Re: Emitting JSON to file using COPY TO

Andrew Dunstan <andrew@dunslane.net>

From: Andrew Dunstan <andrew@dunslane.net>
To: jian he <jian.universality@gmail.com>
Cc: Joe Conway <mail@joeconway.com>, Junwang Zhao <zhjwpku@gmail.com>, Florents Tselai <florents.tselai@gmail.com>, "Andrey M. Borodin" <x4mmm@yandex-team.ru>, Dean Rasheed <dean.a.rasheed@gmail.com>, Daniel Verite <daniel@manitou-mail.org>, Davin Shearer <davin@apache.org>, PostgreSQL-development <pgsql-hackers@postgresql.org>
Date: 2026-03-08T19:44:51Z
Lists: pgsql-hackers, pgsql-general
On 2026-03-08 Su 12:16 PM, jian he wrote:
> hi.
>
> V27-0002 is still not bullet-proof.
>
> drop table if exists t1;
> create table t1(a int);
> insert into t1 values (1);
> copy (select * from t1) to stdout json;
> {"a":1}
> WARNING:  resource was not closed: TupleDesc 0x7171d0ca3440 (18239,-1)
>
> Also see ExecAssignScanProjectionInfo->ExecConditionalAssignProjectionInfo
> So in v28-0002, I changed to
> +    /*
> +     * composite_to_json() requires a stable TupleDesc. Since the slot's
> +     * descriptor (slot->tts_tupleDescriptor) can change during the execution
> +     * of a SELECT query, we use cstate->queryDesc->tupDesc instead. This
> +     * precaution is only necessary when the output slot's TupleDesc is of
> +     * type RECORDOID.
> +     */
> +    if (!cstate->rel && slot->tts_tupleDescriptor->tdtypeid == RECORDOID)
> +        slot->tts_tupleDescriptor = cstate->queryDesc->tupDesc;


Hmm. But should we be scribbling on slot->tts_tupleDescriptor like that? 
How about something like this?:


-        * Full table or query without column list.  Ensure the slot uses
-        * cstate->tupDesc so that the datum is stamped with the right type;
-        * for queries output type is RECORDOID this must be the blessed
-        * descriptor so that composite_to_json can look it up via
-        * lookup_rowtype_tupdesc.
+        * Full table or query without column list.  For queries, the slot's
+        * TupleDesc may carry RECORDOID, which is not registered in the 
type
+        * cache and would cause composite_to_json's lookup_rowtype_tupdesc
+        * call to fail.  Build a HeapTuple stamped with the blessed
+        * descriptor so the type can be looked up correctly.
          */
         if (!cstate->rel && slot->tts_tupleDescriptor->tdtypeid == 
RECORDOID)
-           slot->tts_tupleDescriptor = cstate->queryDesc->tupDesc;
+       {
+           HeapTuple   tup;

-       rowdata = ExecFetchSlotHeapTupleDatum(slot);
+           tup = heap_form_tuple(cstate->tupDesc,
+                                 slot->tts_values,
+                                 slot->tts_isnull);
+           rowdata = HeapTupleGetDatum(tup);
+       }
+       else
+       {
+           rowdata = ExecFetchSlotHeapTupleDatum(slot);
+       }


cheers


andrew


--
Andrew Dunstan
EDB: https://www.enterprisedb.com




Commits

  1. Add option force_array for COPY JSON FORMAT

  2. json format for COPY TO

  3. introduce CopyFormat, refactor CopyFormatOptions

  4. Doc: add IDs to copy.sgml's <varlistentry> and <refsect1>

  5. Refactor COPY TO to use format callback functions.