update io documentation

[usneil]

The HDFS dropna=True parameter does not drop the NaN values. Perhaps this is a bug to the core Pandas, or really a typo in the documentation

• closes REGR: pd.to_hdf(dropna=True) not dropping all nan rows #35719
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff

[rhshadrach]

The way you're changing the documentation, setting dropna=True keeps NA values. That isn't correct, I believe this is a bug, not an issue with the docs.

[simonjayhawkins]

Thanks @grrlic for the PR. see @rhshadrach comment

[usneil]

Ah, glad this bug is found soon, 'cause I myself kinda doubt that the errata comes from the docs, but apparently it's not the case. Thanks to @XiaozhanYang observations.

[usneil]

@simonjayhawkins Thanks for the quick updates with @rhshadrach. I am actually new to contributing to open source. I myself am quite familiar with Python. Perhaps with regards to this bug, you have any suggestion or tips that would allow helping contribute? thanks in advanced, really love pandas hahaha.

[simonjayhawkins]

@grrlic in the first instance, if you are unfamiliar with the codebase, I would probably bisect the regression, see #35685 to get a idea of where a fix should be applied.

although for this case, it might be quite straightforward to follow the code from to_hdf (I notice from git blame that there we changes to the signature in #29957) It appears that dropna is not passed onto store.put in the if append else block