Georgi Peev

@diogobaeder Please refer to my comment on #189.

You need to wrap your column types in `clickhouse_sqlalchemy.types.Nullable` like so: ``` relevance = Column(Nullable(Integer), nullable=True) ``` The `nullable` parameter is confusing, but the docstring of the `Column` constructor explicitly...

In my proposal I was trying to have a minimal code impact. I will, of course, defer to you about what change would make most sense. I get the point...

I posit that streaming would also benefit from grouping small blocks into larger batches, and it seems sensible to determine the appropriate batch size empirically. Letting the client choose the...

Thanks! In the two test cases I ran, the performance of 0.7.2 was indistinguishable from that of 0.6.16. There were some new type-related woes I had to take care of...

Some string columns that used to come back as dtype `object`/`str` now come as `pandas.StringDtype()`. I didn't change the pandas version I was using, it was 2.2.1 in both cases.

I'm sorry, but I don't think I would be able to generate a sanitized set that reproduces the problem. Regarding the schema, the table is partitioned by day, has 6...

I see. Do you plan on modifying the behavior of `query_df` to return consistent dtypes across different versions of pandas, regardless of whether those dtypes are the old ones or...