fix(generic-axes): apply contribution before flatten

[villebro]

SUMMARY

Currently the contribution operation is applied after flattening, despite only being meant to apply to numeric columns. By ensuring that the dataframe is unflattened, the index columns will always be absent from the columns, which ensures that the index is not used in the contribution calculation. Also, in the case of using a numeric index, applying the contribution calculation after flattening would cause the index to leak into the contribution calculation. In addition to the fix, falsy post transformation ops are removed to clean up the request payload (these cause unnecessary cache misses).

There is already a check that should remove any non-numeric columns:

superset/superset/utils/pandas_postprocessing/contribution.py

Line 51 in 4435e53

numeric_df = contribution_df.select_dtypes(include=["number", Decimal])

However, for some reason this currently also returns string-columns (object), possibly due to the Decimal type extending object (seems weird, I know, but the only reason I could come up with). However, I didn't want to make any changes to this logic in this PR, as refining it will probably require additional tests to be added + we should potentially bump Pandas to 1.4 before changing this.

AFTER

Now calculating contributions works when GENERIC_CHART_AXES enabled:

BEFORE

Previously it would fail:

TESTING INSTRUCTIONS

1. Enable GENERIC_CHART_AXES feature flag
2. Create a chart with a non-temporal x-axis
3. Enable contribution mode (either one is ok)

ADDITIONAL INFORMATION

• Has associated issue:
• Required feature flags:
• Changes UI
• Includes DB Migration (follow approval process in SIP-59)
  • Migration is atomic, supports rollback & is backwards-compatible
  • Confirm DB migration upgrade and downgrade tested
  • Runtime estimates and downtime expectations provided
• Introduces new feature or API
• Removes existing feature or API

[zhaoyongjie]

LGTM

[zhaoyongjie]

Nice! The filter should apply in a separate PR and apply in all post operators.

[codecov]

Codecov Report

Merging #20077 (e94f064) into master (4435e53) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master   #20077   +/-   ##
=======================================
  Coverage   66.37%   66.38%           
=======================================
  Files        1715     1715           
  Lines       64179    64180    +1     
  Branches     6753     6753           
=======================================
+ Hits        42602    42603    +1     
  Misses      19859    19859           
  Partials     1718     1718           

Flag	Coverage Δ
javascript	51.33% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
.../plugin-chart-echarts/src/Timeseries/buildQuery.ts	71.42% <100.00%> (+4.76%)	⬆️

---

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 4435e53...e94f064. Read the comment docs.