Re: [Proposal] Global temporary tables

čt 23. 1. 2020 v 17:28 odesílatel 曾文旌(义从) <wenjing.zwj@alibaba-inc.com>
napsal:

>
>
> 2020年1月22日 下午1:29，曾文旌(义从) <wenjing.zwj@alibaba-inc.com> 写道：
>
>
>
> 2020年1月21日 下午1:43，Pavel Stehule <pavel.stehule@gmail.com> 写道：
>
> Hi
>
> I have a free time this evening, so I will check this patch
>
> I have a one question
>
> + /* global temp table get relstats from localhash */
> + if (RELATION_IS_GLOBAL_TEMP(rel))
> + {
> + get_gtt_relstats(RelationGetRelid(rel),
> + &relpages, &reltuples, &relallvisible,
> + NULL, NULL);
> + }
> + else
> + {
> + /* coerce values in pg_class to more desirable types */
> + relpages = (BlockNumber) rel->rd_rel->relpages;
> + reltuples = (double) rel->rd_rel->reltuples;
> + relallvisible = (BlockNumber) rel->rd_rel->relallvisible;
> + }
>
> Isbn't possible to fill the rd_rel structure too, so this branching can be
> reduced?
>
> I'll make some improvements to optimize this part of the code.
>
> I'm trying to improve this part of the implementation in
> global_temporary_table_v7-pg13.patch
> Please check my patch and give me feedback.
>
>
It is looking better, still there are some strange things (I didn't tested
functionality yet)

  elog(ERROR, "invalid relpersistence: %c",
  relation->rd_rel->relpersistence);
@@ -3313,6 +3336,10 @@ RelationBuildLocalRelation(const char *relname,
  rel->rd_backend = BackendIdForTempRelations();
  rel->rd_islocaltemp = true;
  break;
+ case RELPERSISTENCE_GLOBAL_TEMP:
+ rel->rd_backend = BackendIdForTempRelations();
+ rel->rd_islocaltemp = true;
+ break;
  default:

+ rel->rd_islocaltemp = true;  <<<<<<< if this is valid, then the name of
field "rd_islocaltemp" is not probably best



regards

Pavel




>
> Thanks
>
> Wenjing
>
>
>
>
>
> Regards
>
> Pavel
>
> po 20. 1. 2020 v 17:27 odesílatel 曾文旌(义从) <wenjing.zwj@alibaba-inc.com>
> napsal:
>
>>
>>
>> > 2020年1月20日 上午1:32，Erik Rijkers <er@xs4all.nl> 写道：
>> >
>> > On 2020-01-19 18:04, 曾文旌(义从) wrote:
>> >>> 2020年1月14日 下午9:20，Pavel Stehule <pavel.stehule@gmail.com> 写道：
>> >>> út 14. 1. 2020 v 14:09 odesílatel 曾文旌(义从) <
>> wenjing.zwj@alibaba-inc.com <mailto:wenjing.zwj@alibaba-inc.com>> napsal:
>> >
>> >>> [global_temporary_table_v4-pg13.patch ]
>> >
>> > Hi,
>> >
>> > This patch doesn't quiet apply for me:
>> >
>> > patching file src/backend/access/common/reloptions.c
>> > patching file src/backend/access/gist/gistutil.c
>> > patching file src/backend/access/hash/hash.c
>> > Hunk #1 succeeded at 149 (offset 3 lines).
>> > patching file src/backend/access/heap/heapam_handler.c
>> > patching file src/backend/access/heap/vacuumlazy.c
>> > patching file src/backend/access/nbtree/nbtpage.c
>> > patching file src/backend/access/table/tableam.c
>> > patching file src/backend/access/transam/xlog.c
>> > patching file src/backend/catalog/Makefile
>> > Hunk #1 FAILED at 44.
>> > 1 out of 1 hunk FAILED -- saving rejects to file
>> src/backend/catalog/Makefile.rej
>> > [...]
>> >   (The rest applies without errors)
>> >
>> > src/backend/catalog/Makefile.rej contains:
>> >
>> > ------------------------
>> > --- src/backend/catalog/Makefile
>> > +++ src/backend/catalog/Makefile
>> > @@ -44,6 +44,8 @@ OBJS = \
>> >       storage.o \
>> >       toasting.o
>> >
>> > +OBJS += storage_gtt.o
>> > +
>> > BKIFILES = postgres.bki postgres.description postgres.shdescription
>> >
>> > include $(top_srcdir)/src/backend/common.mk
>> > ------------------------
>> >
>> > Can you have a look?
>> I updated the code and remade the patch.
>> Please give me feedback if you have any more questions.
>>
>>
>>
>>
>> >
>> >
>> > thanks,
>> >
>> > Erik Rijkers
>> >
>> >
>> >
>> >
>> >
>>
>>
>
>