Upstream port candidates — week of 2026-05-25

[richardwooding]

Summary

Scanned 8 upstream cel2sql Go commits since ef505ab (v3.8.0, Apache Spark dialect — the last sync point for pycel2sql PR #8). All 8 are dep bumps or CHANGELOG prep; no Go-side features to port this week.

Scanned cel2sql4j cross-reference for the same window (commits after PR #10 / PR #9). Found 1 port candidate and 1 already done.

---

Port candidates

1835215 — fix: pass elem writer to JSON array membership dialect methods (cel2sql4j PR #20)

Upstream commit: SPANDigital/cel2sql4j@1835215

Scope: The write_json_array_membership and write_nested_json_array_membership abstract methods in the Dialect ABC currently receive only a write_array writer — not the element being tested for membership. This matches the pre-fix cel2sql4j state and produces broken SQL in most dialects.

grep result (confirms old signatures in pycel2sql):

src/pycel2sql/dialect/_base.py:131:    def write_json_array_membership(
src/pycel2sql/dialect/_base.py:132:        self, w: StringIO, json_func: str, write_expr: WriteFunc   # ← no write_elem
src/pycel2sql/dialect/_base.py:136:    def write_nested_json_array_membership(
src/pycel2sql/dialect/_base.py:137:        self, w: StringIO, write_expr: WriteFunc                   # ← no write_elem

What the cel2sql4j fix shows needs changing:

Dialect	Current pycel2sql output (broken)	Correct output after fix
PostgreSQL	ANY(ARRAY(SELECT jsonFunc(arr))) (works accidentally)	elem = ANY(ARRAY(SELECT jsonFunc(arr))) — signature update only
DuckDB	elem = (SELECT value FROM json_each(arr)) — scalar subquery returns last row	EXISTS (SELECT 1 FROM json_each(arr) WHERE value = elem)
SQLite	elem = (SELECT value FROM json_each(arr)) — same scalar subquery bug	EXISTS (SELECT 1 FROM json_each(arr) WHERE value = elem)
BigQuery	elem = UNNEST(JSON_VALUE_ARRAY(arr)) — invalid SQL	elem IN UNNEST(JSON_VALUE_ARRAY(arr))
MySQL	elem = JSON_CONTAINS(arr, CAST(? AS JSON)) — compares elem to 0/1	JSON_OVERLAPS(JSON_ARRAY(elem), arr)
Spark	raises UnsupportedDialectFeatureError at conversion time	array_contains(from_json(arr, 'ARRAY<STRING>'), elem)

Additionally: the converter's _visit_in (converter.py:396) never routes to write_json_array_membership at all — it always falls through to write_array_membership. Adding the routing is part of the same fix. The cel2sql4j Converter.visitInJSONArray and the Go cel2sql.go:774–794 both show the required detection logic.

cel2sql4j cross-reference: cel2sql4j PR #20 is the direct source. The Go upstream (cel2sql) still has the old signatures in its dialect/dialect.go — this bug has only been fixed in the Java port so far.

Two-test change in test_spark.py: test_json_array_membership_dialect_method_raises and test_nested_json_array_membership_dialect_method_raises currently assert that the Spark dialect raises — they should be updated to assert the correct array_contains(from_json(...), elem) output once the fix lands.

Portability assessment: medium (6 dialect files + _base.py ABC + _converter.py routing + test updates across dialect test files and test_spark.py).

---

Already done

5e8eaa6 — Add three repo-local agent skills + fix missed Spark writeFormat impl (cel2sql4j PR #11)

The subject mentions "fix missed Spark writeFormat impl" — a Spark writeFormat method that was omitted in the initial cel2sql4j Spark port. pycel2sql already has this: SparkDialect.write_format is implemented at src/pycel2sql/dialect/spark.py:521 (emits format_string(...)) and was included in pycel2sql PR #8. The skills additions are chore and don't port to pycel2sql.

---

Out of scope

All 8 cel2sql Go commits since v3.8.0 (ef505ab..65f4a6c) are dependency bumps and CHANGELOG prep; none contain feature or fix changes to port.

---

Generated by the weekly upstream-scan routine.
https://claude.ai/code/routines/weekly-upstream-port-scan