Jonas Haag comments

Results 800 comments of


                                            Jonas Haag

`sink_parquet(...)` much slower than `.collect().write_parquet(...)`

@hugopendlebury it is macOS Instruments (comes with Xcode). It is pretty unknown outside of iOS app development but it is actually a fantastic performance monitoring tool. Probably the best performance...

`Series(dtype=)` sometimes ignored or nulls data

I've had a look into this. My idea is to convert Python types to Polars dtypes in `Series.__init__`, pass those to the construction code (eg. `sequence_to_pyseries`), and let the Rust...

`Series(dtype=)` sometimes ignored or nulls data

@stinodego any hints how to proceed here?

`Series(dtype=)` sometimes ignored or nulls data

Another example where it's ignored, even with `strict=True`: ```py In [3]: pl.Series(["a"], dtype=pl.UInt64, strict=True) Out[2]: shape: (1,) Series: '' [str] [ "a" ] ```

`np.asarray(series, dtype="object")` should return null values as None

@ritchie46 @stinodego @alexander-beedie if you agree with my opinion I'll take a stab at this

`np.asarray(series, dtype="object")` should return null values as None

Hmm, this is incompatible with `np.asarray(pyarrow.array(...), dtype="object")`. Maybe this is something that should be discussed with PyArrow and/or the array interchange protocol first.

`np.asarray(series, dtype="object")` should return null values as None

I think we should change the labels of this ticket from bug to something else.

Apple Silicon build not supported.

I don't know how to reproduce since it works on my machine ``` ❯ python3 -m venv env ❯ . env/bin/activate.fish ❯ pip install bjoern Collecting bjoern Using cached bjoern-3.2.2.tar.gz...

Apple Silicon build not supported.

You need to install libev, see install instructions

`scan_database` feature

It would be really handy to have streaming support for reading query results. My use case: ```py pl.scan_database(...).stuff().sink_parquet("stuff.parquet") ```