Charles Pritchard
Charles Pritchard
The allocation of splits to available nodes may benefit from testing data locality; mainly, repeated searches would result in the same shards being run on the same nodes (or group...
If it's feasible, an option for echo_location and echo_location_async to ignore the output of the subrequests would be helpful for running some subrequests that handle various cleanup. memcache "get" followed...
There are mixed licenses in the code base -- the subpackages could use clarification as to the licenses, e.g. the Scott Bishel copyright as listed would be all rights reserved
The npm thrift project is not packaged for library use, and it requires about 30 megs of space. It's fine for devDependencies but should not be used for production (distribution)...
This grabs from @livechat/ui-kit, but that package is unlicensed and largely difficult to find source.
I hooked in for a short time to persist.purge -- we need to have a "clear chat history" option for various reasons for users. I was not clear, from our...
I don't know if you want to bring this in yet, but the current RocksDB has a compression stub now for ZSTD as kZSTDNotFinalCompression. zstd is quite a bit faster...
From the blog: > Allow preset compression dictionary for improved compression of block-based tables. This is supported for zlib, zstd, and lz4. The compression dictionary’s size is configurable via CompressionOptions::max_dict_bytes.
I'm going to take a peek at the performance on this -- if it's not worth it, I'll just close out the issue. There may be overhead in the iterkeys/iteritems/itervalues...
Currently the bytes_to_slice does not support bytearray; bytearray is just a standard/compatible pointer. ``` batch.put(key, bytearray("data")) ```