Issue 283 nano plot scaling

[minimav]

Summary

This PR fixes a bug whereby values from unselected rows affect the scaling in nanoplots. In addition to the single values case given in #283, a simple example of the bug in the context of multiple values per row is below (when autoscale is used):

multiple_vals_df = pd.DataFrame(
    {
        "i": list(range(1, 3)),
        "lines_small": ["12.44 6.34", "5.2 -8.2 10"],
        "lines_large": ["12.44, 6.34", "5.2 -8.2 100"],
    }
)
(
    GT(multiple_vals_df, rowname_col="i")
    .fmt_nanoplot(columns="lines_small", rows=[0], plot_type="line", autoscale=True)
    .fmt_nanoplot(columns="lines_large", rows=[0], plot_type="line", autoscale=True)
)

Single values example post-fix:

Multiple values example post-fix:

I have added a couple of unit tests based on existing ones, let me know if they work. Thanks for your work on this library, I have been using it at work and really enjoying it 🙂

Related GitHub Issues and PRs

• Ref: fmt_nanoplot includes excluded rows when scaling plot #283

Checklist

• I understand and agree to the Code of Conduct.
• I have followed the Style Guide for Python Code as best as possible for the submitted code.
• I have added pytest unit tests for any new functionality.

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 92.50%. Comparing base (9874b7c) to head (13b63b0).

Files with missing lines	Patch %	Lines
great_tables/_formats.py	83.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #404      +/-   ##
==========================================
+ Coverage   92.31%   92.50%   +0.18%     
==========================================
  Files          48       48              
  Lines        6039     6043       +4     
==========================================
+ Hits         5575     5590      +15     
+ Misses        464      453      -11     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[machow]

I'm so sorry it took us so long to get to this. @rich-iannone do you mind looking through? If needed, we may just need to start a new PR to address, since we've made a lot of changes since then.

[rich-iannone]

This PR fixes the issue that fmt_nanoplot() used all rows for scaling of single-value nanoplots (one value per cell), even when rows= restricted which rows received nanoplots. This meant that an extreme value in a non-selected row would distort the scale of the selected rows' plots.

With this PR, here's an example where we generate nanoplots in a column of values that are common but exclude the last value (which are different between the two DFs, and very large in df_extreme).

import polars as pl
from great_tables import GT

df_normal = pl.DataFrame({"vals": [-5.3, 6.3, 5.2, -8.2, 10]})
df_extreme = pl.DataFrame({"vals": [-5.3, 6.3, 5.2, -8.2, 500.0]})

gt_normal = GT(df_normal).fmt_nanoplot(
    columns="vals",
    plot_type="bar",
    rows=[0, 1, 2, 3],
)

gt_extreme = GT(df_extreme).fmt_nanoplot(
    columns="vals",
    plot_type="bar",
    rows=[0, 1, 2, 3],
)

In the first table (gt_normal), the excluded value is 10 and so it's difficult to notice any scale effects.

gt_normal.show("browser")

However, the same values that exclude a large value of 500 show the same bar plots. The extreme value of 500 isn't considered in the scaling of the other values (they would appear compressed to the zero line).

gt_extreme.show("browser")

The unit tests compare single SVGs from two different DataFrames (with different unselected values). The equivalence check of the first row's SVG across the DFs indirectly shows that there aren't scaling effects from unselected cells.

[rich-iannone]

LGTM!