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

diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c
index fc2fc5891e20366dc9b06b5580ec65545696afdc..f108ecece64bf3135626b6993ca9e9662b2d2f6a 100644 (file)
--- a/src/bin/pg_dump/pg_dump.c
+++ b/src/bin/pg_dump/pg_dump.c
@@ -5067,6 +5067,7 @@ 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));
@@ -18735,12 +18736,11 @@ 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;
@@ -18753,6 +18753,11 @@ 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);
@@ -18765,6 +18770,10 @@ 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 0224911b502fe1ff184a1bfce6a5c1110e80926c..7d41064b8a1f8d136af2835a17cc2328d50df578 100644 (file)
--- a/src/bin/pg_dump/pg_dump.h
+++ b/src/bin/pg_dump/pg_dump.h
@@ -163,9 +163,11 @@ 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;