sql: fix crash when using a partial index with predicate that has a UDF

[dils2k]

Using a partial index with predicate that references a UDF produces a Project expression that was not handled during partial index contruction in the optimizer, causing an error. This commit fixes it by handling Project expressions and also inlining additional expressions.

Epic: CRDB-55454
Release note: None
Fixes: #155488

[trunk-io]

Merging to master in this repository is managed by Trunk.

• To merge this pull request, check the box to the left or comment /trunk merge below.

After your PR is submitted to the merge queue, this comment will be automatically updated with its status. If the PR fails, failure details will also be posted here

[blathers-crl]

It looks like your PR touches production code but doesn't add or edit any test code. Did you consider adding tests to your PR?

Thank you for contributing to CockroachDB. Please ensure you have followed the guidelines for creating a PR.

My owl senses detect your PR is good for review. Please keep an eye out for any test failures in CI.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}

[cockroach-teamcity]

This change is 

[blathers-crl]

It looks like your PR touches production code but doesn't add or edit any test code. Did you consider adding tests to your PR?

Thank you for updating your pull request.

My owl senses detect your PR is good for review. Please keep an eye out for any test failures in CI.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}

[DrewKimball]

This approach looks good to me. We'll need tests for both changes, though.

@DrewKimball reviewed 2 files and all commit messages, and made 1 comment.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on dils2k).

[dils2k]

@DrewKimball Actually I was thinking about this approach and it turns out we will get an error every time a partial index predicate has a non-inlinable UDF. For example, if we use a bitwise "and" instead of a module operator, then we will get the same error:

CREATE FUNCTION is_even(i INT) RETURNS BOOL AS $$
  -- NOTE: bitwise operator, not module
  SELECT i & 1 = 0;
$$ LANGUAGE SQL IMMUTABLE;

CREATE TABLE t (
  k INT PRIMARY KEY,
  a INT,
  INDEX (a) WHERE is_even(k)
);

CREATE INDEX partial_idx ON t(a) where is_even(k);

-- ERROR: index "t_a_idx" is a partial index that does not contain all the rows needed to execute this query SQLSTATE:
-- 42809

Should I also include bitwise operations in CanInline or we should reconsider this approach, because I think there are more operations that could trigger this error, for example consider this:

CREATE FUNCTION is_cat(s varchar) RETURNS BOOL AS $$
  SELECT lower(s) = 'cat';
$$ LANGUAGE SQL IMMUTABLE;

CREATE TABLE tt ( a varchar );

CREATE INDEX partial_idx ON tt(a) WHERE is_cat(a);

-- ERROR: index "partial_idx" is a partial index that does not contain all the rows needed to execute this query
-- SQLSTATE: 42809

[DrewKimball]

@dils2k I think we can safely add simple operators like bitwise AND to that list, maybe in separate commit or PR. You're right that in the general case, this isn't enough to prevent is a partial index that does not contain all the rows needed, but that seems OK for now, since still an improvement over the internal error. We could open a new issue to track that limitation. Also, I wonder if marking the UDF as VOLATILE works as a workaround?

[blathers-crl]

Thank you for updating your pull request.

My owl senses detect your PR is good for review. Please keep an eye out for any test failures in CI.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}

[dils2k]

@DrewKimball PTAL when you have time. There were places in the tests where division was (AFAIU) expected to be a non-inlinable operation, I replaced it with a bitwise AND in those cases.

[dils2k]

I wonder if marking the UDF as VOLATILE works as a workaround?

I receive this error:

ERROR: is_even(): volatile functions are not allowed in INDEX PREDICATE
SQLSTATE: 0A000