lance icon indicating copy to clipboard operation
lance copied to clipboard

feat: allow general compression for string and binary types

Open niyue opened this issue 1 year ago • 2 comments

This is an attempt to allow users to apply general compression such as zstd to compress string and binary types when they explicitly specify the relevant field metadata

This PR tries to address https://github.com/lancedb/lance/issues/2876

niyue avatar Sep 13 '24 13:09 niyue

Right after submitting the PR, I realized that the relevant code in encoder.rs was significantly updated yesterday, making the PR largely irrelevant 🤣. I'll need some time to review the new code and figure out how the PR should be revised.

niyue avatar Sep 13 '24 13:09 niyue

Right after submitting the PR, I realized that the relevant code in encoder.rs was significantly updated yesterday, making the PR largely irrelevant 🤣. I'll need some time to review the new code and figure out how the PR should be revised.

Sorry about that. Yes, some things have been shaking up a bit recently and I do think I disconnected zstd as part of that. My primary goal is to move scheduling out of the lower level "compressive encoders". I do want to reconnect zstd (and friends) as part of 2.1. I'll also look at this a bit today.

westonpace avatar Sep 13 '24 13:09 westonpace

@niyue I've made an attempt in https://github.com/lancedb/lance/pull/2878

westonpace avatar Sep 13 '24 16:09 westonpace

Close this PR in favor of https://github.com/lancedb/lance/pull/2878

niyue avatar Sep 14 '24 00:09 niyue