Closed WoMayr closed 5 months ago
Table API will be deprecated soon. Are you able to use class serialisation?
Oh.. good to know. Unfortunately class serialization is not an option for my use case.
Fair enough. I'm trying to rewrite row API to use the same engine as class serialization as it follows parquet specification 100% but it's a long shot. It is a priority for me though.
Is there a workaround? Files generated using this method also fail to read using s3 object select (this is crucial to our ETL pipelines). Class based serialization is not an option - we have dynamic schemas that get determined at run time.
Were now looking at serializing this to JSON - not my favorite approach.
Sorry as of now row API development is frozen unless you contribute with a PR. The best option would be to resort to low level API which is as flexible as it gets, but needs special handling for lists . There are plenty of examples though.
Closing due to inactivity and deprecation of row api soon.
Library Version
4.16.4
OS
Windows 10
OS Architecture
64 bit
How to reproduce?
I also tried opening the file with DBeaver using the DuckDB driver and get the following error: "SQL Error: Invalid Input Error: Mismatch in parquet read for column 6, expected 1000 rows, got 665"
Failing test