File Coverage

blib/lib/MIME/Parser/Filer.pm

Criterion	Covered	Total	%
statement	153	167	91.6
branch	49	72	68.0
condition	12	29	41.3
subroutine	31	33	93.9
pod	15	19	78.9
total	260	320	81.2

package MIME::Parser::Filer;

=head1 NAME

MIME::Parser::Filer - manage file-output of the parser


=head1 SYNOPSIS

Before reading further, you should see L to make sure that
you understand where this module fits into the grand scheme of things.
Go on, do it now.  I'll wait.

Ready?  Ok... now read L<"DESCRIPTION"> below, and everything else
should make sense.


=head2 Public interface

    ### Create a "filer" of the desired class:
    my $filer = MIME::Parser::FileInto->new($dir);
    my $filer = MIME::Parser::FileUnder->new($basedir);
    ...

    ### Want added security?  Don't let outsiders name your files:
    $filer->ignore_filename(1);

    ### Prepare for the parsing of a new top-level message:
    $filer->init_parse;

    ### Return the path where this message's data should be placed:
    $path = $filer->output_path($head);


=head2 Semi-public interface

These methods might be overridden or ignored in some subclasses,
so they don't all make sense in all circumstances:

    ### Tweak the mapping from content-type to extension:
    $emap = $filer->output_extension_map;
    $emap->{"text/html"} = ".htm";




=head1 DESCRIPTION


=head2 How this class is used when parsing

When a MIME::Parser decides that it wants to output a file to disk,
it uses its "Filer" object -- an instance of a MIME::Parser::Filer
subclass -- to determine where to put the file.

Every parser has a single Filer object, which it uses for all
parsing.  You can get the Filer for a given $parser like this:

    $filer = $parser->filer;

At the beginning of each C, the filer's internal state
is reset by the parser:

    $parser->filer->init_parse;

The parser can then get a path for each entity in the message
by handing that entity's header (a MIME::Head) to the filer
and having it do the work, like this:

    $new_file = $parser->filer->output_path($head);

Since it's nice to be able to clean up after a parse (especially
a failed parse), the parser tells the filer when it has actually
used a path:

    $parser->filer->purgeable($new_file);

Then, if you want to clean up the files which were created for a
particular parse (and also any directories that the Filer created),
you would do this:

    $parser->filer->purge;



=head2 Writing your own subclasses

There are two standard "Filer" subclasses (see below):
B, which throws all files from all parses
into the same directory, and B (preferred), which
creates a subdirectory for each message.  Hopefully, these will be
sufficient for most uses, but just in case...

The only method you have to override is L:

    $filer->output_path($head);

This method is invoked by MIME::Parser when it wants to put a
decoded message body in an output file.  The method should return a
path to the file to create.  Failure is indicated by throwing an
exception.

The path returned by C should be "ready for open()":
any necessary parent directories need to exist at that point.
These directories can be created by the Filer, if course, and they
should be marked as B if a purge should delete them.

Actually, if your issue is more I the files go than
what they're named, you can use the default L
method and just override one of its components:

    $dir  = $filer->output_dir($head);
    $name = $filer->output_filename($head);
    ...



=head1 PUBLIC INTERFACE


=head2 MIME::Parser::Filer

This is the abstract superclass of all "filer" objects.

=over 4

=cut

use strict;

### Kit modules:
use MIME::Tools qw(:msgtypes);
use File::Spec;
use File::Path qw(rmtree);
use MIME::WordDecoder;

### Output path uniquifiers:
my $GFileNo = 0;
my $GSubdirNo = 0;

### Map content-type to extension.
### If we can't map "major/minor", we try "major/*", then use "*/*".
my %DefaultTypeToExt =
qw(

application/andrew-inset	.ez
application/octet-stream	.bin
application/oda			.oda
application/pdf			.pdf
application/pgp			.pgp
application/postscript		.ps
application/rtf			.rtf
application/x-bcpio		.bcpio
application/x-chess-pgn		.pgn
application/x-cpio		.cpio
application/x-csh		.csh
application/x-dvi		.dvi
application/x-gtar		.gtar
application/x-gunzip		.gz
application/x-hdf		.hdf
application/x-latex		.latex
application/x-mif		.mif
application/x-netcdf		.cdf
application/x-netcdf		.nc
application/x-sh		.sh
application/x-shar		.shar
application/x-sv4cpio		.sv4cpio
application/x-sv4crc		.sv4crc
application/x-tar		.tar
application/x-tcl		.tcl
application/x-tex		.tex
application/x-texinfo		.texi
application/x-troff		.roff
application/x-troff		.tr
application/x-troff-man		.man
application/x-troff-me		.me
application/x-troff-ms		.ms
application/x-ustar		.ustar
application/x-wais-source	.src
application/zip			.zip

audio/basic			.snd
audio/ulaw			.au
audio/x-aiff			.aiff
audio/x-wav			.wav

image/gif			.gif
image/ief			.ief
image/jpeg			.jpg
image/png                       .png
image/xbm                       .xbm
image/tiff			.tif
image/x-cmu-raster		.ras
image/x-portable-anymap		.pnm
image/x-portable-bitmap		.pbm
image/x-portable-graymap	.pgm
image/x-portable-pixmap		.ppm
image/x-rgb			.rgb
image/x-xbitmap			.xbm
image/x-xpixmap			.xpm
image/x-xwindowdump		.xwd

text/*                          .txt
text/html			.html
text/plain			.txt
text/richtext			.rtx
text/tab-separated-values	.tsv
text/x-setext			.etx
text/x-vcard                    .vcf

video/mpeg			.mpg
video/quicktime			.mov
video/x-msvideo			.avi
video/x-sgi-movie		.movie

message/*                       .msg

*/*                             .dat

);

#------------------------------

=item new INITARGS...

I
Create a new outputter for the given parser.
Any subsequent arguments are given to init(), which subclasses should
override for their own use (the default init does nothing).

=cut

sub new {
    my ($class, @initargs) = @_;
    my $self = bless {
	MPF_Prefix    => "msg",
	MPF_Dir       => ".",
	MPF_Ext       => { %DefaultTypeToExt },
	MPF_Purgeable => [],       ### files created by the last parse

	MPF_MaxName   => 80,       ### max filename before treated as evil
	MPF_TrimRoot  => 14,       ### trim root to this length
	MPF_TrimExt   => 3,        ### trim extension to this length
    }, $class;
    $self->init(@initargs);
    $self;
}

sub init {
    ### no-op
}

#------------------------------
# My PID, but untainted
#
sub untainted_pid
{
        if ($$ =~ /^(\d+)$/) {
                return $1;
        }
        # Can't happen...
        return "0";
}

#------------------------------
#
# cleanup_dir
#
# Instance method, private.
# Cleanup a directory, defaulting empty to "."
#
sub cleanup_dir {
    my ($self, $dir) = @_;
    $dir = '.' if (!defined($dir) || ($dir eq ''));   # coerce empty to "."
    $dir = '/.' if ($dir eq '/');   # coerce "/" so "$dir/$filename" works
    $dir =~ s|/$||;                 # be nice: get rid of any trailing "/"
    $dir;
}

#------------------------------

=item results RESULTS

I
Link this filer to a MIME::Parser::Results object which will
tally the messages.  Notice that we avoid linking it to the
parser to avoid circular reference!

=cut

sub results {
    my ($self, $results) = @_;
    $self->{MPF_Results} = $results if (@_ > 1);
    $self->{MPF_Results};
}

### Log debug messages:
sub debug {
    my $self = shift;
    if (MIME::Tools->debugging()) {
	if ($self->{MPF_Results}) {
	    unshift @_, $self->{MPF_Results}->indent;
	    $self->{MPF_Results}->msg($M_DEBUG, @_);
	}
	MIME::Tools::debug(@_);
    }
}

### Log warning messages:
sub whine {
    my $self = shift;
    if ($self->{MPF_Results}) {
	unshift @_, $self->{MPF_Results}->indent;
	$self->{MPF_Results}->msg($M_WARNING, @_);
    }
    MIME::Tools::whine(@_);
}

#------------------------------

=item init_parse

I
Prepare to start parsing a new message.
Subclasses should always be sure to invoke the inherited method.

=cut

sub init_parse {
    my $self = shift;
    $self->{MPF_Purgeable} = [];
}

#------------------------------

=item evil_filename FILENAME

I
Is this an evil filename; i.e., one which should not be used
in generating a disk file name?  It is if any of these are true:

    * it is empty or entirely whitespace
    * it contains leading or trailing whitespace
    * it is a string of dots: ".", "..", etc.
    * it contains characters not in the set: "A" - "Z", "a" - "z",
      "0" - "9", "-", "_", "+", "=", ".", ",", "@", "#",
      "$", and " ".
    * it is too long

If you just want to change this behavior, you should override
this method in the subclass of MIME::Parser::Filer that you use.

B at the time this method is invoked, the FILENAME has
already been unmime'd into the local character set.
If you're using any character set other than ASCII, ISO-8859-*,
or UTF-8, the interpretation of the "path" characters might be
very different, and you will probably need to override this method.
See L for more details.

B subclasses of MIME::Parser::Filer which override
output_path() might not consult this method; note, however, that
the built-in subclasses do consult it.

I
version.  Thanks to Nickolay Saukh for noting that evil is in the
eye of the beholder.>

=cut

sub evil_filename {
    my ($self, $name) = @_;

    $self->debug("is this evil? '$name'");

    return 1 if (!defined($name) or ($name eq ''));   ### empty
    return 1 if ($name =~ m{(^\s)|(\s+\Z)});  ### leading/trailing whitespace
    return 1 if ($name =~ m{^\.+\Z});         ### dots
    return 1 if ($name =~ /[^-A-Z0-9_+=.,@\#\$\% ]/i); # Only allow good chars
    return 1 if ($self->{MPF_MaxName} and
		 (length($name) > $self->{MPF_MaxName}));
    $self->debug("it's ok");
    0;
}

#------------------------------

=item exorcise_filename FILENAME

I
If a given filename is evil (see L) we try to
rescue it by performing some basic operations: shortening it,
removing bad characters, etc., and checking each against
evil_filename().

Returns the exorcised filename (which is guaranteed to not
be evil), or undef if it could not be salvaged.

B at the time this method is invoked, the FILENAME has
already been unmime'd into the local character set.
If you're using anything character set other than ASCII, ISO-8859-*,
or UTF-8, the interpretation of the "path" characters might be very
very different, and you will probably need to override this method.
See L for more details.

=cut

sub exorcise_filename {
    my ($self, $fname) = @_;

    ### Isolate to last path element:
    my $last = $fname;

    ### Path separators are / or \
    $last =~ s{^.*[/\\]}{};

    ### Convert semi-evil characters to underscores
    $last =~ s/[\/\\\[\]:]/_/g;
    if ($last and !$self->evil_filename($last)) {
	$self->debug("looks like I can use the last path element");
	return $last;
    }

    ### Break last element into root and extension, and truncate:
    my ($root, $ext) = (($last =~ /^(.*)\.([^\.]+)\Z/)
			? ($1, $2)
			: ($last, ''));
    ### Delete leading and trailing whitespace
    $root =~ s/^\s+//;
    $ext  =~ s/\s+$//;
    $root = substr($root, 0, ($self->{MPF_TrimRoot} || 14));
    $ext  = substr($ext,  0, ($self->{MPF_TrimExt}  ||  3));
    $ext =~ /^\w+$/ or $ext = "dat";
    my $trunc = $root . ($ext ? ".$ext" : '');
    if (!$self->evil_filename($trunc)) {
	$self->debug("looks like I can use the truncated last path element");
	return $trunc;
    }

    ### Remove all bad characters
    $trunc =~ s/([^-A-Z0-9_+=.,@\#\$ ])/sprintf("%%%02X", unpack("C", $1))/ige;
    if (!$self->evil_filename($trunc)) {
	$self->debug("looks like I can use a munged version of the truncated last path element");
	return $trunc;
    }

    ### Hope that works:
    undef;
}

#------------------------------

=item find_unused_path DIR, FILENAME

I
We have decided on an output directory and tentative filename,
but there is a chance that it might already exist.  Keep
adding a numeric suffix "-1", "-2", etc. to the filename
until an unused path is found, and then return that path.

The suffix is actually added before the first "." in the filename
is there is one; for example:

    picture.gif       archive.tar.gz      readme
    picture-1.gif     archive-1.tar.gz    readme-1
    picture-2.gif     archive-2.tar.gz    readme-2
    ...               ...                 ...
    picture-10.gif
    ...

This can be a costly operation, and risky if you don't want files
renamed, so it is in your best interest to minimize situations
where these kinds of collisions occur.  Unfortunately, if
a multipart message gives all of its parts the same recommended
filename, and you are placing them all in the same directory,
this method might be unavoidable.

=cut

sub find_unused_path {
    my ($self, $dir, $fname) = @_;
    my $i = 0;
    while (1) {

	### Create suffixed name (from filename), and see if we can use it:
	my $suffix = ($i ? "-$i" : "");
	my $sname = $fname; $sname =~ s/^(.*?)(\.|\Z)/$1$suffix$2/;
	my $path = File::Spec->catfile($dir, $sname);
	if (! -e $path) {   ### it's good!
	    $i and $self->whine("collision with $fname in $dir: using $path");
	    return $path;
	}
	$self->debug("$path already taken");
    } continue { ++$i; }
}

#------------------------------

=item ignore_filename [YESNO]

I
Return true if we should always ignore recommended filenames in
messages, choosing instead to always generate our own filenames.
With argument, sets this value.

B subclasses of MIME::Parser::Filer which override
output_path() might not honor this setting; note, however, that
the built-in subclasses honor it.

=cut

sub ignore_filename {
    my $self = shift;
    $self->{MPF_IgnoreFilename} = $_[0] if @_;
    $self->{MPF_IgnoreFilename};
}

#------------------------------

=item output_dir HEAD

I
Return the output directory for the given header.
The default method returns ".".

=cut

sub output_dir {
    my ($self, $head) = @_;
    return ".";
}

#------------------------------

=item output_filename HEAD

I
A given recommended filename was either not given, or it was judged
to be evil.  Return a fake name, possibly using information in the
message HEADer.  Note that this is just the filename, not the full path.

Used by L.
If you're using the default C, you probably don't
need to worry about avoiding collisions with existing files;
we take care of that in L.

=cut

sub output_filename {
    my ($self, $head) = @_;

    ### Get the recommended name:
    my $recommended = $head->recommended_filename;

    ### Get content type:
    my ($type, $subtype) = split m{/}, $head->mime_type; $subtype ||= '';

    ### Get recommended extension, being quite conservative:
    my $recommended_ext = (($recommended and ($recommended =~ m{(\.\w+)\Z}))
			   ? $1
			   : undef);

    ### Try and get an extension, honoring a given one first:
    my $ext = ($recommended_ext ||
	       $self->{MPF_Ext}{"$type/$subtype"} ||
	       $self->{MPF_Ext}{"$type/*"} ||
	       $self->{MPF_Ext}{"*/*"} ||
	       ".dat");

    ### Get a prefix:
    ++$GFileNo;
    return ($self->output_prefix . "-" . untainted_pid() . "-$GFileNo$ext");
}

#------------------------------

=item output_prefix [PREFIX]

I
Get the short string that all filenames for extracted body-parts
will begin with (assuming that there is no better "recommended filename").
The default is F<"msg">.

If PREFIX I given, the current output prefix is returned.
If PREFIX I given, the output prefix is set to the new value,
and the previous value is returned.

Used by L.

B subclasses of MIME::Parser::Filer which override
output_path() or output_filename() might not honor this setting;
note, however, that the built-in subclasses honor it.

=cut

sub output_prefix {
    my ($self, $prefix) = @_;
    $self->{MPF_Prefix} = $prefix if (@_ > 1);
    $self->{MPF_Prefix};
}

#------------------------------

=item output_type_ext

I
Return a reference to the hash used by the default
L for mapping from content-types
to extensions when there is no default extension to use.

    $emap = $filer->output_typemap;
    $emap->{'text/plain'} = '.txt';
    $emap->{'text/html'}  = '.html';
    $emap->{'text/*'}     = '.txt';
    $emap->{'*/*'}        = '.dat';

B subclasses of MIME::Parser::Filer which override
output_path() or output_filename() might not consult this hash;
note, however, that the built-in subclasses consult it.

=cut

sub output_type_ext  {
    my $self = shift;
    return $self->{MPF_Ext};
}

#------------------------------

=item output_path HEAD

I
Given a MIME head for a file to be extracted, come up with a good
output pathname for the extracted file.  This is the only method
you need to worry about if you are building a custom filer.

The default implementation does a lot of work; subclass
implementers I should try to just override its components
instead of the whole thing.  It works basically as follows:

    $directory = $self->output_dir($head);

    $filename = $head->recommended_filename();
    if (!$filename or
	 $self->ignore_filename() or
	 $self->evil_filename($filename)) {
 	$filename = $self->output_filename($head);
    }

    return $self->find_unused_path($directory, $filename);

B There are many, many, many ways you might want to control
the naming of files, based on your application.  If you don't like
the behavior of this function, you can easily define your own subclass
of MIME::Parser::Filer and override it there.

B Nickolay Saukh pointed out that, given the subjective nature of
what is "evil", this function really shouldn't I about an evil
filename, but maybe just issue a I message.  I considered that,
but then I thought: if debugging were off, people wouldn't know why
(or even if) a given filename had been ignored.  In mail robots
that depend on externally-provided filenames, this could cause
hard-to-diagnose problems.  So, the message is still a warning.

I
implementation, and for making some good suggestions.  Thanks also to
Achim Bohnet for pointing out that there should be a hookless, OO way of
overriding the output path.>

=cut

sub output_path {
    my ($self, $head) = @_;

    ### Get the output directory:
    my $dir = $self->output_dir($head);

    ### Get the output filename as UTF-8
    my $fname = $head->recommended_filename;

    ### Can we use it:
    if    (!defined($fname)) {
	$self->debug("no filename recommended: synthesizing our own");
	$fname = $self->output_filename($head);
    }
    elsif ($self->ignore_filename) {
	$self->debug("ignoring all external filenames: synthesizing our own");
	$fname = $self->output_filename($head);
    }
    elsif ($self->evil_filename($fname)) {

	### Can we save it by just taking the last element?
	my $ex = $self->exorcise_filename($fname);
	if (defined($ex) and !$self->evil_filename($ex)) {
	    $self->whine("Provided filename '$fname' is regarded as evil, ",
			 "but I was able to exorcise it and get something ",
			 "usable.");
	    $fname = $ex;
	}
	else {
	    $self->whine("Provided filename '$fname' is regarded as evil; ",
			 "I'm ignoring it and supplying my own.");
	    $fname = $self->output_filename($head);
	}
    } else {
            # Untaint $fname... we know it's not evil
            if ($fname =~ /^(.*)$/) {
                    $fname = $1;
            }
    }

    $self->debug("planning to use '$fname'");

    ### Resolve collisions and return final path:
    return $self->find_unused_path($dir, $fname);
}

#------------------------------

=item purge

I
Purge all files/directories created by the last parse.
This method simply goes through the purgeable list in reverse order
(see L) and removes all existing files/directories in it.
You should not need to override this method.

=cut

sub purge {
    my ($self) = @_;
    foreach my $path (reverse @{$self->{MPF_Purgeable}}) {
	(-e $path) or next;   ### must check: might delete DIR before DIR/FILE
	rmtree($path, 0, 1);
	(-e $path) and $self->whine("unable to purge: $path");
    }
    1;
}

#------------------------------

=item purgeable [FILE]

I
Add FILE to the list of "purgeable" files/directories (those which
will be removed if you do a C).
You should not need to override this method.

If FILE is not given, the "purgeable" list is returned.
This may be used for more-sophisticated purging.

As a special case, invoking this method with a FILE that is an
arrayref will replace the purgeable list with a copy of the
array's contents, so [] may be used to clear the list.

Note that the "purgeable" list is cleared when a parser begins a
new parse; therefore, if you want to use purge() to do cleanup,
you I do so I starting a new parse!

=cut

sub purgeable {
    my ($self, $path) = @_;
    return @{$self->{MPF_Purgeable}} if (@_ == 1);

    if (ref($path)) { $self->{MPF_Purgeable} = [ @$path ]; }
    else            { push @{$self->{MPF_Purgeable}}, $path; }
    1;
}

=back

=cut


#------------------------------------------------------------
#------------------------------------------------------------

=head2 MIME::Parser::FileInto

This concrete subclass of MIME::Parser::Filer supports filing
into a given directory.

=over 4

=cut

package MIME::Parser::FileInto;

use strict;
use vars qw(@ISA);
@ISA = qw(MIME::Parser::Filer);

#------------------------------

=item init DIRECTORY

I
Set the directory where all files will go.

=cut

sub init {
    my ($self, $dir) = @_;
    $self->{MPFI_Dir} = $self->cleanup_dir($dir);
}

#------------------------------
#
# output_dir HEAD
#
# I
# Return the output directory where the files go.
#
sub output_dir {
    shift->{MPFI_Dir};
}

=back

=cut




#------------------------------------------------------------
#------------------------------------------------------------

=head2 MIME::Parser::FileUnder

This concrete subclass of MIME::Parser::Filer supports filing under
a given directory, using one subdirectory per message, but with
all message parts in the same directory.

=over 4

=cut

package MIME::Parser::FileUnder;

use strict;
use vars qw(@ISA);
@ISA = qw(MIME::Parser::Filer);

#------------------------------

=item init BASEDIR, OPTSHASH...

I
Set the base directory which will contain the message directories.
If used, then each parse of begins by creating a new subdirectory
of BASEDIR where the actual parts of the message are placed.
OPTSHASH can contain the following:

=over 4

=item DirName

Explicitly set the name of the subdirectory which is created.
The default is to use the time, process id, and a sequence number,
but you might want a predictable directory.

=item Purge

Automatically purge the contents of the directory (including all
subdirectories) before each parse.  This is really only needed if
using an explicit DirName, and is provided as a convenience only.
Currently we use the 1-arg form of File::Path::rmtree; you should
familiarize yourself with the caveats therein.

=back

The output_dir() will return the path to this message-specific directory
until the next parse is begun, so you can do this:

    use File::Path;

    $parser->output_under("/tmp");
    $ent = eval { $parser->parse_open($msg); };   ### parse
    if (!$ent) {	 ### parse failed
	rmtree($parser->output_dir);
	die "parse failed: $@";
    }
    else {               ### parse succeeded
	...do stuff...
    }

=cut

sub init {
    my ($self, $basedir, %opts) = @_;

    $self->{MPFU_Base}    = $self->cleanup_dir($basedir);
    $self->{MPFU_DirName} = $opts{DirName};
    $self->{MPFU_Purge}   = $opts{Purge};
}

#------------------------------
#
# init_parse
#
# I
# Prepare to start parsing a new message.
#
sub init_parse {
    my $self = shift;

    ### Invoke inherited method first!
    $self->SUPER::init_parse;

    ### Determine the subdirectory of their base to use:
    my $subdir = (defined($self->{MPFU_DirName})
		  ?       $self->{MPFU_DirName}
		  :       ("msg-" . scalar(time) . "-" .
                           MIME::Parser::Filer::untainted_pid() . "-" . $GSubdirNo++));
    $self->debug("subdir = $subdir");

    ### Determine full path to the per-message output directory:
    $self->{MPFU_Dir} = File::Spec->catfile($self->{MPFU_Base}, $subdir);

    ### Remove and re-create the per-message output directory:
    rmtree $self->output_dir if $self->{MPFU_Purge};
    (-d $self->output_dir) or
	mkdir $self->output_dir, 0700 or
	    die "mkdir ".$self->output_dir.": $!\n";

    ### Add the per-message output directory to the puregables:
    $self->purgeable($self->output_dir);
    1;
}

#------------------------------
#
# output_dir HEAD
#
# I
# Return the output directory that we used for the last parse.
#
sub output_dir {
    shift->{MPFU_Dir};
}

=back

=cut

1;
__END__