After merging #8303 we got a regression on zstd-encoded arrays: #8310 (comment).
There are two issues with that:
- You can't push zstd-encoded byte_length as it decompresses.
- You need to reuse the cache, as branching over ZSTD doesn't solve the issue you recompute values over and over again for each Dict request.
However, why do we need SharedArray and caching? If Dict itself was cached, we'd compute the (canonicalized or not) values once and then hold on to this reference till the end of this query.
As we have execution context now, we can cache the dict object on it.
After merging #8303 we got a regression on zstd-encoded arrays: #8310 (comment).
There are two issues with that:
However, why do we need SharedArray and caching? If Dict itself was cached, we'd compute the (canonicalized or not) values once and then hold on to this reference till the end of this query.
As we have execution context now, we can cache the dict object on it.