fix: json_each on SQLite generates invalid SQL

[scottlaird]

Fixes #168.

SQLIte's json_each and MySQL's JSON_TABLE are table-valued functions, and cel2sql generates invalid SQL for them, breaking the ability to do comprehensions across JSON fields.

For example, for SQLite approval.exists(a, a == "alice") generates this SQL:

EXISTS (SELECT 1 FROM json_each(approvals) AS a WHERE a = ?)

However, for SQLite rejects this with no such column: a at query time. The correct SQL would be

EXISTS (SELECT 1 FROM json_each(approvals) AS a WHERE a.value = ?)

Since JSON handling isn't standardized in SQL, each dialect support is slightly different:

• BigQuery, DuckDB, PostgreSQL, Spark: UNNEST and EXPLODE are row-valued and work as-is.
• SQLite and MySQL: json_each and JSON_TABLE are table-valued and do not work.

This PR allows dialects to generate the correct syntax for JSON comprehensions and fixes SQLite and MySQL.

Scope

all, exists, exists_one, filter and map all go through the same source writer, so all five were affected.

Tests

sqlite_iteration_variable_test.go converts and then executes against an in-memory SQLite database, asserting the rows that come back. That is the check this bug asks for: the previous expectations were SQL that parsed and could not run. Against main it fails with no such column: a.

The shared SQLite expectations in testcases/comprehension_tests.go encoded the old output and are updated; they are the same five comprehensions.

go test ./... is otherwise unchanged from main on this machine — the 27 failures before and after are the testcontainers suites, which need Docker.

Note on the MySQL change

I included MySQL because WriteUnnest there emits JSON_TABLE(… COLUMNS(value TEXT PATH '$')), which is also table-valued, although the exact structure is slightly different. I could not execute against MySQL here, so that half is untested. I can split it out into its own PR if you'd prefer.

[scottlaird]

@richardwooding SQLite and MySQL have broken implementations of JSON decoding; this should fix them.

[richardwooding]

Thanks @scottlaird I will get this PR in

[codecov-commenter]

Welcome to Codecov 🎉

Once you merge this PR into your default branch, you're all set! Codecov will compare coverage reports and display results in all future pull requests.

Thanks for integrating Codecov - We've got you covered ☂️

[richardwooding]

Thanks @scottlaird for the excellent fix and the thorough write-up — the executable SQLite test that actually runs the generated SQL is exactly the right kind of regression coverage for this, and including the MySQL JSON_TABLE half saves us a follow-up. I pushed a small lint suppression for the intentional SQL concatenation in the new test. The remaining red checks are pre-existing dependency vulnerabilities failing on main (being fixed in #170) and the benchmark-storage step that can't run from forks. Merging!