Markus Klein

Hello @andreypanchenko , Sorry you encountered a bug. You could help me by logging with verbose output (add -vvv to the command line) and sharing the results with me. Thanks

Thank you. This is very enlightening. I think the most relevant clue is here: > 2024-10-16T08:27:57+00:00 - INFO Fetched batch 32769 with 15 rows. 32769 is 2^15 + 1, or...

sorry, my bad `column-length-limit` is indeed in characters rather than memory, forget the unit.

The newest `odbc2parquet 6.3.1` also ships an upstream fix of parquet, which would replace the panic with a proper error. Still the column-length-limit is needed. Also want to keep the...

`odbc2parquet 7.0.0` is released and sets `--column-length-limit 4096` by default.

Hello @allurefx , yeah, this value is simply out of range and can not be represented by the associated arrow type. I can improve the error message though. If a...

Hello @allurefx , I would still like to see the logs, to learn about the involved data types. We might get this to work, but I currently do not want...

Thanks for the log and the additional information. It seems to me that before the the iterator which yields `None` you should have gotten one `Err` with this error: https://github.com/pacman82/arrow-odbc/blob/08c5de1b5e5f1b3c4e0f47d1fe9d0cc7d77c7c20/src/reader/map_odbc_to_arrow.rs#L202...

Here the entire text of the error message: > Timestamp is not representable in arrow: {value} > Timestamps with nanoseconds precision are represented using a signed 64 Bit integer. This...

@allurefx You may be happy to learn that `arrow-odbc 12.0.0` has been released. It features a new method for the `OdbcReaderBuilder` called `value_errors_as_null`. Setting the attribute to `true` should yield...