Python and SQLAlchemy gotcha: ORM's RowMapping walks like a dict and quacks like a dict, yet isn't one. Spent hours debugging why a Polars DataFrame test failed. The culprit? Duck typing: a sequence of RowMapping differs from a dictionary sequence when evaluated in Polars, causing the bug.
Comments
And your point is sharp: Python might ike duck typing, Polars does not lol…
And yes, not blaming, just root-cause analyzing XD
BTW in terms of memory usage, Polars streaming is really interesting. We've been able to process huge Parquet files in standard VMs, combining LazyFrame, a generator as data input, and sink_parquet to write the output. Neat.