Horton

Re: Emitting JSON to file using COPY TO

Andrew Dunstan <andrew@dunslane.net>

View thread
From: Andrew Dunstan <andrew@dunslane.net>
To: Joe Conway <mail@joeconway.com>, Davin Shearer <davin@apache.org>, PostgreSQL-development <pgsql-hackers@postgreSQL.org>
Date: 2023-12-03T19:52:29Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  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>

On 2023-12-03 Su 14:24, Joe Conway wrote:
> On 12/3/23 11:03, Joe Conway wrote:
>> On 12/3/23 10:10, Andrew Dunstan wrote:
>>> I  realize this is just a POC, but I'd prefer to see 
>>> composite_to_json()
>>> not exposed. You could use the already public datum_to_json() instead,
>>> passing JSONTYPE_COMPOSITE and F_RECORD_OUT as the second and third
>>> arguments.
>>
>> Ok, thanks, will do
>
> Just FYI, this change does loose some performance in my not massively 
> scientific A/B/A test:
>
> 8<---------------------------
> -- with datum_to_json()
> test=# \timing
> Timing is on.
> test=# copy foo to '/tmp/buf' (format json, force_array);
> COPY 10000000
> Time: 37196.898 ms (00:37.197)
> Time: 37408.161 ms (00:37.408)
> Time: 38393.309 ms (00:38.393)
> Time: 36855.438 ms (00:36.855)
> Time: 37806.280 ms (00:37.806)
>
> Avg = 37532
>
> -- original patch
> test=# \timing
> Timing is on.
> test=# copy foo to '/tmp/buf' (format json, force_array);
> COPY 10000000
> Time: 37426.207 ms (00:37.426)
> Time: 36068.187 ms (00:36.068)
> Time: 38285.252 ms (00:38.285)
> Time: 36971.042 ms (00:36.971)
> Time: 35690.822 ms (00:35.691)
>
> Avg = 36888
>
> -- with datum_to_json()
> test=# \timing
> Timing is on.
> test=# copy foo to '/tmp/buf' (format json, force_array);
> COPY 10000000
> Time: 39083.467 ms (00:39.083)
> Time: 37249.326 ms (00:37.249)
> Time: 38529.721 ms (00:38.530)
> Time: 38704.920 ms (00:38.705)
> Time: 39001.326 ms (00:39.001)
>
> Avg = 38513
> 8<---------------------------
>
> That is somewhere in the 3% range.


I assume it's because datum_to_json() constructs a text value from which 
you then need to extract the cstring, whereas composite_to_json(), just 
gives you back the stringinfo. I guess that's a good enough reason to go 
with exposing composite_to_json().


cheers


andrew

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