REF/CLN: test_get_dummies

[jbrockmendel]

No description provided.

[jreback]

lgtm. merge on green.

[jschendel]

lgtm, one small comment

[datapythonista]

Is there an ongoing effort to replace pd.DataFrame by DataFrame in our code and to be redundant with the test method names (TestGetDummmies.test_basic by TestGetDummmies.test_get_dummies_basic)?

Replacing loops by parametrization improves readability, but these other changes look very personal and opinionated, and if we want to make them a standard, would be nice to have an issue referenced here. Otherwise we can end up in a situation where every time someone works with a file, we'll get some pd. added and removed (for example, I would personally prefer to always have the module name and not import objects).

[jbrockmendel]

and to be redundant with the test method names (TestGetDummmies.test_basic by TestGetDummmies.test_get_dummies_basic)?

In this case it is redundant, but in many other cases we'll see something like test_basic in a class without an informative name. It's also not uncommon to see tests for something other than get_dummies living inside a TestGetDummies class. I think being explicit is worthwhile here.

Is there an ongoing effort to replace pd.DataFrame by DataFrame in our code

Not that I'm aware of. For files testing lower-level code, I like to have all the relevant imports at the top so that it is easy to tell what is tested in this file. If pd is in the namespace, I can't rule anything out. Not really relevant for this file, just a habit I've gotten into.

[jbrockmendel]

updated with typo fixup + green

[WillAyd]

(for example, I would personally prefer to always have the module name and not import objects).

Yea @jorisvandenbossche and I agree, but at the same time I think someone needs to champion that cause if we want to establish a standard

[WillAyd]

Thanks @jbrockmendel

[jreback]

(for example, I would personally prefer to always have the module name and not import objects).

Yea @jorisvandenbossche and I agree, but at the same time I think someone needs to champion that cause if we want to establish a standard

i have always been in favor of this :)

moving to this will require a good code check though - s may need be that easy

[jorisvandenbossche]

IMO this doesn't necessarily require a code check initially.
But we could already just stop merging PRs that do changes like pd.DataFrame -> DataFrame (or ask them to revert that part). @WillAyd gave the comment like #33184 (comment) on several PRs recently I think, but in the end we always just merged the PR..

[jreback]

I care most about consistency in a single file meaning we shouldn’t mix ways of referencing common things like Series vs pd.Series

I do prefer using import of Series and not pd.Series

[jorisvandenbossche]

I do prefer using import of Series and not pd.Series

I am confused now. I interpreted the above "i have always been in favor of this" as the opposite (#33184 (comment)), since that is replying to a quote of Will that is basically saying that he, Marc and I prefer to not import objects but modules (so using pd.Series instead of Series)

[jorisvandenbossche]

I think someone needs to champion that cause if we want to establish a standard

I opened a separate issue about it: #33203. Let's try to decide there which of the two ways has our preference.