feat(python): Add a show method to DataFrame and LazyFrame

[guilhem-dvr]

This adds a show method for both DataFrame and LazyFrame objects, taking inspiration from pyspark's show method and taking into account the requirements from @stinodego in #16534.

I choose to only expose config options that influence the result width's size, to mimic pyspark's truncate option.

I've provided tests, but I'm not super satisfied with them: they could break when changing the default display options. I was thinking of mocking Config, print and display_html, would that be okay?

[alexander-beedie]

I've provided tests, but I'm not super satisfied with them: they could break when changing the default display options. I was thinking of mocking Config, print and display_html, would that be okay?

No need to mockConfig; it can act as a decorator¹, so you could decorate your tests such that you explicitly set the Config to some known values (eg: the current defaults) and then you can modify the values away from these defaults inside the test 👍

There are a few other options it would be nice to expose as well, such as tbl_formatting, tbl_cell_alignment, tbl_cell_numeric_alignment, etc.

I'd also change the parameter name n to limit, and allow it to be None (but maintain the default of 5). This would allow the caller to print the entire table (eg: with no limit).

Footnotes

1. Config as a decorator:
   https://docs.pola.rs/api/python/stable/reference/config.html#use-as-a-decorator ↩

[codecov]

Codecov Report

Attention: Patch coverage is 82.14286% with 5 lines in your changes missing coverage. Please review.

Project coverage is 79.88%. Comparing base (e8e0295) to head (399ddb1).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
py-polars/polars/dataframe/frame.py	77.27%	3 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #19634      +/-   ##
==========================================
- Coverage   79.88%   79.88%   -0.01%     
==========================================
  Files        1593     1593              
  Lines      227649   227676      +27     
  Branches     2600     2607       +7     
==========================================
+ Hits       181860   181880      +20     
- Misses      45192    45198       +6     
- Partials      597      598       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[guilhem-dvr]

@alexander-beedie I've added all config options that impact the display format of a dataframe.

I was thinking of hiding the dataframe shape by default because I find it a bit irrelevant when showing a dataframe, usually you would know what is the shape of the frame you are working with. But with the limitless option and for the sake of consistency I think I will leave it visible.

[alexander-beedie]

@alexander-beedie I've added all config options that impact the display format of a dataframe.

Good stuff, will take a look.

I was thinking of hiding the dataframe shape by default because I find it a bit irrelevant when showing a dataframe, usually you would know what is the shape of the frame you are working with.

Not necessarily - if it's wide (so cols are truncated in the repr) or you've just filtered the data you won't know the shape; definitely want to keep it 👍

[alexander-beedie]

@guilhem-dvr: the tests seem to show a few issues still to resolve. Looks like some state might be escaping (either from the function itself or, more likely, the tests?) 🤔

[guilhem-dvr]

@guilhem-dvr: the tests seem to show a few issues still to resolve. Looks like some state might be escaping (either from the function itself or, more likely, the tests?) 🤔

Somehow I was expecting this to be unrelated 😓 I will have a look.

[guilhem-dvr]

I found the issue @alexander-beedie: decorating tests with Config was causing a config leak to other tests. I tried to replace the config decorators with calls to Config.set inside the tests, but this made it worst and tests became flaky. I ended up wrapping each test inside a context managed config, and this worked.

[guilhem-dvr]

Hi @alexander-beedie, I think this PR is ready, would you care to have a look at it?

[alexander-beedie]

Apologies, lost track of this over the end of year crunch at work and Xmas break - will take a fresh look over it at the weekend, as I think this will be a nice addition 👌