What is FASTDIFF?FASTDIFF is a binary diffing engine, intended for use both as a stand-alone utility, and as a library to be integrated into other products such as: Communication devices or software packages that wish to send "differential data" over a slow communication link Storage devices or software packages that wish store or transmit diffs between versions of data or files Version control systems Applications with internal versioning of documents, such as CAD or Document Management systemsWhat is Binary Diffing?Binary Diffing is an operation that takes two files (or arrays of data), and extracts the difference (also known as a diff or a delta) between them. DIFF = DIFF (SOURCE, TARGET) = TARGET - SOURCEWhere TARGET and SOURCE are files that we believe to be related (i.e. some portion of their data is identical), and DIFF is the resulting difference between the file.The DIFF may then be stored, or transmitted over a communication line, and eventually used to recreate the TARGET file, using the reverse operation, known as PATCH:TARGET = PATCH (SOURCE, DIFF) = SOURCE + DIFFBecause the DIFF is typically 20x to 100x smaller than the TARGET file, storing or transmitting the DIFF only, rather than the whole TARGET file, is much more efficient.FASTDIFF is a binary diff - meaning that it works on arbitrary data types. This is in contrast with other diff algorithms, such as Unix's diff command, that work on text files only.What makes FASTDIFF unique?FASTDIFF's uniqueness is its high performance and low memory consumption. FASTDIFF is therefore capable of operating in a real-time environment, such as in a communication device. FASTDIFF requires a fixed amount of working memory, regardless of file size (roughly 1.25MB). See benchmark to a get a sense of FASTDIFF's performance.Does FASTDIFF work on any file type?Generally - yes. FASTDIFF has no problem working on files such as files with embedded graphics, word processing files, CAD files, database files, email archives etc.However, there are specific types of files on which FASTDIFF will work poorly. The reason for this is that FASTDIFF extracts similarities between the files - on a byte level. Here are two examples in which even though the files contain similar information, the representation of that information as a string of bytes is very different, and therefore the FASTDIFF algorithm will perform poorly:(1) Image files with global modifications. If, for example, you change the contrast of an image file using an image editor, the result is that almost every byte of the file is changed.(2) Compressed files - consider two zip archives of two versions of a file - even though the underlying files in the compressed archives may be very similar, the compression algorithm may result in compressed files that differ in almost every byte.Does FASTDIFF detect shifts in a file? For example, when a new paragraph is inserted into a word processing document - the data following the new paragraph shifts in position in the file. Does FASTDIFF still detect the similarities of this portion of data, even though it has shifted?Yes.What interface do you provide to FASTDIFF?FASTDIFF has two types of interfaces:Command line interface:fastdiff <source file> <target file> <diff file>fastpatch <source file> <diff file> <target file>'C' library interface: Take a look at the API.Do you provide the source code of the FASTDIFF library?Both source code and binary-only licensing options are available.What platforms are currently supported? How portable is the FASTDIFF code?The binary version of the library runs on both Windows and Linux platforms, on Intel CPUs or compatible.The FASTDIFF code itself was designed with portability in mind: The code is pure ANSI 'C' - compiles cleanly with GCC and MS VC Should run with almost no modifications on other processors, including big endian processors and 64 bit processors. The resulting DIFF file is identical regardless of processor type, endianity or bus width. Can you help us adapt the FASTDIFF library to our specific needs?In some cases we are able to provide consulting services to adapt the FASTDIFF library to your specific needs, including optimization for certain file types, and porting to other platforms.
