [ sort keys %{ $self->{missing_modules_files} } ];
    }
}

# Returns the missing modules.
sub missing_modules {
    my $self = shift;
    $self->__missing_modules;
    return @{ $self->{result_cache}{missing_modules} };
}

# Returns the total reference count of a module name.
sub module_count {
    my ($self, $module) = @_;
    $self->__modules;
    return 0 unless $self->{modules};
    return $self->{modules}->{$module} || 0;
}

# Returns the files referring a missing module.
sub missing_module_files {
    my ($self, $module) = @_;
    $self->__missing_modules;
    return unless $self->{missing_modules_files}{$module};
    return sort keys %{ $self->{missing_modules_files}{$module} };
}

# Returns the lines referring a missing module.
sub missing_module_lines {
    my ($self, $module) = @_;
    $self->__missing_modules;
    return unless $self->{missing_modules_lines}{$module};
    return map  { "$_->[0]:$_->[1]" }
           sort { $a->[0] cmp $b->[0] || $a->[1] <=> $b->[1] }
           map  { /^(.+):(\d+)$/ ? [ $1, $2 ] : [ $_, 0 ] }
           keys %{ $self->{missing_modules_lines}{$module} };
}

# Returns the times a missing module was referred.
sub missing_module_count {
    my ($self, $module) = @_;
    $self->__missing_modules;
    return 0 unless $self->{missing_modules_count}{$module};
    return $self->{missing_modules_count}{$module} || 0;
}

# Computes the parse errors.
sub __parse_errors {
    my $self = shift;
    unless (defined $self->{result_cache}{parse_errors_files}) {
        $self->{result_cache}{parse_errors_files} //= [];
        return unless exists $self->{file_parse_errors};
        delete $self->{parse_errors_files};
        for my $f ($self->__file_ids) {
            for my $l (keys %{ $self->{file_parse_errors}{$f} }) {
                $self->{parse_errors_files}{$FILE_BY_ID{$f}}++;
            }
        }
        $self->{result_cache}{parse_errors_files} =
            [ sort keys %{ $self->{parse_errors_files} } ];
    }
}

# Return the files with parse errors.
sub parse_errors_files {
    my $self = shift;
    $self->__parse_errors;
    return @{ $self->{result_cache}{parse_errors_files} };
}

# Return the parse errors in a file, as a hash of filelocation -> error.
sub file_parse_errors {
    my ($self, $file) = @_;
    $self->__parse_errors;
    return unless exists $self->{file_parse_errors};
    return unless defined $file;
    my $file_id = $self->{file_id}{$file};
    return unless defined $file_id;
    return unless exists $self->{file_parse_errors}{$file_id};
    return %{ $self->{file_parse_errors}{$file_id} };
}

# Generates the subs or packages.
sub __subs_or_packages {
    my ($self, $key, $cache) = @_;
    unless (defined $self->{result_cache}{$cache}) {
        return unless $self->{$key};
        my %uniq;
        for my $f ($self->__file_ids) {
            @uniq{ map { $_->[0] } @{ $self->{$key}{$f} } } = ();
        }
        $self->{result_cache}{$cache} = [ sort keys %uniq ];
    }
    return @{ $self->{result_cache}{$cache} };
}

# Returns the subs.
sub subs {
    my $self = shift;
    return $self->__subs_or_packages(file_subs => 'subs');
}

# Returns the packages.
sub packages {
    my $self = shift;
    return $self->__subs_or_packages(file_packages => 'packages');
}

# Generates the subs' or packages' files.
sub __subs_or_packages_and_files {
    my ($self, $key, $cache) = @_;
    unless (defined $self->{result_cache}{$cache}) {
        return [] unless $self->{$key};
        my @cache;
        for my $f ($self->__file_ids) {
            for my $s (@{ $self->{$key}{$f} }) {
                push @cache,
                     [ $s->[0], $f, @{$s}[1..$#$s] ];
            }
        }
        $self->{result_cache}{$cache} = [
            sort { $a->[0] cmp $b->[0] ||
                   $FILE_BY_ID{$a->[1]} cmp $FILE_BY_ID{$b->[1]} ||
                   $a->[2] <=> $b->[2] ||
                   $a->[3] <=> $b->[3] }
            @cache ];
    }
    return $self->{result_cache}{$cache};
}

# Returns the subs' files (an aref to iterate through).
sub __subs_files_full {
    my $self = shift;
    $self->__subs_or_packages_and_files(file_subs => '__subs_files');
}

# Returns the packages' files (an aref to iterate through).
sub __packages_files_full {
    my $self = shift;
    $self->__subs_or_packages_and_files(file_packages => '__packages_files');
}

# Base class for returning iterator results.
package PPI::Xref::IterResultBase {
    # Result as a string.
    # The desired fields are selected, and concatenated.
    sub string {
        my $self = shift;
        return $self->{cb}->($self->{it});
    }
    # Result as an array.
    sub array {
        my $self = shift;
        return @{ $self->{it} };
    }
}

package PPI::Xref::FileIterResult {
    use parent -norequire, 'PPI::Xref::IterResultBase';
}

package PPI::Xref::FileIterBase {
    # Base class method for stepping through the subs' and packages' files.
    # Converts the file id into filename, and returns iterator result.
    sub next {
        # NOTE: while this is an iterator in the sense of returning the next
        # result, this does not compute the next result because all the results
        # have already been computed when the iterator was constructed.
        my $self = shift;
        if ($self->{ix} < @{$self->{it}}) {
            my @it = @{ $self->{it}->[$self->{ix}++] };
            $it[1] = $FILE_BY_ID{$it[1]};
            return bless { it => \@it, cb => $self->{cb} },
                         'PPI::Xref::FileIterResult'
        }
        return;
    }
}

package PPI::Xref::SubsFilesIter {
    use parent -norequire, 'PPI::Xref::FileIterBase';
}

package PPI::Xref::PackagesFilesIter {
    use parent -norequire, 'PPI::Xref::FileIterBase';
}

# Callback generator for subs' and packages' files.
# Selects the desired fields, and concatenates the fields.
sub __file_iter_callback {
    my $opt = shift;
    sub {
        my $self = shift;
        join($opt->{separator} // "\t",
             @{$self}[0, 1, 2],
             @{$self}[$opt->{column} ?
                      ($opt->{finish} ? (3, 4, 5) : (3)) :
                      ($opt->{finish} ? (4) : ())]
            );
    }
}

# Constructor for iterating through the subs' files.
sub subs_files_iter {
    my ($self, $opt) = @_;
    bless {
        it => $self->__subs_files_full,
        ix => 0,
        cb => __file_iter_callback($opt),
    }, 'PPI::Xref::SubsFilesIter';
}

# Constructor for iterating through the packages' files.
sub packages_files_iter {
    my ($self, $opt) = @_;
    bless {
        it => $self->__packages_files_full,
        ix => 0,
        cb => __file_iter_callback($opt),
    }, 'PPI::Xref::PackagesFilesIter';
}

# Generates all the inclusion files and caches them.
# The inclusion files are the file id, followed all its inclusions.
# Sorting is by filename, line, column, and include_string (e.g. Data::Dumper).
sub __incs_files_full {
    my ($self) = @_;
    unless (defined $self->{result_cache}{__incs_files}) {
        my @cache;
        for my $f ($self->__file_ids) {
            for my $i (@{ $self->{file_incs}{$f} }) {
                push @cache, [ $f,      # 0: fileid1
                               @{$i}[1, # 1: line
                                     7, # 2: fileid2
                                     0, # 3: stmt
                                     5, # 4: include_string
                                     2, # 5: col
                                     3, # 6: line
                                     4, # 7: col
                               ] ],
            }
        }
        $self->{result_cache}{__incs_files} = [
            sort { $FILE_BY_ID{$a->[0]} cmp $FILE_BY_ID{$b->[0]} ||
                   $a->[1] <=> $b->[1] ||  # line
                   $a->[2] <=> $b->[2] ||  # column
                   $a->[5] cmp $b->[5] }   # include_string
            @cache ];
    }
    return $self->{result_cache}{__incs_files};
}

# Callback generator for inclusion files.
# Selects the desired fields, and concatenates the fields with the separator.
sub __incs_files_iter_callback {
    my $opt = shift;
    sub {
        my $self = shift;
        join($opt->{separator} // "\t",
             @{$self}[0..4],
             @{$self}[$opt->{column} ?
                      ($opt->{finish} ? (5, 6, 7) : (5)) :
                      ($opt->{finish} ? (6) : ())],
            );
    }
}

package PPI::Xref::IncsFilesIterResult {
    use parent -norequire, 'PPI::Xref::IterResultBase';
}

package PPI::Xref::IncsFilesIter {
    # Iterator stepper for iterating through the inclusion files.
    # Converts the file ids to filenames, and returns iterator result.
    sub next {
        # NOTE: while this is an iterator in the sense of returning the next
        # result, this does not compute the next result because all the results
        # have already been computed when the iterator was constructed.
        my $self = shift;
        if ($self->{ix} < @{$self->{it}}) {
            my @it = @{ $self->{it}->[$self->{ix}++] };
            $it[0] = $FILE_BY_ID{$it[0]};
            $it[2] = $FILE_BY_ID{$it[2]};
            return bless { it => \@it, cb => $self->{cb} },
                         'PPI::Xref::IncsFilesIterResult'
        }
        return;
    }
}

# Constructor for iterating through the inclusion files.
sub incs_files_iter {
    my ($self, $opt) = @_;
    bless {
        it => $self->__incs_files_full,
        ix => 0,
        cb => __incs_files_iter_callback($opt),
    }, 'PPI::Xref::IncsFilesIter';
}

# Recursive generator for inclusion chains.  If there are inclusions
# from this file, recurse for them; if not, aggregate into the result.
sub __incs_chains_recurse {
    my ($self, $file_id, $path, $seen, $result) = @_;
    my @s =
        exists $self->{file_incs}{$file_id} ?
        @{ $self->{file_incs}{$file_id} } : ();
    $seen->{$file_id}++;
    # print "recurse: $FILE_BY_ID{$file_id} path: [@{[map { $FILE_BY_ID{$_} // $_ } @$path]}] \n";
    my $s = 0;
    for my $i (@s) {
        my ($line, $next_file_id) = ($i->[1], $i->[-1]);
        # print "recurse: $FILE_BY_ID{$file_id}:$line -> $FILE_BY_ID{$next_file_id} path: [@{[map { $FILE_BY_ID{$_} // $_ } @$path]}] seen: [@{[sort map { $FILE_BY_ID{$_} } keys %$seen]}]\n";
        # E.g. Carp uses strict, strict requires Carp.
        unless ($seen->{$next_file_id}++) {
            $self->__incs_chains_recurse($next_file_id, [ @$path, $line, $next_file_id ], $seen, $result);
            $s++;
        }
    }
    if ($s == 0) {  # If this was a leaf (no paths leading out), aggregrate result.
        push @{$result}, [ @$path ];
    }
    delete $seen->{$file_id};
}

sub __incs_deps {
    my ($self) = @_;
    unless (defined $self->{result_cache}{__incs_deps}) {
        my %pred;
        my %succ;
        my %line;
        for my $fi ($self->__file_ids) {
            if (exists $self->{file_incs}{$fi}) {
                for my $g (@{ $self->{file_incs}{$fi} }) {
                    my ($gl, $gi) = @{ $g }[ 1, 7 ];
                    $succ{$fi}{$gi}{$gl}++;
                    $pred{$gi}{$fi}{$gl}++;
                    $line{$fi}{$gl}{$gi}++;
                }
            }
        }
        my %singleton;
        my %leaf;
        my %root;
        my %branch;
        for my $s ($self->__file_ids) {
            my @s = exists $succ{$s} ? keys %{$succ{$s}} : ();
            my @p = exists $pred{$s} ? keys %{$pred{$s}} : ();
            if (@s == 0) {
                if (@p == 0) {
                    $singleton{$s}++;
                } else {
                    $leaf{$s}++;
                }
            } elsif (@p == 0) {
                $root{$s}++;
            } else {
                $branch{$s}++;
            }
        }
        $self->{result_cache}{__incs_deps} = {
            pred => \%pred,
            succ => \%succ,
            line => \%line,
            singleton => \%singleton,
            leaf => \%leaf,
            root => \%root,
            branch => \%branch,
            parent => $self,
        };
    }
    return $self->{result_cache}{__incs_deps};
}

package PPI::Xref::IncsDeps {
    sub files {
        my ($self) = @_;
        return $self->{parent}->files;
    }
    sub __file_id {
        my ($self, $file) = @_;
        return $self->{parent}{file_id}{$file};
    }
    sub __by_file {
        my ($self, $key, $file) = @_;
        my $file_id = $self->__file_id($file);
        return unless defined $file_id && exists $self->{$key}{$file_id};
        return keys %{ $self->{$key}{$file_id} };
    }
    sub __filenames {
        my $self = shift;
        return map { $FILE_BY_ID{$_} } @_;
    }
    sub __predecessors {
        my ($self, $file) = @_;
        return _$self->__by_file(pred => $file);
    }
    sub __successors {
        my ($self, $file) = @_;
        return $self->__by_file(succ => $file);
    }
    sub predecessors {
        my ($self, $file) = @_;
        return $self->__filenames(_$self->__predecessors($file));
    }
    sub successors {
        my ($self, $file) = @_;
        return $self->__filenames($self->__successors($file));
    }
    sub __files {
        my ($self, $key) = @_;
        return exists $self->{$key} ?
            map { $FILE_BY_ID{$_} } keys %{ $self->{$key} } : ();
    }
    sub __roots {
        my ($self) = @_;
        return exists $self->{root} ? keys %{ $self->{root} } : ();
    }
    sub __singletons {
        my ($self) = @_;
        return exists $self->{singleton} ? keys %{ $self->{singleton} } : ();
    }
    sub roots {
        my ($self) = @_;
        return $self->__files('root');
    }
    sub leaves {
        my ($self) = @_;
        return $self->__files('leaf');
    }
    sub singletons {
        my ($self) = @_;
        return $self->__files('singleton');
    }
    sub branches {
        my ($self) = @_;
        return $self->__files('branch');
    }
    sub __file_kind {
        my ($self, $file_id) = @_;
        return unless defined $file_id;
        return 'branch'    if exists $self->{branch}   {$file_id};
        return 'leaf'      if exists $self->{leaf}     {$file_id};
        return 'root'      if exists $self->{root}     {$file_id};
        return 'singleton' if exists $self->{singleton}{$file_id};
        return;
    }
    sub file_kind {
        my ($self, $file) = @_;
        return $self->__file_kind($self->__file_id($file));
    }
}

sub incs_deps {
    my ($self) = @_;
    bless $self->__incs_deps, 'PPI::Xref::IncsDeps';
}

sub __incs_chains_iter {
    my ($self, $opt) = @_;
    my %iter;
    my $deps = $self->incs_deps;
    if (defined $deps) {
        $iter{next} = sub {
            my ($iterself) = @_;
            until ($iterself->{done}) {
                unless (defined $iterself->{path} && @{ $iterself->{path} }) {
                    unless (defined $iterself->{roots}) {
                        my @roots = (
                            $deps->__roots,
                            $deps->__singletons,
                            );
                        my %roots;
                        @roots{@roots} = ();
                        if (exists $self->{__process}) {
                            for my $id (keys %{ $self->{__process} }) {
                                push @roots, $id unless exists $roots{$id};
                            }
                        }
                        $iterself->{roots} = \@roots;
                    }
                    my $root = shift @{ $iterself->{roots} };
                    unless (defined $root) {
                        $iterself->{done}++;
                        return;
                    }
                    $iterself->{path} = [ $root ];
                    # E.g. Carp uses strict, strict requires Carp, and also
                    # the dependency trees are very probably not clean DAGs.
                    $iterself->{seen} = { $root => { 0 => 1 } };
                };
                while (@{ $iterself->{path} }) {
                    my $curr = $iterself->{path}[-1];
                    my $pushed = 0;
                  SUCC: {
                      if (exists $deps->{line}{$curr}) {
                          for my $line (sort { $a <=> $b }
                                        keys %{ $deps->{line}{$curr} }) {
                              for my $succ (sort { $a cmp $b }
                                            keys %{ $deps->{line}{$curr}{$line} }) {
                                  unless ($iterself->{seen}{$succ}{$line}++) {
                                      push @{ $iterself->{path} }, $line, $succ;
                                      $pushed++;
                                      last SUCC;
                                  }
                              }
                          }
                      }
                    }
                    unless ($pushed) {
                        if (my @path = @{ $iterself->{path} }) {
                            @path = reverse @path if $opt->{reverse_chains};
                            if (@path > 1) {
                                splice @{ $iterself->{path} }, -2;  # Double-pop.
                                if ($self->{lastpush}) {
                                    $self->{lastpush} = $pushed;
                                    return @path;
                                }
                            } else {
                                $iterself->{path} = [];
                                my $kind = $deps->__file_kind($curr);
                                if (defined $kind && $kind eq 'singleton') {
                                    return @path;
                                }
                            }
                        }
                    }
                    $self->{lastpush} = $pushed;
                }  # while
            }
        };
    }
    return \%iter;
}

# Callback generator for inclusion chains.
# Simply concatenates the fields with the separator.
sub __incs_chains_iter_callback {
    my $opt = shift;
    sub {
        my $self = shift;
        join($opt->{separator} // "\t", @{$self} );
    }
}

package PPI::Xref::IncsChainsIterResult {
    use parent -norequire, 'PPI::Xref::IterResultBase';
}

package PPI::Xref::IncsChainsIter {
    # Iterator stepper for iterating through the inclusion chains.
    # Converts the file ids to filenames, and returns iterator result.
    sub next {
        my $self = shift;
        if (my @it = $self->{it}{next}->($self)) {
            for (my $i = 0; $i < @it; $i += 2) {
                $it[$i] = $FILE_BY_ID{$it[$i]};
            }
            return bless { it => \@it, cb => $self->{cb} },
                         'PPI::Xref::IncsChainsIterResult';
        }
        return;
    }
}

# Constructor for iterating through the inclusion chains.
sub incs_chains_iter {
    my ($self, $opt) = @_;
    bless {
        it => $self->__incs_chains_iter($opt),
        ix => 0,
        cb => __incs_chains_iter_callback($opt),
    }, 'PPI::Xref::IncsChainsIter';
}

sub looks_like_cache_file {
    my ($self, $file) = @_;

    my $cache_directory = $self->{opt}{cache_directory};
    return unless defined $cache_directory;

    return 0 if $file =~ m{\.\.};

    return $file =~ m{^\Q$cache_directory\E/.+\Q$CACHE_EXT\E$};
}

sub cache_delete {
    my $self = shift;
    my $cache_directory = $self->{opt}{cache_directory};
    unless (defined $cache_directory) {
        warn "$Sub: cache_directory undefined\n";
        return;
    }
    my $delete_count = 0;
    for my $file (@_) {
        if (!File::Spec->file_name_is_absolute($file) ||
            $file =~ m{\.\.} ||
            ($file !~ m{_p[ml](?:\Q$CACHE_EXT\E)?$} &&
             $file !~ m{.p[ml]$})) {
            # Paranoia check one.
            warn "$Sub: Skipping unexpected file: '$file'\n";
            next;
        }
        my $cache_file =
            $file =~ /\Q$CACHE_EXT\E$/ ?
            $file : $self->__cache_filename($file);
        # Paranoia check two.  Both paranoia checks are needed.
        unless ($self->looks_like_cache_file($cache_file)) {
            warn "$Sub: Skipping unexpected cache file: '$cache_file'\n";
            next;
        }
        if ($self->{opt}{cache_verbose}) {
            print "cache_delete: deleting $cache_file\n";
        }
        if (unlink $cache_file) {
            $delete_count++;
            $self->{__cachedeletes}++;
        }
    }
    return $delete_count;
}

sub __unparse_cache_filename {
    my ($self, $cache_filename) = @_;

    my $cache_directory = $self->{opt}{cache_directory};
    return unless defined $cache_directory;

    return unless $cache_filename =~ s{\Q$CACHE_EXT\E$}{};

    my $cache_prefix_length = $self->{__cache_prefix_length};
    return unless length($cache_filename) > $cache_prefix_length;

    my $prefix = substr($cache_filename, 0, $cache_prefix_length);
    return unless $prefix =~ m{^\Q$cache_directory\E(?:/|\\)$};

    my $path = substr($cache_filename, $cache_prefix_length - 1);

    $path =~ s{_(p[ml])$}{\.$1};  # _pm -> .pm, _pl -> .pl

    if ($^O eq 'MSWin32') {
      # \c\a\b -> c:/a/b
      $path =~ s{\\}{/}g;
      if ($path =~ m{^/([A-Z])(/.+)}) {
        my $volpath = "$1:$2";
        if (-f $volpath) {
          $path = $volpath;
        }
      }
    }

    return $path;
}

# Given an xref, find all the cache files under its cache directory,
# and add their filenames to href.
sub find_cache_files {
    my ($self, $files) = @_;

    my $cache_directory = $self->{opt}{cache_directory};
    unless (defined $cache_directory) {
        warn "$Sub: cache_directory undefined\n";
        return;
    }

    use File::Find qw[find];

    find(
        sub {
            if (/\.p[ml]\Q$CACHE_EXT\E$/) {
                my $name = $self->__unparse_cache_filename($File::Find::name);
                $files->{$name} = $File::Find::name;
            }
        },
        $cache_directory);
}

# Given an xref, find all the pm files under its INC,
# and add their filenames to href.
sub find_system_files {
    my ($self, $files) = @_;

    use File::Find qw[find];

    for my $d (@{ $self->INC }) {
        find(
            sub {
                if (/\.p[ml]$/) {
                    $files->{$File::Find::name} = $File::Find::name;
                }
            },
            $d);
    }
}

1;
__DATA__