File List
Not logged in

Files of check-in [c7456d90e9] in directory bdelta   [history]

Author: John Whitney (
License: Mozilla Public License, v. 2.0.


BDelta is a sophisticated sequence matching library bundled with a delta
creator and patch tool. This code is being used in production systems. Recent
releases should be stable, fast, and accurate.

Limitations / Warranty

Note that there is currently a 4GB file size limitation.

This software does not come with any guarantees. However, if you have any
problems, please send me an e-mail and I'll likely be able to help. I would
especially like to know about any bugs that you find.

Library API C / C++

BDelta can be used in a wide range of applications. It is easy for the library
user to optimize its behavior to obtain better performance and results.

The only header you need to include to use the library is "bdelta.h".

The key thing to know about BDelta's algorithm is that it is designed to
pass over the data multiple times, finding ever-smaller matches. The API
is designed to give a great deal of control over each pass. This functionality
is not well-documented, so contact me if you need help. I hope to add more
comments to the code in future releases.

Other Bindings

A simple interface for accessing the API with Python is included. This
interface is kept up-to-date since I do a lot of work in Python these days. To
istall it, just type "python install" in the root folder. You'll need 
Cython to build/install, but Cython is not required after installation.

See test/ for a usage example.

Delta File Format

The included delta creator / patch tool uses this simple binary
format. All integers are written in protobuf format, either unsigned
(upb) or zig-zag encoded signed (ipb).

(All numbers are stored in little-endian format)
char[3] magic "BDT"
upb version (3, but will increment if this binary format changes)
upb file 1 size
upb file 2 size
upb number of matches
for each match {
    <upb n><string size <n>> insert string into file
    <ipb seek>               seek in file 1
    <upb m>		     copy m bytes from file 1