fix: % replace in values_for_column

[betodealmeida]

SUMMARY

When we run a query we replace %% (in certain SQLAlchemy dialects) with %:

superset/superset/models/core.py

Lines 668 to 669 in 76d897e

	if engine.dialect.identifier_preparer._double_percents: # noqa
	sql = sql.replace("%%", "%")

These dialects use pyformat (%(name)s) or format (%s) to parameterize arguments passed to the cursor. Normally we'd need to pass %% for literal percent signs, but the current behavior for most databases (Postgres, Presto, MySQL, Hive, Druid) is to accept unescaped percent symbols when no parameters are passed — which is the case for the calls Superset does to cursor.execute.

This logic is missing from values_for_column, which results in queries that are valid but, for some reason, extremely slow in Druid:

SELECT * FROM t WHERE name LIKE '%%a';  -- slow

While the expected query is fast:

SELECT * FROM t WHERE name LIKE '%a';  -- fast

This PR adds the same behavior to the values_for_column method.

Note that while drivers accept unescaped percent symbols when no parameters are passed to the cursor, this behavior is undocument in the DB API 2.0 spec. Ideally we'd not do this replacement in Superset, and instead:

1. In SQL Lab we'd escape % by replacing it with %% before sending the query to the DB.
2. When going from Explore to SQL Lab we'd unescape %% back to %.

This way Superset is not relying on undocument behavior. But that fix is more complicated and requires a lot of testing.

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