fix(csv): Do not coerce persisted data integer columns to float

[john-bodley]

SUMMARY

Regrettably #20151 wasn't suffice is the result set was stored prior to downloading the CSV file. More specifically Pandas coerces an integer array with None to a float—likely because of the Numpy coercion, i.e.,

>>> pd.DataFrame.from_records([{"foo": 1}, {"foo": None}])
   foo
0  1.0
1  NaN

The fix is to explicitly define the dtype, using the standard DataFrame constructor, i.e.,

>>> pd.DataFrame(data=[{"foo": 1}, {"foo": None}], dtype=object)
    foo
0     1
1  None

Long term we should probably replace quirky Pandas with PyArrow globally.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

TESTING INSTRUCTIONS

CI.

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

[codecov]

Codecov Report

Merging #20760 (4d2f439) into master (e60083b) will decrease coverage by 11.48%.
The diff coverage is 0.00%.

@@             Coverage Diff             @@
##           master   #20760       +/-   ##
===========================================
- Coverage   66.35%   54.87%   -11.49%     
===========================================
  Files        1754     1754               
  Lines       66689    66688        -1     
  Branches     7049     7049               
===========================================
- Hits        44253    36595     -7658     
- Misses      20639    28296     +7657     
  Partials     1797     1797               

Flag	Coverage Δ
hive	53.23% <0.00%> (+<0.01%)	⬆️
mysql	?
postgres	?
presto	53.09% <0.00%> (+<0.01%)	⬆️
python	58.00% <0.00%> (-23.69%)	⬇️
sqlite	?
unit	50.57% <0.00%> (+<0.01%)	⬆️

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

Impacted Files	Coverage Δ
superset/views/core.py	34.46% <0.00%> (-43.43%)	⬇️
superset/utils/dashboard_import_export.py	0.00% <0.00%> (-100.00%)	⬇️
superset/key_value/commands/update.py	0.00% <0.00%> (-88.89%)	⬇️
superset/key_value/commands/delete.py	0.00% <0.00%> (-85.30%)	⬇️
superset/key_value/commands/delete_expired.py	0.00% <0.00%> (-80.77%)	⬇️
superset/dashboards/commands/importers/v0.py	15.62% <0.00%> (-76.25%)	⬇️
superset/datasets/commands/update.py	25.30% <0.00%> (-68.68%)	⬇️
superset/datasets/commands/create.py	29.41% <0.00%> (-68.63%)	⬇️
superset/datasets/commands/importers/v0.py	24.03% <0.00%> (-67.45%)	⬇️
superset/reports/commands/execute.py	24.45% <0.00%> (-67.16%)	⬇️
... and 275 more

---

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 e60083b...4d2f439. Read the comment docs.

[mbcsa]

Hi @john-bodley

This fix introduces a new problem when user exports CSV file from a cached Query.
I've created a new issue #20919

The thing is, when Dataframe is created dinamically from cached data, it is not respecting column formats.
This is a problem when decimal separator is configured by CSV_EXPORT, "sep" attribute

I'm testing this, and it works well when changing:

df = pd.DataFrame(
    data=obj["data"],
    dtype=object,
    columns=[c["name"] for c in obj["columns"]],
)

to

df = pd.DataFrame(
    data=obj["data"],
    columns=[c["name"] for c in obj["columns"]],
)

Thank you