Horton

Re: row filtering for logical replication

Amit Kapila <amit.kapila16@gmail.com>

View thread
From: Amit Kapila <amit.kapila16@gmail.com>
To: Peter Smith <smithpb2250@gmail.com>
Cc: "houzj.fnst@fujitsu.com" <houzj.fnst@fujitsu.com>, Ajin Cherian <itsajin@gmail.com>, Greg Nancarrow <gregn4422@gmail.com>, Dilip Kumar <dilipbalaut@gmail.com>, Euler Taveira <euler@eulerto.com>, Rahila Syed <rahilasyed90@gmail.com>, Peter Eisentraut <peter.eisentraut@enterprisedb.com>, Önder Kalacı <onderkalaci@gmail.com>, japin <japinli@hotmail.com>, Michael Paquier <michael@paquier.xyz>, David Steele <david@pgmasters.net>, Craig Ringer <craig@2ndquadrant.com>, Amit Langote <amitlangote09@gmail.com>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2021-11-15T10:31:41Z
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. Release cache tuple when no longer needed

  2. Add some additional tests for row filters in logical replication.

  3. Fix one of the tests introduced in commit 52e4f0cd47.

  4. Allow specifying row filters for logical replication of tables.

  5. Move scanint8() to numutils.c

  6. Replace Test::More plans with done_testing

  7. Reduce relcache access in WAL sender streaming logical changes

  8. Small cleanups related to PUBLICATION framework code

  9. Add a view to show the stats of subscription workers.

  10. Allow publishing the tables of schema.

  11. Doc: improve documentation of CREATE/ALTER SUBSCRIPTION.

  12. Add PublicationTable and PublicationRelInfo structs

  13. Remove unused argument "txn" in maybe_send_schema().

  14. Add prepare API support for streaming transactions in logical replication.

  15. Unify PostgresNode's new() and get_new_node() methods

  16. Use l*_node() family of functions where appropriate

  17. Add support for prepared transactions to built-in logical replication.

  18. Restore the portal-level snapshot after procedure COMMIT/ROLLBACK.

  19. Rename a parse node to be more general

  20. Remove unused column atttypmod from initial tablesync query

  21. SEARCH and CYCLE clauses

On Wed, Nov 10, 2021 at 12:36 PM Peter Smith <smithpb2250@gmail.com> wrote:
>
> On Mon, Nov 8, 2021 at 5:53 PM houzj.fnst@fujitsu.com
> <houzj.fnst@fujitsu.com> wrote:
> >
> > 3) v37-0005
> >
> > - no parse nodes of any kind other than Var, OpExpr, Const, BoolExpr, FuncExpr
> >
> > I think there could be other node type which can also be considered as simple
> > expression, for exmaple T_NullIfExpr.
>
> The current walker restrictions are from a previously agreed decision
> by Amit/Tomas [1] and from an earlier suggestion from Andres [2] to
> keep everything very simple for a first version.
>
> Yes, you are right, there might be some additional node types that
> might be fine, but at this time I don't want to add anything different
> without getting their approval to do so. Anyway, additions like this
> are all candidates for a future version of this row-filter feature.
>

I think we can consider T_NullIfExpr unless you see any problem with the same.

> >
> > Personally, I think it's natural to only check the IMMUTABLE and
> > whether-user-defined in the new function rowfilter_walker. We can keep the
> > other row-filter errors which were thrown for EXPR_KIND_PUBLICATION_WHERE in
> > the 0001 patch.
> >
>
> YMMV. IMO it is much more convenient for all the filter validations to
> be centralized just in one walker function instead of scattered all
> over the place like they were in the 0001 patch.
>

+1.

Few comments on the latest set of patches (v39*)
=======================================
0001*
1.
 ObjectAddress
-publication_add_relation(Oid pubid, PublicationRelInfo *targetrel,
+publication_add_relation(Oid pubid, PublicationRelInfo *pri,
  bool if_not_exists)
 {
  Relation rel;
  HeapTuple tup;
  Datum values[Natts_pg_publication_rel];
  bool nulls[Natts_pg_publication_rel];
- Oid relid = RelationGetRelid(targetrel->relation);
+ Relation    targetrel = pri->relation;

I don't think such a renaming (targetrel-->pri) is warranted for this
patch. If we really want something like this, we can probably do it in
a separate patch but I suggest we can do that as a separate patch.

2.
+ * The OptWhereClause (row-filter) must be stored here
+ * but it is valid only for tables. If the ColId was
+ * mistakenly not a table this will be detected later
+ * in preprocess_pubobj_list() and an error thrown.

/error thrown/error is thrown

0003*
3. In pgoutput_row_filter(), the patch is finding pub_relid when it
should already be there in RelationSyncEntry->publish_as_relid found
during get_rel_sync_entry call. Is there a reason to do this work
again?

4. I think we should add some comments in pgoutput_row_filter() as to
why we are caching the row_filter here instead of
get_rel_sync_entry()? That has been discussed multiple times so it is
better to capture that in comments.

5. Why do you need a separate variable rowfilter_valid to indicate
whether a valid row filter exists? Why exprstate is not sufficient?
Can you update comments to indicate why we need this variable
separately?

0004*
6. In rowfilter_expr_checker(), the expression tree is traversed
twice, can't we traverse it once to detect all non-allowed stuff? It
can be sometimes costly to traverse the tree multiple times especially
when the expression is complex and it doesn't seem acceptable to do so
unless there is some genuine reason for the same.

7.
+static void
+rowfilter_expr_checker(Publication *pub, Node *rfnode, Relation rel)

Keep the rel argument before whereclause as that makes the function
signature better.


With Regards,
Amit Kapila.