Horton

Re: POC, WIP: OR-clause support for indexes

jian he <jian.universality@gmail.com>

View thread
From: jian he <jian.universality@gmail.com>
To: Alexander Korotkov <aekorotkov@gmail.com>
Cc: Andrei Lepikhov <lepihov@gmail.com>, Alena Rybakina <a.rybakina@postgrespro.ru>, Nikolay Shaplov <dhyan@nataraj.su>, pgsql-hackers@lists.postgresql.org, Robert Haas <robertmhaas@gmail.com>, pgsql-hackers@postgresql.org, Peter Geoghegan <pg@bowt.ie>, Marcos Pegoraro <marcos@f10.com.br>, teodor@sigaev.ru, Peter Eisentraut <peter@eisentraut.org>, Ranier Vilela <ranier.vf@gmail.com>
Date: 2024-08-26T03:12:00Z
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. Make group_similar_or_args() reorder clause list as little as possible

  2. Allow usage of match_orclause_to_indexcol() for joins

  3. Skip not SOAP-supported indexes while transforming an OR clause into SAOP

  4. Remove the wrong assertion from match_orclause_to_indexcol()

  5. Teach bitmap path generation about transforming OR-clauses to SAOP's

  6. Transform OR-clauses to SAOP's during index matching

  7. Fix the value of or_to_any_transform_limit in postgresql.conf.sample

  8. Transform OR clauses to ANY expression

  9. MergeAttributes code deduplication

  10. SEARCH and CYCLE clauses

  11. Improve estimation of OR clauses using extended statistics.

  12. Teach btree to handle ScalarArrayOpExpr quals natively.

  13. Revise collation derivation method and expression-tree representation.

  14. Instead of trying to force WHERE clauses into CNF or DNF normal form,

On Fri, Aug 23, 2024 at 8:58 PM Alexander Korotkov <aekorotkov@gmail.com> wrote:
>
based on v37.


+ {
+ /*
+ * We have only Const's.  In this case we can construct an array
+ * directly.
+ */
+ int16 typlen;
+ bool typbyval;
+ char typalign;
+ Datum   *elems;
+ int i = 0;
+ ArrayType  *arrayConst;
+
+ get_typlenbyvalalign(consttype, &typlen, &typbyval, &typalign);
+
+ elems = (Datum *) palloc(sizeof(Datum) * list_length(consts));
+ foreach(lc, consts)
+ elems[i++] = ((Const *) lfirst(lc))->constvalue;
+
+ arrayConst = construct_array(elems, i, consttype,
+ typlen, typbyval, typalign);
+ arrayNode = (Node *) makeConst(arraytype, -1, inputcollid,
+   -1, PointerGetDatum(arrayConst),
+   false, false);
+
+ pfree(elems);
+ list_free(consts);
+ }
List "consts" elements can be NULL?
I didn't find a query to trigger that.
but construct_array comments says
"elems (NULL element values are not supported)."
Do we need to check Const->constisnull for the Const node?


+ /* Construct the list of nested OR arguments */
+ for (j = group_start; j < i; j++)
+ {
+ Node   *arg = list_nth(orargs, matches[j].argindex);
+
+ rargs = lappend(rargs, arg);
+ if (IsA(arg, RestrictInfo))
+ args = lappend(args, ((RestrictInfo *) arg)->clause);
+ else
+ args = lappend(args, arg);
+ }
the ELSE branch never reached?



+ /* Construct the nested OR and wrap it with RestrictInfo */
+ *subrinfo = *rinfo;
+ subrinfo->clause = make_orclause(args);
+ subrinfo->orclause = make_orclause(rargs);
+ result = lappend(result, subrinfo);
should we use memcpy instead of " *subrinfo = *rinfo;"?



+ /* Sort clauses to make similar clauses go together */
+ pg_qsort(matches, n, sizeof(OrArgIndexMatch), or_arg_index_match_cmp);
Should we use qsort?
since comments in pg_qsort:
/*
 * Callers should use the qsort() macro defined below instead of calling
 * pg_qsort() directly.
 */



+/*
+ * Data structure representing information about OR-clause argument and its
+ * matching index key.  Used for grouping of similar OR-clause arguments in
+ * group_similar_or_args().
+ */
+typedef struct
+{
+ int indexnum; /* index of the matching index */
+ int colnum; /* index of the matching column */
+ Oid opno; /* OID of the OpClause operator */
+ Oid inputcollid; /* OID of the OpClause input collation */
+ int argindex; /* index of the clause in the list of
+ * arguments */
+} OrArgIndexMatch;

I am not 100% sure about the comments.
indexnum:  index of the matching index reside in rel->indexlist that
matches (counting from 0)
colnum: the column number of the matched index (counting from 0)