fix: add nil-pointer guard to the bind for driver.Valuer args

[ciricc]

Solve #1823

[CLAassistant]

All committers have signed the CLA.

[github-actions]

Summary

This PR guards against a nil-pointer panic that occurs when a *T (where T implements driver.Valuer with a value receiver) is passed as a bind argument. A nil *T satisfies the interface but calling fn.Value() dereferences the nil pointer and panics. The fix adds an isNilPointer helper and calls it before fn.Value() in all three bind paths (bindPositional, bindNumeric, bindNamed). It also opportunistically refactors the pre-existing nil-Stringer guard in format() to use the same helper. The core fix is correct and the approach is sound.

Should fix

• Missing regression test in tests/issues/: Per the project convention (CLAUDE.md), regression tests for bug fixes must live in tests/issues/issue_1823_test.go. The unit tests added to bind_test.go are valuable and should stay, but they only test the bind() helper in isolation. An integration test exercising the full query path is also needed — one that passes a nil driver.Valuer arg to an actual Exec or Query call — covering both TCP and HTTP protocols, and both the native (clickhouse.Open) and database/sql (clickhouse_std) API surfaces.

Nits

• nit: The refactoring of the fmt.Stringer case in format() is a subtle behavioral change. The original code also checked v.Type().Elem().Implements(...), which meant a nil *T where String() is a pointer receiver (only *T satisfies Stringer, not T) would NOT get the "NULL" short-circuit and would fall through to v.String(). The new isNilPointer(v) returns "NULL" for all nil pointers unconditionally. This is arguably more correct (calling a nil pointer receiver panics unless the method explicitly handles nil), but the change is worth a one-line comment so future readers understand why the elem-implements check was dropped.

• nit (pre-existing, not introduced here): bindPositional (line 128) and bindNumeric (line 170) both do return "", nil instead of return "", err when fn.Value() fails. The caller silently sees a success with an empty string rather than the error. Worth a follow-up fix — noting it here so it doesn't get buried.

Verdict

Approve — the nil-pointer guard is correct and the unit tests confirm the fix. Add the tests/issues/issue_1823_test.go integration test before or shortly after merge to meet the project's regression-test convention.