Author image Kevin Ryde
and 1 contributors


Filter::gunzip - gunzip Perl source code for execution


 perl -MFilter::gunzip

 # or in a script
 use Filter::gunzip;
 ... # inline gzipped source code bytes


This filter uncompresses gzipped Perl source code at run-time. It can be used from the command-line to run a .pl.gz file,

    perl -MFilter::gunzip

Or in a self-uncompressing executable beginning with a use Filter::gunzip then gzip bytes immediately following that line,

    use Filter::gunzip;
    ... raw gzip bytes here

The filter is implemented in one of two ways.

  • If there are no other filters and PerlIO is available (now usual) then push a PerlIO::gzip layer.

  • Otherwise add a block-oriented source filter per perlfilter.

In both cases, the compressed code executed can apply further source filters in usual ways.

DATA Handle

The __DATA__ token (see "Special Literals" in perldata) and DATA handle can be used in the compressed source, but only some of the time.

For the PerlIO::gzip case, the DATA handle is simply the input, including the :gzip uncompressing layer, positioned just after the __DATA__ token. It can be read in the usual way. However PerlIO::gzip as of its version 0.19 cannot dup() or seek(), which limits what can be done with the DATA handle. In particular for example SelfLoader requires seek() and so doesn't work on compressed source. (Duping and seeking in PerlIO::gzip are probably both feasible, though seeking backward could be slow.)

For the perlfilter case, DATA doesn't work properly. Perl stops reading from the source filters at the __DATA__ token, because that's where the source ends. But a block oriented filter like Filter::gunzip may read ahead in the input file which means the position of the DATA handle is unpredictable, especially if there's more than one block-oriented filter stacked up.


If Filter::gunzip sees a :crlf layer on top of the input then it pushes :gzip underneath that, since the CRLF is almost certainly meant to apply to the text, not to the raw gzip bytes. In particular this should let it work with the global PERLIO=crlf (see "PERLIO" in perlrun) suggested by README.cygwin, and which you could conceivably use elsewhere too.

The Perl tokenizer has some of its own CRLF understanding (unless a build-time strictness option is used) so it can normally read source code in either CRLF or binary (and it translates any CRLF in literal strings or here documents to newline). This allows the default read mode to be chosen according to what you might need for reading text files, including DATA parts of source files.

Read Errors

The gzip format has a CRC checksum at the end of the data. This might catch subtle corruption in the compressed bytes, but as of Perl 5.10 the parser usually doesn't report a read error from the source and in any case the code is compiled and BEGIN blocks are executed immediately, before the CRC is reached, so corruption will likely provoke a syntax error or similar first.

Only the gzip format (RFC 1952) is supported. Zlib format (RFC 1950) differs only in the header, but PerlIO::gzip (version 0.18) doesn't allow it. The actual gunzip program can handle some other formats too, like Unix .Z compress, but those formats are probably best left to other modules.


Filter::exec and the zcat program can do the same thing, either from the command line or self-expanding,

    perl -MFilter::exec=zcat

Because Filter::exec is a block-oriented filter (as of its version 1.37) a compressed DATA section within the script doesn't work, the same as it doesn't in the filter method here.

In the past it was possible to apply PerlIO::gzip to a script with the open pragma and a require of the script filename, though circa Perl 5.30 this doesn't seem to work any more. It was something like the following from the command line. Since the open pragma is lexical, it doesn't affect other later loads or opens.

    perl -e '{use open IN=>":gzip";require shift}' \
            ./ arg1 arg2

It doesn't work to set a PERLIO environment variable for a global :gzip layer, like PERLIO=':gzip(autopop)', because such default layers are restricted to Perl builtin layers (see "PERLIO" in perlrun), and PerlIO::gzip is not a builtin.

Bzip2, lz, or other compression formats could be handled by a very similar filter module. Bz2 has the disadvantage of its decompressor using at least 2.5 Mbytes of memory, so there'd have to be a big disk saving before it was worth that much memory at runtime. Filter::exec above can be used for other formats since the various compression tools normally offer something similar to zcat, such as bzcat or lzcat, or equivalent command line options as for example in lz4.


Gzip format has an end of data indication (cf the CRC described above). Reading stops at that point, and it should be end of file too. Behaviour is unspecified if there's anything after. The gunzip program will read multiple gzips which have been concatenated together, and uncompresses them as a single stream. Maybe similar would be helpful, or maybe something cute like returning to plain text after the gzip. Suspect PerlIO::gzip treats end of gzip as end of file, and so ignores anything after. For now think either concats or more text should be unusual and probably better handled other ways.


PerlIO::gzip, PerlIO, Filter::Util::Call, Filter::exec, gzip(1), zcat(1), open

The author's compile-command-default.el can setup Emacs to run a visited .pl.gz by either Filter::gunzip or other ways, according to what's available.



Filter-gunzip is Copyright 2010, 2011, 2013, 2014, 2019 Kevin Ryde

Filter-gunzip 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.

Filter-gunzip 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 Filter-gunzip. If not, see <>.