diff options
| author | Tom Lane <tgl@sss.pgh.pa.us> | 2021-08-31 13:53:33 -0400 |
|---|---|---|
| committer | Tom Lane <tgl@sss.pgh.pa.us> | 2021-08-31 13:53:50 -0400 |
| commit | 2f1ed9d98c38a62ffa4b0a6589c9fcc529ae0883 (patch) | |
| tree | 30dba6d0503345d5e9cf4bcc00be52f29e81d4d1 /src | |
| parent | 5f8dd5dc1701348bce659b9689d085e011399d61 (diff) | |
| download | postgresql-2f1ed9d98c38a62ffa4b0a6589c9fcc529ae0883.tar.gz postgresql-2f1ed9d98c38a62ffa4b0a6589c9fcc529ae0883.zip | |
Cache the results of format_type() queries in pg_dump.
There's long been a "TODO: there might be some value in caching
the results" annotation on pg_dump's getFormattedTypeName function;
but we hadn't gotten around to checking what it was costing us to
repetitively look up type names. It turns out that when dumping the
current regression database, about 10% of the total number of queries
issued are duplicative format_type() queries. However, Hubert Depesz
Lubaczewski reported a not-unusual case where these account for over
half of the queries issued by pg_dump. Individually these queries
aren't expensive, but when network lag is a factor, they add up to a
problem. We can very easily add some caching to getFormattedTypeName
to solve it.
Since this is such a simple fix and can have a visible performance
benefit, back-patch to all supported branches.
Discussion: https://postgr.es/m/20210826084430.GA26282@depesz.com
Diffstat (limited to 'src')
| -rw-r--r-- | src/bin/pg_dump/pg_dump.c | 13 | ||||
| -rw-r--r-- | src/bin/pg_dump/pg_dump.h | 6 |
2 files changed, 15 insertions, 4 deletions
diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c index d181946ad7a..9aae8f8f55c 100644 --- a/src/bin/pg_dump/pg_dump.c +++ b/src/bin/pg_dump/pg_dump.c @@ -4981,6 +4981,7 @@ getTypes(Archive *fout, int *numTypes) tyinfo[i].dobj.namespace = findNamespace(fout, atooid(PQgetvalue(res, i, i_typnamespace))); + tyinfo[i].ftypname = NULL; /* may get filled later */ tyinfo[i].rolname = pg_strdup(PQgetvalue(res, i, i_rolname)); tyinfo[i].typacl = pg_strdup(PQgetvalue(res, i, i_typacl)); tyinfo[i].rtypacl = pg_strdup(PQgetvalue(res, i, i_rtypacl)); @@ -18611,12 +18612,11 @@ findDumpableDependencies(ArchiveHandle *AH, DumpableObject *dobj, * * This does not guarantee to schema-qualify the output, so it should not * be used to create the target object name for CREATE or ALTER commands. - * - * TODO: there might be some value in caching the results. */ static char * getFormattedTypeName(Archive *fout, Oid oid, OidOptions opts) { + TypeInfo *typeInfo; char *result; PQExpBuffer query; PGresult *res; @@ -18633,6 +18633,11 @@ getFormattedTypeName(Archive *fout, Oid oid, OidOptions opts) return pg_strdup("NONE"); } + /* see if we have the result cached in the type's TypeInfo record */ + typeInfo = findTypeByOid(oid); + if (typeInfo && typeInfo->ftypname) + return pg_strdup(typeInfo->ftypname); + query = createPQExpBuffer(); appendPQExpBuffer(query, "SELECT pg_catalog.format_type('%u'::pg_catalog.oid, NULL)", oid); @@ -18645,6 +18650,10 @@ getFormattedTypeName(Archive *fout, Oid oid, OidOptions opts) PQclear(res); destroyPQExpBuffer(query); + /* cache a copy for later requests */ + if (typeInfo) + typeInfo->ftypname = pg_strdup(result); + return result; } diff --git a/src/bin/pg_dump/pg_dump.h b/src/bin/pg_dump/pg_dump.h index 162a524303d..0a29919c13c 100644 --- a/src/bin/pg_dump/pg_dump.h +++ b/src/bin/pg_dump/pg_dump.h @@ -167,9 +167,11 @@ typedef struct _typeInfo DumpableObject dobj; /* - * Note: dobj.name is the pg_type.typname entry. format_type() might - * produce something different than typname + * Note: dobj.name is the raw pg_type.typname entry. ftypname is the + * result of format_type(), which will be quoted if needed, and might be + * schema-qualified too. */ + char *ftypname; char *rolname; /* name of owner, or empty string */ char *typacl; char *rtypacl; |