A Go API for zdelta
This is a Go (golang) interface to the zdelta delta-compression library.
zdelta creates and applies binary deltas (aka diffs) between arbitrary strings of bytes. Delta encoding is very useful as a compact representation of changes in files or data records. For example, all software version control systems use chains of deltas to store the history of a file over time, and most software update systems contain packages of deltas (patches) instead of entire files.
Using It
This package can be directly imported into a Go source file via:
import "github.com/snej/zdelta-go"The zdelta source code (in C) is contained in this package, so it's standalone and you don't need to install any shared libraries into your system. (There's no official up-to-date zdelta source repository, so I copied the source from my unofficial (but up-to-date, as far as I know) one.)
Here's the Godoc documentation. And here's an example of basic usage: Let's say you have two byte arrays, vers1 and vers2, containing two revisions of a file. You want to send the second version to someone who only has the first. So you generate a delta:
delta, err := zdelta.CreateDelta(vers1, vers2)The delta is just a byte array you can send however you like. (There are alternate APIs that directly write the delta to an output stream.) In general it will be much smaller than vers2, especially if the two versions are almost the same; in the worst case it could be as much as 12 bytes larger (but no more.)
On the receiving end, you just gather up vers1 and the delta you received, and:
vers2, err := zdelta.ApplyDelta(vers1, delta)(Again, there's an alternate API that allows you to write the output to a stream.)
Performance
The runtime speed should be nearly identical to the base zdelta library. Since zdelta is based on zlib, the speed of generating a delta is about the same as that of zipping (deflating) the concatenation of the two versions. I'm not so sure about the speed of applying a delta; it should be faster than unzipping the target would be, but somewhat slower than unzipping an archive the size of the delta.
Status
As of now (22 Feb 2015) I consider this pre-alpha. The API isn't nailed down, it hasn't been seriously used yet, and this is admittedly my first attempt at using CGo. On the other hand, the core functionality comes from a very stable C library that's been in use for over a decade.
If you find any problems or have suggestions, please file an issue here on Github. Thanks!
—Jens Alfke