Skipif no scipy

[WillAyd]

• progress towards TST: make _skip_if into pytest decorators #18190
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[WillAyd]

I intentionally did not replace this with a decorator because it looks unnecessary (none of the subsequent code requires spicy)

[jreback]

ok

[WillAyd]

There were a few instances where part of the function required scipy and the other part did not. I parametrized these using kwargs to give better visibility into what is being passed / skipped, especially on machines that don't have scipy installed

[WillAyd]

I was thinking of using kwargs for this function as noted in the subsequent comment, but it would slightly change the test given expected is a copy of the interpolation being done without scipy. The subsequent scipy interpolation is still compared back to that object for equality, which may or may not be intentional

[WillAyd]

As I'm adding comments I just realized that I can almost completely get rid of the tm.skip_if_no_package function as part of this change. It's practically always used used to check scipy (there's one check for pytables) and could be replaced by the skip_if_no function in the _test_decorators module. Will clean up in next push

[codecov]

Codecov Report

Merging #18794 into master will decrease coverage by 0.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #18794      +/-   ##
==========================================
- Coverage   91.63%   91.62%   -0.02%     
==========================================
  Files         154      154              
  Lines       51422    51428       +6     
==========================================
- Hits        47121    47120       -1     
- Misses       4301     4308       +7

Flag	Coverage Δ
#multiple	89.49% <100%> (ø)	⬆️
#single	40.83% <100%> (-0.1%)	⬇️

Impacted Files	Coverage Δ
pandas/util/_test_decorators.py	94.54% <100%> (+0.66%)	⬆️
pandas/io/gbq.py	25% <0%> (-58.34%)	⬇️
pandas/core/frame.py	97.68% <0%> (-0.11%)	⬇️
pandas/util/testing.py	82.91% <0%> (+0.19%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fe900cb...4aa0b7f. Read the comment docs.

[WillAyd]

@jreback how would you feel if we got rid of a dedicated skip_if_no_scipy decorator? The current dedicated function goes through importing the stats, sparse and interpolate subpackages. Looking at the SciPy source all three of those are available as far back as v0.6 released in 2007.

The test cases that use the existing function all seem to require just one and only one of those sub-packages, so we could either:

1. Be explicit about the sub package required (ex: skip_if_no('scipy.stats') OR
2. Simply do skip_if_no('scipy`), reasonably assuming the required sub-package is there

[jreback]

ok with removing extra decorators
in the one hand it is nice to have a dedicated function
if we r using it a lot but scipy skips are not so often

[jreback]

lgtm. lmk when you are finished. (you mentioned might be changing the decorator)

[jreback]

ok

[WillAyd]

I'm OK to have this merged as is. Not sure the juice is worth the squeeze yet in going back and getting rid of that decorator, so I'll defer for now.

tm.skip_if_no_package still needs to be replaced but that's already called out as a separate checklist item. While most calls of that can be replaced with the scipy decorator, there's one instance containing a max_version that isn't built out in the new module already, so I'd rather tackle that in another change.

[jreback]

thanks @WillAyd keep em coming!