[DEPR] deprecate pivot_wider

[samukweku]

PR Description

Please describe the changes proposed in the pull request:

• deprecate pivot_wider in favor of pandas pivot
• minor edits to pivot_longer

This PR relates to #1045 .

PR Checklist

Please ensure that you have done the following:

1. PR in from a fork off your branch. Do not PR from <your_username>:dev, but rather from <your_username>:<feature-branch_name>.

2. If you're not on the contributors list, add yourself to AUTHORS.md.

3. Add a line to CHANGELOG.md under the latest version header (i.e. the one that is "on deck") describing the contribution.
   • Do use some discretion here; if there are multiple PRs that are related, keep them in a single line.

Automatic checks

There will be automatic checks run on the PR. These include:

• Building a preview of the docs on Netlify
• Automatically linting the code
• Making sure the code is documented
• Making sure that all tests are passed
• Making sure that code coverage doesn't go down.

Relevant Reviewers

Please tag maintainers to review.

• @ericmjl

[ericmjl]

🚀 Deployed on https://deploy-preview-1263--pyjanitor.netlify.app

[codecov]

Codecov Report

Merging #1263 (4491b1f) into dev (4da9a90) will increase coverage by 10.36%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##              dev    #1263       +/-   ##
===========================================
+ Coverage   87.15%   97.51%   +10.36%     
===========================================
  Files          78       78               
  Lines        3705     3702        -3     
===========================================
+ Hits         3229     3610      +381     
+ Misses        476       92      -384     

[ericmjl]

LGTM!

[fkgruber]

Too bad you are deprecating pivot_wider. Seems like a better useful function. What is the most efficient way to use pivot to get the same outputs pivot_wider? Because by default you get multi-index which are very inconvenient

[samukweku]

Hi @fkgruber kindly share an example you have and I can show how to deal with it. There is also the collapse_levels function to handle multiindex. We would like to not duplicate existing pandas functions. Open to suggestions/feedback. Thanks

[samukweku]

@ericmjl @fkgruber I'll write up a series of examples on the alternatives for the deprecated functions on here. Thoughts ?

[fkgruber]

Here is an example in R. Would love to see the equivalent in pyjanitor using non-deprecated functions

library(tidyverse)
mtcars %>% 
  rownames_to_column() %>%
  group_by(cyl) %>%
  summarise(mean=mean(mpg))%>%
  pivot_wider(names_from = cyl, values_from = mean)%>%
  mutate(Average= mean(c(`6`, `4`, `8`)),
         Diff_4= `4`-Average)

[samukweku]

@fkgruber this is a similar approach in pandas

import pandas as pd
url='https://gist.githubusercontent.com/seankross/a412dfbd88b3db70b74b/raw/5f23f993cd87c283ce766e7ac6b329ee7cc2e1d1/mtcars.csv'
mtcars = pd.read_csv(url)
(mtcars
.groupby('cyl', as_index=False)
.agg(mean=('mpg','mean'))
.assign(index=1)
.pivot(index='index', columns='cyl',values='mean')
.rename_axis(columns=None,index=None)
.assign(Average=lambda df: df.mean(axis='columns'), 
        Diff_4= lambda df: df[4] - df.Average)
.round(2)
)

[fkgruber]

Thanks,
so with pure pandas the pivot_wider provided in R becoes 3 calls
.assign(index=1)
.pivot(index='index', columns='cyl',values='mean')
.rename_axis(columns=None,index=None)
(by the way I have seen examples where .rename_axis(columns=None,index=None) does not work so it is not a general solution)

With janitor and pivot_wider we could do:

(mtcars
 .groupby('cyl', as_index=False)
 .agg(mean=('mpg','mean'))
 .assign(index=1)
 .pivot_wider(index="index",names_from='cyl', values_from='mean')
 .assign(Average=lambda df: df.mean(axis='columns'),
         Diff_4= lambda df: df[4] - df.Average)
)

And if pivot_wider would take of the index as dplyr does, then it would almost be the same number of lines as in R.

Seems like pivot_wider is helpful by reducing extra coding

[samukweku]

Kindly share an example where rename_axis does not work. We could raise an issue with the pandas team. The assign(index=1) is not required in all cases; tidyverse does not assign the same level of importance that pandas does to index.

Pandas pivot already does a great job and there is no need to repeat it; if there are edge cases it can be handled (happy to be proven wrong). Pandas does not have pivot_wider_spec or something close to it; pyjanitor has that.

If you have any other scenario where pivot is not sufficient happy to work through it with you