Open sylvinus opened 8 years ago
Decompressing the WARC files from Common Crawl is a relatively slow step in the indexing process. It would be great to see how much improvement CloudFlare's version of zlib can bring.
Some benchmarks here: http://www.snellman.net/blog/archive/2015-06-05-updated-zlib-benchmarks/
We should keep the ability to fallback on the regular implementation.
To do this we may have to fork commoncrawl/gzipstream, which is the place where zlib is imported: https://github.com/commoncrawl/gzipstream/blob/master/gzipstream/gzipstreamfile.py#L2
zlib
If the improvement is significant we should use it and add the build commands to our Dockerfile.
Decompressing the WARC files from Common Crawl is a relatively slow step in the indexing process. It would be great to see how much improvement CloudFlare's version of zlib can bring.
Some benchmarks here: http://www.snellman.net/blog/archive/2015-06-05-updated-zlib-benchmarks/
We should keep the ability to fallback on the regular implementation.
To do this we may have to fork commoncrawl/gzipstream, which is the place where
zlib
is imported: https://github.com/commoncrawl/gzipstream/blob/master/gzipstream/gzipstreamfile.py#L2If the improvement is significant we should use it and add the build commands to our Dockerfile.