feat: support None operand in EQUAL operator

[zhaoyongjie]

SUMMARY

This PR intends to support the None operand with Equal and Not-Equal operators in Filters in QueryObject. Currently, the In operator has supported None operand calculation, in order to align that the Equal and Not-Equal should support as well.

After the PR, there are 4 approaches to generate is null or is not null filter in SQL.

1. EQUAL / NOT EQUAL operator

>>> result = mock_dataset.query({
    "metrics": ["count"],
    "filter": [{"col": "col", "val": None, "op": "=="}],
    "is_timeseries": False,
})
>>> result.query
SELECT count(*) AS count
FROM mock_dataset
WHERE col IS NULL

2. IS NULL / IS NOT NULL operator

>>> result = mock_dataset.query({
    "metrics": ["count"],
    "filter": [{"col": "col", "op": "IS NULL"}],
    "is_timeseries": False,
})
>>> result.query
SELECT count(*) AS count
FROM mock_dataset
WHERE col IS NULL

3. IN operator

>>> result = mock_dataset.query({
    "metrics": ["count"],
    "filter": [{"col": "col", val: ["a", None], "op": "IN"}],
    "is_timeseries": False,
})
>>> result.query
SELECT count(*) AS count
FROM mock_dataset
WHERE col IS NULL
  OR col IN ('a')

4. There is an additional case that <NULL> as filter operand literal constant in the EQ or IN operators

>>> result = mock_dataset.query({
    "metrics": ["count"],
    "filter": [{"col": "col", "val": '<NULL>', "op": "=="}],
    "is_timeseries": False,
})
>>> result.query
SELECT count(*) AS count
FROM mock_dataset
WHERE col IS NULL

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

TESTING INSTRUCTIONS

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 #21713 (3614b05) into master (51c54b3) will decrease coverage by 0.02%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master   #21713      +/-   ##
==========================================
- Coverage   66.84%   66.82%   -0.03%     
==========================================
  Files        1799     1799              
  Lines       68881    68881              
  Branches     7319     7319              
==========================================
- Hits        46044    46028      -16     
- Misses      20950    20966      +16     
  Partials     1887     1887              

Flag	Coverage Δ
hive	52.91% <100.00%> (ø)
mysql	78.24% <100.00%> (+<0.01%)	⬆️
postgres	78.30% <100.00%> (+<0.01%)	⬆️
presto	52.81% <100.00%> (ø)
python	81.39% <100.00%> (-0.05%)	⬇️
sqlite	?
unit	50.95% <0.00%> (ø)

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

Impacted Files	Coverage Δ
superset/charts/schemas.py	99.35% <ø> (ø)
superset/connectors/sqla/models.py	90.68% <100.00%> (+0.08%)	⬆️
superset/db_engine_specs/sqlite.py	89.28% <0.00%> (-7.15%)	⬇️
superset/utils/celery.py	86.20% <0.00%> (-3.45%)	⬇️
superset/connectors/sqla/utils.py	87.25% <0.00%> (-1.97%)	⬇️
superset/result_set.py	96.29% <0.00%> (-1.49%)	⬇️
superset/views/core.py	75.45% <0.00%> (-0.61%)	⬇️
superset/db_engine_specs/base.py	89.38% <0.00%> (-0.32%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[zhaoyongjie]

bycatch: use only_postgresql decorator to narrow the test case, and use login_as_admin to define the Flask app context, there's no logical change.

[zhaoyongjie]

This test case is the new one.

[villebro]

I hadn't noticed it actually allowed selecting the empty value despite the API not allowing it 😵‍💫 Works great! However, one bug that I noticed when testing + a few stylistic nits:

1. When creating a new filter, then selecting "IS NOT NULL" operator and then switching back to "equals" or "not equals", the save button becomes disabled:
   

2. We should also update the pill labels to reflect that they are checking for NULL. For the "equals" case I'd do something like "col = NULL" and "col <> NULL":
   
   

3. Also a bycatch, but I'd actually prefer for the NOT EQUALS operator to use the same ≠ character in the pill label that we're using in the dropdown:
   

[villebro]

LGTM - leaving my comments on the table for a potential follow-up PR as they're mostly unrelated

[zhaoyongjie]

@villebro Thanks a lot, I will fix the frontend problem in the separated PR.

[john-bodley]

@zhaoyongjie did you address the frontend issues that @villebro mentioned in a separate PR? If so could you reference this PR in said PR so there's a edge connecting the two PRs?

[zhaoyongjie]

@john-bodley I will do these cases. thanks for the mention.