Patrick Hoefler

Results 345 comments of Patrick Hoefler

CoW: Use weakref callbacks to track dead references

How does this correspond to the other timings in the issue?

ENH: Add dtype_backend parameter to to_datetime

Yep this sounds good. Same as for to_numeric basically. Pinging @MarcoGorelli

BUG: Memory leak after pd.read_csv() with default parameters

Hi, thanks for your report. Contributions are welcome

PERF: Inefficient data representation when building dataframe from 2D NumPy array

I think we might have a chance to fix this with Copy on write. When using CoW we have to/should copy a given array to avoid that changing the array...

PERF: Inefficient data representation when building dataframe from 2D NumPy array

Most of them are not using 2D numpy arrays as far as I know

apache-airflow-providers-amazon added xmlsec as new dependency and pinned to a version that doesn't have wheels for new python versions

Fixing the root cause sounds good, but currently it is **not possible** to install the amazon provider with any configuration that has Python >=3.11 or a non Windows machine on...

Nunique dropping NaN values

Hi, thanks for your report. dropna and observed are not passed through for nunique, which means the defaults from pandas are used. Are you interested in contributing a fix?

Improper conversions of object dtype to string[pyarrow]

https://pandas.pydata.org/docs/whatsnew/v2.1.0.html#avoid-numpy-object-dtype-for-strings-by-default This is the best we can do, it converts the data through looking at them. It would cause meta and the actual data get out of sync, but that's...

Improper conversions of object dtype to string[pyarrow]

So I looked a little into the types mapper, converting dates and decimals to numpy dtypes is not supported out of the box. We could theoretically cast them ourselves if...

Use 3.12 as the canonical environment

needs a rebase