DOC: Replace .values with .to_numpy() in enhancingperf

[huizew]

As suggested in #24807 (comment)

Replace .values with .to_numpy() in the benchmark demonstration code.

[WillAyd]

Thanks for the PR! How much effort do you think it would it be to swap out all of these instances across the documentation?

[codecov]

Codecov Report

Merging #26313 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #26313      +/-   ##
==========================================
- Coverage   92.04%   92.03%   -0.01%     
==========================================
  Files         175      175              
  Lines       52302    52302              
==========================================
- Hits        48142    48137       -5     
- Misses       4160     4165       +5

Flag	Coverage Δ
#multiple	90.59% <ø> (ø)	⬆️
#single	40.73% <ø> (-0.17%)	⬇️

Impacted Files	Coverage Δ
pandas/io/gbq.py	78.94% <0%> (-10.53%)	⬇️
pandas/core/frame.py	97.01% <0%> (-0.12%)	⬇️
pandas/util/testing.py	90.6% <0%> (-0.11%)	⬇️

---

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 20fa58d...05789ca. Read the comment docs.

[codecov]

Codecov Report

Merging #26313 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #26313      +/-   ##
==========================================
- Coverage   92.04%   92.03%   -0.01%     
==========================================
  Files         175      175              
  Lines       52302    52302              
==========================================
- Hits        48142    48137       -5     
- Misses       4160     4165       +5

Flag	Coverage Δ
#multiple	90.59% <ø> (ø)	⬆️
#single	40.73% <ø> (-0.17%)	⬇️

Impacted Files	Coverage Δ
pandas/io/gbq.py	78.94% <0%> (-10.53%)	⬇️
pandas/core/frame.py	97.01% <0%> (-0.12%)	⬇️
pandas/util/testing.py	90.6% <0%> (-0.11%)	⬇️

---

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 20fa58d...bd2b70e. Read the comment docs.

[huizew]

@WillAyd All the .values in this file have been replaced.

Took a look at the other files under pandas/doc/source. Apart from whatsnew folder, no other file uses .values for pandas objects. And I suppose whatsnew files shouldn't be changed, right?

[WillAyd]

Great thanks for checking. I think this looks good - @gfyoung any thoughts?

[gfyoung]

For our own edification, can those benchmark numbers be double-checked (i.e. the ones that follow the timeit commands)? Are those still approximately correct when we swap in to_numpy?

And how much better is it vs. using .values ?

[huizew]

Thanks for the review.

I just checked that changing to .to_numpy() doesn’t change the benchmark time very much (less than 1%), and all the points that this guide tries to demonstrate still hold.

The performance comparison between .to_numpy() and .values is not discussed in this guide page. The difference seems negligible when compared to other things discussed in this guide page, such as Cython and Jit compiling. Personally I think the main reason behind this PR is to follow the documentation/community’s encouragement to use .to_numpy() rather than .values

[gfyoung]

Personally I think the main reason behind this PR is to follow the documentation/community’s encouragement to use .to_numpy() rather than .values

True, but always good to double check to make sure we aren't actually proposing a performance regression in our docs.

[gfyoung]

Thanks @huizew !