Gerd Zellweger
Gerd Zellweger
the clock has been ticktocking too long for this PR I'm going to close it
I wouldn't expect it to add a lot of overhead if you measure during constructing of the badge then cache (you already have to memcpy the thing anyways)? size-of for...
Ultimately this happens because JSON doesn't support non-strings as field names -- and the arrow-json not having a setting to work around this. https://github.com/apache/arrow-rs/blob/a01886db20982155fb38f101181ae14f31ed9256/arrow-json/src/writer/encoder.rs#L726 and https://github.com/apache/arrow-rs/issues/6525 Might be easiest to...
most likely happens because of a JSON conversion somewhere in there (keys in JSON have to be string). Note that if you're using ad-hoc queries this should be fixable by...
I added an arrow format recently that can represent this. Eg if you pass --format arrow to fda you will get the correct map. We should probably use this for...
might be in our serialization code then let me check
Noticed that by looking at the log here https://github.com/feldera/feldera/actions/runs/17777972992/job/50530875984#step:4:854
arrow has a decimal256 type. maybe it would be better to use this instead of dec: https://docs.rs/arrow/latest/arrow/datatypes/struct.Decimal256Type.html
It implements this trait https://docs.rs/arrow/latest/arrow/datatypes/trait.ArrowNativeTypeOp.html so it appears so
can you check if this is now supported by using the arrow-ipc format for ad-hoc queries (it probably should)