Обсуждение: IS JSON predicate support for domain base type as JSON/JSONB/BYTEA/TEXT
hi.
src/backend/executor/execExpr.c
case T_JsonIsPredicate:
{
JsonIsPredicate *pred = (JsonIsPredicate *) node;
ExecInitExprRec((Expr *) pred->expr, state, resv, resnull);
scratch.opcode = EEOP_IS_JSON;
scratch.d.is_json.pred = pred;
ExprEvalPushStep(state, &scratch);
break;
}
gram.y:
a_expr IS json_predicate_type_constraint
the above shows the a_expr will be transformed and then evaluated.
The exprType type of a_expr as domain should work just fine.
The attached patch implements this, and it seems to be quite straightforward.
(extensive regress tests added)
CREATE DOMAIN jd1 AS JSON CHECK ((VALUE ->'a')::text <> '3');
CREATE DOMAIN jd2 AS JSONB CHECK ((VALUE ->'a') = '1'::jsonb);
CREATE DOMAIN jd4 AS bytea CHECK (VALUE <> '\x61');
SELECT NULL::jd1 IS JSON;
SELECT NULL::jd2 IS JSON;
SELECT NULL::jd4 IS JSON;
in the master, the above 3 IS JSON would return error,
with the attached patch, it will return NULL.
I checked the discussion links [1], but couldn’t find the reason domains aren’t
supported. I guess at that time, we didn't think about this issue.
[1] https://git.postgresql.org/cgit/postgresql.git/commit/?id=6ee30209a6f161d0a267a33f090c70c579c87c00
--
jian
https://www.enterprisedb.com/
Вложения
jian he <jian.universality@gmail.com> writes:
> [ v1-0001-IS-JSON-predicate-work-with-domain-type.patch ]
This looks like a large patch with a small patch struggling to
get out of it. Why didn't you simply do
- *exprtype = exprType(expr);
+ *exprtype = getBaseType(exprType(expr));
in transformJsonParseArg?
regards, tom lane
On Wed, Nov 19, 2025 at 1:01 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> jian he <jian.universality@gmail.com> writes:
> > [ v1-0001-IS-JSON-predicate-work-with-domain-type.patch ]
>
> This looks like a large patch with a small patch struggling to
> get out of it. Why didn't you simply do
>
> - *exprtype = exprType(expr);
> + *exprtype = getBaseType(exprType(expr));
>
> in transformJsonParseArg?
>
yech.
While at it, I added parser_errposition to the transformJsonIsPredicate ereport.
errmsg("cannot use type %s in IS JSON predicate",
format_type_be(exprtype))
we don't need to worry about exprtype as InvalidOid, because
transformJsonParseArg (exprType(expr)) would fail already in that case.
--
jian
https://www.enterprisedb.com/
Вложения
Hi. The regression test was very verbose; I removed some of it. Also polished function ExecEvalJsonIsPredicate a little bit. -- jian https://www.enterprisedb.com/
Вложения
On 2026-03-12 Th 11:55 PM, jian he wrote: > Hi. > > The regression test was very verbose; I removed some of it. > Also polished function ExecEvalJsonIsPredicate a little bit. Here's a v4. I changed resultBaseType to exprBaseType - I think it's clearer. I also trimmed the tests a bit more, and dropped the new objects after testing them. The error message now shows the domain name rather than the underlying base type. I think that's more useful. cheers andrew -- Andrew Dunstan EDB: https://www.enterprisedb.com
Вложения
On Fri, 13 Mar 2026 at 17:30, Andrew Dunstan <andrew@dunslane.net> wrote: > > > On 2026-03-12 Th 11:55 PM, jian he wrote: > > Hi. > > > > The regression test was very verbose; I removed some of it. > > Also polished function ExecEvalJsonIsPredicate a little bit. > > > > Here's a v4. I changed resultBaseType to exprBaseType - I think it's > clearer. I also trimmed the tests a bit more, and dropped the new > objects after testing them. The error message now shows the domain name > rather than the underlying base type. I think that's more useful. > Hi! V4 looks good. The only thing that I cannot explain is removing the `exprtype` variable inside ExecEvalJsonIsPredicate. We can just change its declaration to exprtype = pred->exprBaseType if i'm not mistaken? Anyway, LGTM > -- > Andrew Dunstan > EDB: https://www.enterprisedb.com -- Best regards, Kirill Reshke
On 2026-03-13 Fr 9:16 AM, Kirill Reshke wrote: > On Fri, 13 Mar 2026 at 17:30, Andrew Dunstan <andrew@dunslane.net> wrote: >> >> On 2026-03-12 Th 11:55 PM, jian he wrote: >>> Hi. >>> >>> The regression test was very verbose; I removed some of it. >>> Also polished function ExecEvalJsonIsPredicate a little bit. >> >> >> Here's a v4. I changed resultBaseType to exprBaseType - I think it's >> clearer. I also trimmed the tests a bit more, and dropped the new >> objects after testing them. The error message now shows the domain name >> rather than the underlying base type. I think that's more useful. >> > > Hi! > V4 looks good. The only thing that I cannot explain is removing the > `exprtype` variable inside ExecEvalJsonIsPredicate. We can just change > its declaration to > > exprtype = pred->exprBaseType if i'm not mistaken? Yeah, we can do that. > > Anyway, LGTM > > Thanks for reviewing. cheers andrew -- Andrew Dunstan EDB: https://www.enterprisedb.com
On 2026-03-13 Fr 9:47 AM, Andrew Dunstan wrote: > > On 2026-03-13 Fr 9:16 AM, Kirill Reshke wrote: >> On Fri, 13 Mar 2026 at 17:30, Andrew Dunstan <andrew@dunslane.net> >> wrote: >>> >>> On 2026-03-12 Th 11:55 PM, jian he wrote: >>>> Hi. >>>> >>>> The regression test was very verbose; I removed some of it. >>>> Also polished function ExecEvalJsonIsPredicate a little bit. >>> >>> >>> Here's a v4. I changed resultBaseType to exprBaseType - I think it's >>> clearer. I also trimmed the tests a bit more, and dropped the new >>> objects after testing them. The error message now shows the domain >>> name >>> rather than the underlying base type. I think that's more useful. >>> >> >> Hi! >> V4 looks good. The only thing that I cannot explain is removing the >> `exprtype` variable inside ExecEvalJsonIsPredicate. We can just change >> its declaration to >> >> exprtype = pred->exprBaseType if i'm not mistaken? > > > > Yeah, we can do that. > > >> >> Anyway, LGTM >> >> > > Committed with that change. cheers andrew -- Andrew Dunstan EDB: https://www.enterprisedb.com