https://delta.io logo
c

Cole MacKenzie

05/16/2023, 7:08 PM
Found an issue where the column statistics fail to parse during query time for
timestamp
fields. https://github.com/delta-io/delta-rs/issues/1372
I am seeing this in
0.10
. Haven't tried
0.11
yet but I think this was a change in arrow
r

rtyler

05/16/2023, 7:43 PM
I wonder if this might be the same issue as https://github.com/delta-io/delta-rs/issues/1286 showing up in a different spot
c

Cole MacKenzie

05/16/2023, 8:40 PM
It might be. I spot checked a few
.json
files in the
_delta_log
and they look to be proper RFC3339
Looking at the
.checkpoint.parquet
files I see them as
Copy code
"Tag": "name=EdgeEndTimestamp, type=INT64, convertedtype=TIMESTAMP_MICROS, repetitiontype=OPTIONAL"
r

rtyler

05/16/2023, 8:54 PM
parquet things timestamps should be one thing, Delta likes to pretend timestamps are INT96's which is super deprecated
c

Cole MacKenzie

05/16/2023, 8:55 PM
Yeah, I cry every time I see
INT96
anywhere
🤬 1
Is that just missing
TimestampMicros
block?
🎉
I hope
r

rtyler

05/16/2023, 10:16 PM
does this work for your use-case with a real delta table already?
c

Cole MacKenzie

05/16/2023, 10:21 PM
Copying the table over to try now. If the encoding in parquet is
type=INT64, convertedtype=TIMESTAMP_MICROS
, then the match statement should have that type, right?
These parquet files were generated by delta-rs v0.10
r

rtyler

05/16/2023, 10:30 PM
it might? I'm not sure tbh