Search code, repositories, users, issues, pull requests...
Imagine if you could randomly access any byte in a gzipped content, with a minimal seeking and skipping cost using indexes encoded in gzip's metadata...
Ragzip stands for randomly-accessible gzip. The ragzip format is a compatible gzip (.gz) file. It can be produced over a strictly appending output stream like gzip, or with random access to the destination file to write.
However, the ragzip-compatible consumers can perform random accesses on a seekable storage api (ex: file system api or http bytes ranges requests) at strategic positions in order to benefit from the ragzip-specific metadata.
This ragzip metadata is stored in the gzip header of "empty" gzip members. The word "empty" means no actual compressed bytes. This means, such empty gzip members do not interfere with the decompressed data.
In particular, the readers would seek the 64 bytes fixed length footer, the indexes tree and the pages. Most importantly, the footer is documenting how to read the rest of the ragzip metadata and content, like a header normally does in many other formats.
