| From: | Neil Chen <carpenter(dot)nail(dot)cz(at)gmail(dot)com> |
|---|---|
| To: | Postgres hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: [PATCH] Handle SQL NULL results from casts in JSON generation |
| Date: | 2026-11-20 06:09:33 |
| Message-ID: | CAA3qoJknBrQWz3ObHrMsyqmW-oZrtiU_V1iRyGV57qhdjrHj9Q@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi,
On Fri, Nov 20, 2026 at 11:36 AM Neil Chen <carpenter(dot)nail(dot)cz(at)gmail(dot)com>
wrote:
>
> With the patch, to_json('unset'::json_cast_flag) IS NULL is false.
> Adding ::json to the argument makes the same test true. Should
> to_json() and to_jsonb() return SQL NULL for this top-level case?
>
>
I lean toward returning SQL NULL here, so that's what v2 does.
It gives the same result as casting to json first, and callers can
still distinguish SQL NULL from JSON null with IS NULL or COALESCE.
Using the enum and cast from the original reproducer:
-- [null] with either patch
SELECT jsonb_build_array('unset'::json_cast_flag);
-- v1 returns f | f. v2 returns t | t.
SELECT to_json('unset'::json_cast_flag) IS NULL,
to_jsonb('unset'::json_cast_flag) IS NULL;
Would you prefer SQL NULL or JSON null for the direct calls?
make check passed with v2 on master (d94964f14f3, 243 tests).
--
Ze Chen (Neil)
HighGo Software Co., Ltd.
https://www.highgo.com/
| Attachment | Content-Type | Size |
|---|---|---|
| v2-0001-Handle-SQL-NULL-results-from-casts-in-JSON-generation.patch | text/x-patch | 10.0 KB |
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Fujii Masao | 2026-11-20 06:16:50 | Re: psql: avoid over-reading unterminated prompt escapes |
| Previous Message | Sergey Tatarintsev | 2026-11-20 06:08:41 | [PATCH] Clamp auto-tuned transaction_buffers to at least 8 SLRU banks |