File Coverage

blib/lib/File/Locate/Iterator.pm

Criterion	Covered	Total	%
statement	80	91	87.9
branch	53	66	80.3
condition	2	3	66.6
subroutine	10	12	83.3
pod	3	3	100.0
total	148	175	84.5

# Copyright 2009, 2010, 2011, 2012, 2013, 2014, 2017, 2018, 2019 Kevin Ryde.
#
# This file is part of File-Locate-Iterator.
#
# File-Locate-Iterator is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License as published
# by the Free Software Foundation; either version 3, or (at your option)
# any later version.
#
# File-Locate-Iterator is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with File-Locate-Iterator; see the file COPYING.  Failing that, go to
# .


# Maybe:
#     ignore_case      globs and suffixes, not easily done to regexs
#     glob_ignore_case
#     suffix_ignore_case
#


package File::Locate::Iterator;
use 5.006;  # for qr//, and open anonymous handles
use strict;
use warnings;
use Carp;

use DynaLoader;
our @ISA = ('DynaLoader');

our $VERSION = 28;

# uncomment this to run the ### lines
#use Devel::Comments;


if (eval { __PACKAGE__->bootstrap($VERSION) }) {
  ### FLI next() from XS ...
} else {
  ### FLI next() in perl, XS didn't load: $@

  eval "\n#line ".(__LINE__+1)." \"".__FILE__."\"\n" . <<'HERE' or die;
use strict;
use warnings;
use File::FnMatch;

sub _UNEXPECTED_EOF {
  my ($self) = @_;
  undef $self->{'entry'};
  croak 'Invalid database contents (unexpected EOF)';
}
sub _ERROR_READING {
  my ($self) = @_;
  undef $self->{'entry'};
  croak 'Error reading database: ',$!;
}
sub _BAD_SHARE {
  my ($self, $sharelen) = @_;
  undef $self->{'entry'};
  croak "Invalid database contents (bad share length $sharelen)";
}
sub next {
  my ($self) = @_;
  ### FLI PP next()

  my $sharelen = $self->{'sharelen'};
  my $entry = $self->{'entry'};
  my $regexp = $self->{'regexp'};
  my $globs = $self->{'globs'};

  if (my $mref = $self->{'mref'}) {
    my $pos = $self->{'pos'};
  MREF_LOOP: for (;;) {
      #### pos in map: sprintf('%#x', $pos)
      if ($pos >= length ($$mref)) {
        undef $self->{'entry'};
        return; # end of file
      }

      my ($adjshare) = unpack 'c', substr ($$mref, $pos++, 1);
      if ($adjshare == -128) {
        #### 2byte pos: sprintf('%#X', $pos)
        # print ord(substr ($$mref,$pos,1)),"\n";
        # print ord(substr ($$mref,$pos+1,1)),"\n";

        if ($pos+2 > length ($$mref)) { goto &_UNEXPECTED_EOF; }

        # for perl 5.10 up could use 's>' for signed 16-bit big-endian,
        # instead of getting unsigned and stepping down
        ($adjshare) = unpack 'n', substr ($$mref, $pos, 2);
        if ($adjshare >= 32768) { $adjshare -= 65536; }

        $pos += 2;
      }
      ### $adjshare
      $sharelen += $adjshare;
      # print "share now $sharelen\n";
      if ($sharelen < 0 || $sharelen > length($entry)) {
        push @_, $sharelen; goto &_BAD_SHARE;
      }

      my $end = index ($$mref, "\0", $pos);
      # print "$pos to $end\n";
      if ($end < 0) { goto &_UNEXPECTED_EOF; }

      $entry = (substr($entry,0,$sharelen)
                . substr ($$mref, $pos, $end-$pos));
      $pos = $end + 1;

      if ($regexp) {
        last if $entry =~ $regexp;
      } elsif (! $globs) {
        last;
      }
      if ($globs) {
        foreach my $glob (@$globs) {
          last MREF_LOOP if File::FnMatch::fnmatch($glob,$entry)
        }
      }
    }
    $self->{'pos'} = $pos;

  } else {
    local $/ = "\0"; # readline() to \0

    my $fh = $self->{'fh'};
    ### pos tell(fh): sprintf('%#x',tell($fh))
  IO_LOOP: for (;;) {
      my $adjshare;
      unless (my $got = read $fh, $adjshare, 1) {
        if (defined $got) {
          undef $self->{'entry'};
          return; # EOF
        }
        goto &_ERROR_READING;
      }

      ($adjshare) = unpack 'c', $adjshare;
      if ($adjshare == -128) {
        my $got = read $fh, $adjshare, 2;
        if (! defined $got) { goto &_ERROR_READING; }
        if ($got != 2) { goto &_UNEXPECTED_EOF; }

        # for perl 5.10 up could use 's>' for signed 16-bit big-endian
        # pack, instead of getting unsigned and stepping down
        ($adjshare) = unpack 'n', $adjshare;
        if ($adjshare >= 32768) { $adjshare -= 65536; }
      }
      ### $adjshare

      $sharelen += $adjshare;
      ### share now: $sharelen
      if ($sharelen < 0 || $sharelen > length($entry)) {
        push @_, $sharelen; goto &_BAD_SHARE;
      }

      my $part;
      {
        # perlfunc.pod of 5.10.0 for readline() says you can clear $!
        # then check it afterwards for an error indication, but that's
        # wrong, $! ends up set to EBADF when filling the PerlIO buffer,
        # which means if the readline crosses a 1024 byte boundary
        # (something in attempting a fast gets then falling back ...)

        $part = readline $fh;
        if (! defined $part) { goto &_UNEXPECTED_EOF; }

        ### part: $part
        chomp $part or goto &_UNEXPECTED_EOF;
      }

      $entry = substr($entry,0,$sharelen) . $part;

      if ($regexp) {
        last if $entry =~ $regexp;
      } elsif (! $globs) {
        last;
      }
      if ($globs) {
        foreach my $glob (@$globs) {
          last IO_LOOP if File::FnMatch::fnmatch($glob,$entry)
        }
      }
    }
  }

  $self->{'sharelen'} = $sharelen;
  return ($self->{'entry'} = $entry);
}

1;

HERE
}

use constant default_use_mmap => 'if_sensible';
my $header = "\0LOCATE02\0";


# Default path these days is /var/cache/locate/locatedb.
#
# Back in findutils 4.1 it was $(localstatedir)/locatedb, but there seems to
# have been no way to ask about the location.
#
sub default_database_file {
  # my ($class) = @_;
  if (defined (my $env = $ENV{'LOCATE_PATH'})) {
    return $env;
  } else {
    return '/var/cache/locate/locatedb';
  }
}

# The fields, all meant to be private, are:
#
# regexp
#     qr// regexp of all the 'regexp', 'regexps', 'suffix' and 'suffixes'
#     parameters.  If no such matches then no such field.  When the field
#     exists an entry must match the regexp or is skipped.
#
# globs
#     arrayref of strings which are globs to fnmatch().  If no globs then no
#     such field.  When the field exists an entry must match at least one of
#     the globs.
#
# mref
#     Ref to a scalar which holds the database contents, or undef if using
#     fh instead.  It's either a ref to the 'database_str' parameter passed
#     in, or a ref to a scalar created as an mmap of the file.  The mmap one
#     is shared among iterators through the File::Locate::Iterator::FileMap
#     caching.
#
# fh
#     When mref is undef, ref file handle which is to be read from,
#     otherwise no such field.  This can be either the 'database_fh'
#     parameter or an opened anonymous handle of the 'database_file'
#     parameter.
#
#     When mmap is used the 'database_fh' is not held here.  The mmap is
#     made (or rather, looked up in the FileMap cache), and the handle is
#     then no longer needed and can be closed or garbage collected in the
#     caller.
#
# fh_start
#     When fh is set, the tell($fh) position just after the $header in that
#     fh.  This is where to seek() back to for a $it->rewind.  If tell()
#     failed then this is -1 and $it->rewind is not possible.
#
#     Normally fh_start is simply length($header) for a database starting at
#     the start of the file, but a database_fh arg which is positioned at
#     some offset into a file can be read and remembering an fh_start
#     position lets $it->rewind work on it too.
#
# fm
#     When using mmap, a File::Locate::Iterator::FileMap object which is the
#     cache entry for the database file, otherwise no such field.  This is
#     hung onto to keep it alive while in use.  $self->{'mref'} is
#     $fm->mmapref in this case.
#
# pos
#     When mref is not undef, an integer offset into the $$mref string which
#     is the current read position.  The file header is checked in new() so
#     the initial value is length($header), ie. 10, the position of the
#     first entry (or possibly EOF).
#
# entry
#     String of the last database entry returned, or no such field before
#     the first is read, or undef after EOF is hit.  Might be undef instead
#     of not existing if a hypothetical seek() goes back to the start of the
#     file.
#
# sharelen
#     Integer which is the number of leading bytes of 'entry' which the next
#     entry will share with that previous entry.  Initially 0.
#
#     This is modified successively by the "adjshare" of each entry as each
#     takes more or less of the preceding entry.  An adjshare can range from
#     -sharelen to take nothing at all of the previous entry, up to
#     length($entry)-sharelen to increment up to take all of the previous
#     entry.
#
sub new {
  my ($class, %options) = @_;
  ### FLI new(): %options

  # delete 'regexp' field if it's undef, as the XS code wants no 'regexp'
  # field for no regexps, not a field set to undef
  my @regexps;
  if (defined (my $regexp = delete $options{'regexp'})) {
    push @regexps, $regexp;
  }
  if (my $regexps = delete $options{'regexps'}) {
    push @regexps, @$regexps;
  }
  foreach my $suffix (defined $options{'suffix'} ? $options{'suffix'} : (),
                      @{$options{'suffixes'}}) {
    push @regexps, quotemeta($suffix) . '$';
  }
  ### @regexps

  # as per findutils locate.c locate() function, pattern with * ? or [ is a
  # glob, anything else is a literal match
  #
  my @globs = (defined $options{'glob'} ? $options{'glob'} : (),
               @{$options{'globs'} || []});
  @globs = grep { ($_ =~ /[[*?]/
                   || do { push @regexps, quotemeta($_); 0 })
                } @globs;
  ### @globs

  my $self = bless { entry    => '',
                     sharelen => 0,
                   }, $class;

  if (@regexps) {
    my $regexp = join ('|', @regexps);
    $self->{'regexp'} = qr/$regexp/s;
  }
  if (@globs) {
    $self->{'globs'} = \@globs;
  }

  ### regexp: $self->{'regexp'}
  ### globs : $self->{'globs'}

  if (defined (my $ref = $options{'database_str_ref'})) {
    $self->{'mref'} = $ref;

  } elsif (defined $options{'database_str'}) {
    $self->{'mref'} = \$options{'database_str'};

  } else {
    my $use_mmap = (defined $options{'use_mmap'}
                    ? $options{'use_mmap'}
                    : $class->default_use_mmap);
    ### $use_mmap
    if ($use_mmap) {
      if (! eval { require File::Locate::Iterator::FileMap }) {
        ### FileMap not possible: $@
        $use_mmap = 0;
      }
    }

    my $fh = $options{'database_fh'};
    if (defined $fh) {
      if ($use_mmap eq 'if_sensible'
          && File::Locate::Iterator::FileMap::_have_mmap_layer($fh)) {
        ### already have mmap layer, not sensible to mmap again
        $use_mmap = 0;
      }
    } else {
      my $file = (defined $options{'database_file'}
                  ? $options{'database_file'}
                  : $class->default_database_file);
      ### open database_file: $file

      # Crib note: '<:raw' means without :perlio buffering, whereas
      # binmode() preserves that buffering, assuming it's in the $ENV{'PERLIO'}
      # defaults.  Also :raw is not available in perl 5.6.
      open $fh, '<', $file
        or croak "Cannot open $file: $!";
      binmode($fh)
        or croak "Cannot set binary mode";
    }

    if ($use_mmap eq 'if_sensible') {
      $use_mmap = (File::Locate::Iterator::FileMap::_mmap_size_excessive($fh)
                   ? 0
                   : 'if_possible');
      ### if_sensible after size check becomes: $use_mmap
    }

    if ($use_mmap) {
      ### attempt mmap: $fh, (-s $fh)

      # There's many ways an mmap can fail, just chuck an eval on FileMap /
      # File::Map it to catch them all.
      # - An ordinary readable file of length zero may fail per POSIX, and
      #   that's how it is in linux kernel post 2.6.12.  However File::Map
      #   0.20 takes care of returning an empty string for that.
      # - A char special usually gives 0 for its length, even for instance
      #   linux kernel special files like /proc/meminfo.  Char specials can
      #   often be mapped perfectly well, but without a length don't know
      #   how much to look at.  For that reason "if_possible" restricts to
      #   ordinary files, though forced "use_mmap=>1" just goes ahead anyway.
      #
      if ($use_mmap eq 'if_possible') {
        if (! -f $fh) {
          ### if_possible, not a plain file, consider not mmappable
        } else {
          if (! eval { $self->{'fm'}
                         = File::Locate::Iterator::FileMap->get($fh) }) {
            ### mmap failed: $@
          }
        }
      } else {
        $self->{'fm'} = File::Locate::Iterator::FileMap->get($fh);
      }
    }
    if ($self->{'fm'}) {
      $self->{'mref'} = $self->{'fm'}->mmap_ref;
    } else {
      $self->{'fh'} = $fh;
    }
  }

  if (my $mref = $self->{'mref'}) {
    unless ($$mref =~ /^\Q$header/o) { goto &_ERROR_BAD_HEADER }
    $self->{'pos'} = length($header);
  } else {
    my $got = '';
    read $self->{'fh'}, $got, length($header);
    if ($got ne $header) { goto &_ERROR_BAD_HEADER }
    $self->{'fh_start'} = tell $self->{'fh'};
  }

  return $self;
}
sub _ERROR_BAD_HEADER {
  croak 'Invalid database contents (no LOCATE02 header)';
}

sub rewind {
  my ($self) = @_;

  $self->{'sharelen'} = 0;
  $self->{'entry'} = '';
  if ($self->{'mref'}) {
    $self->{'pos'} = length($header);
  } else {
    $self->{'fh_start'} > 0
      or croak "Cannot seek database";
    seek ($self->{'fh'}, $self->{'fh_start'}, 0)
      or croak "Cannot seek database: $!";
  }
}

# return true if mmap is in use
# (an actual mmap, not the slightly similar 'database_str' option)
# this is meant for internal use as a diagnostic ...
sub _using_mmap {
  my ($self) = @_;
  return defined $self->{'fm'};
}

# Not yet documented, likely worthwhile as long as it works properly.
# Return empty list for nothing yet?  Same as next().
# Return empty list at EOF?  At EOF 'entry' is undefed out.
#
# =item C<< $entry = $it->current >>
#
# Return the current entry from the database, meaning the same as the last
# call to C returned.  At the start of the database (before the first
# C) or at end of the database the return is an empty list.
#
#     while (defined $it->next) {
#         ...
#         print $it->current,"\n";
#     }
#
sub _current {
  my ($self) = @_;
  if (defined $self->{'entry'}) {
    return $self->{'entry'};
  } else {
    return;
  }
}


1;
__END__