NAME Vcdiff - diff and patch for binary data SYNOPSIS In order to use this module you must install one or more backend modules (see below) use Vcdiff; my $delta = Vcdiff::diff($source, $target); ## ... send the $delta string to someone who has $source ... my $target2 = Vcdiff::patch($source, $delta); ## $target2 is the same as $target DESCRIPTION Given source data and target data, the "Vcdiff::diff" function computes a "delta" that encodes the information needed to turn source into target. Anyone who has source and delta can compute target with the "Vcdiff::patch" function. If the source and target inputs are related then delta can be very small relative to target, meaning it may be more efficient to send the delta string instead of the whole target. Even though source and target don't necessarily have to be binary data (regular data is fine too), the delta will always contain binary data including NUL bytes so if your transport protocols don't support this you will have to encode or escape it in some way (ie Base64). Compressing the delta before you do this might be worthwhile depending on the size of your changes and the entropy of your data. The delta format is described by , "The VCDIFF Generic Differencing and Compression Data Format". STREAMING API The streaming API is sometimes more convenient than the in-memory API. It can also be more efficient since it uses less memory. Also, you can start processing output before Vcdiff has finished. Sometimes you have to use the streaming API in order to handle files that are too large to fit into your virtual address space (though note that some backends have size limitations apart from this). In order to send output to a stream, a file handle should be passed in as the 3rd argument to "diff" or "patch": Vcdiff::diff("hello", "hello world", \*STDOUT); In order to fully take advantage of streaming, the source and target parameters of "diff" and the source and delta parameters of "patch" can be file handles instead of strings: open(my $source_fh, '<', 'source.dat') || die $!; open(my $target_fh, '<', 'target.dat') || die $!; open(my $delta_fh, '>', 'delta.dat') || die $!; Vcdiff::diff($source_fh, $target_fh, $delta_fh); Note that in all current backends the source file handle must be backed by an lseek(2)able and/or mmap(2)able file descriptor (in other words, a real file, not a pipe or socket). Vcdiff will throw an exception if the source file handle is unsuitable. MEMORY MAPPED INPUTS If the source or target/delta data is in a file, an alternative to the streaming API is to map the files into memory with mmap(2) and then pass the mappings in to "diff"/"patch" as strings. Doing so is more efficient than the streaming API for large files because fewer system calls are made and a kernel-space to user-space copy is avoided. As mentioned above, files that are too large to fit in your virtual address must be diffed with the streaming API (this is only an issue when diffing multi-gigabyte files on 32 bit systems). Here is an example using Sys::Mmap: use Sys::Mmap; open(my $source_fh, '<', 'source.dat') || die $!; open(my $target_fh, '<', 'target.dat') || die $!; open(my $delta_fh, '>', 'delta.dat') || die $!; my ($source_str, $target_str); mmap($source_str, 0, PROT_READ, MAP_SHARED, $source_fh) || die $!; mmap($target_str, 0, PROT_READ, MAP_SHARED, $target_fh) || die $!; Vcdiff::diff($source_str, $target_str, $delta_fh); munmap($source_str); munmap($target_str); BACKENDS Vcdiff doesn't itself implement delta compression. Instead, it provides a consistent interface to various open-source VCDIFF (RFC 3284) implementations. The implementation libraries it interfaces to are called "backends". In other words, Vcdiff aims to be "the DBI" of VCDIFF implementations. The currently supported backends are described below. See the POD documentation in the backend module distributions for more details on the pros and cons of each backend. XDELTA3 BACKEND The Vcdiff::Xdelta3 backend module bundles Joshua MacDonald's library. OPEN-VCDIFF BACKEND The Vcdiff::OpenVcdiff backend module depends on Alien::OpenVcdiff which configures, builds, and installs Google's library. FUTURE BACKENDS Another possible candidate would be Kiem-Phong Vo's library which contains a vcdiff implementation. A really cool project would be a pure-perl VCDIFF implementation that could be used in environments that are unable to compile XS modules. CHOOSING A BACKEND In order to choose which backend to use, Vcdiff will first check to see if the $Vcdiff::backend variable is populated. If so, it will attempt to load that backend. This variable can be used to force a particular backend: { local $Vcdiff::backend = 'Vcdiff::OpenVcdiff'; $delta = Vcdiff::diff($source, $target); } Otherwise, Vcdiff will check to see if any backends have been loaded already. If so, it will choose the first one it finds: use Vcdiff::Xdelta3; $delta = Vcdiff::diff($source, $target); If it doesn't find any loaded backends, it will try to load them in the following order: Xdelta3, OpenVcdiff. Finally, if none of these backends can be loaded, an exception is thrown. SEE ALSO , "The VCDIFF Generic Differencing and Compression Data Format" AUTHOR Doug Hoyte, "" COPYRIGHT & LICENSE Copyright 2013 Doug Hoyte. This module is licensed under the same terms as perl itself.