pd.Series to_csv method does not recognize "compression"

[jolespin]

There looks like something is wrong with the compression argument in to_csv forpd.Series but pd.DataFrame. This error is also in 0.20.3.

In [1]: Se_tmp = pd.Series(list("ACGT"))
   ...: Se_tmp.to_csv("~/test.tsv.gz", sep="\t", compression="gzip")
   ...:
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-1-de7ce483228f> in <module>()
      1 Se_tmp = pd.Series(list("ACGT"))
----> 2 Se_tmp.to_csv("~/test.tsv.gz", sep="\t", compression="gzip")

TypeError: to_csv() got an unexpected keyword argument 'compression'

In [2]: Se_tmp.to_frame("testing_column").to_csv("~/test.tsv.gz", sep="\t", compression="gzip")

In [3]: pd.__version__
Out[3]: '0.21.1'

[gfyoung]

@jolespin : What can I say but oops? 😄 By all means, surface this parameter for Series.to_csv !

[gfyoung]

@jreback : From a maintenance perspective, having to keep all of these signatures updated is kind of annoying, especially since all that Series.to_csv does is call DataFrame.to_csv. Is there a way that we could create one function (similar to how we implement DataFrame.merge for example) that we can then call for both Series and DataFrame and then not have to replicate the signature everywhere?

[jreback]

@gfyoung not like merge, rather it is straightforward to move the .to_csv functions to generic.py to accomplish this, though its slightly tricky because the doc-string needs to be reflexive of the class (e.g. say Series or DataFrame) as appropriate, but that can be accomplished thru shared_kwargs , but that itself is a cost, e.g. look at .reindex we still define it in series.py though we use *args, **kwargs (we could/should actually write down the signature here though).

so yes I think it is possible to be simpler / less error prone. but you have to pick your poison a bit. e.g. we can share doc-strings and simpl pretty easy, the signature itself is slightly tricker.

that said I think we should move the .to_csv impl to generic. pls create an issue (and if you want to create a more general issue about doing this, though IIRC there might be one already).