From 73cc7d3b240e1d46b1996382e5735a820f8bc3f7 Mon Sep 17 00:00:00 2001 From: Tom Lane Date: Fri, 25 May 2012 14:35:37 -0400 Subject: [PATCH] Use binary search instead of brute-force scan in findNamespace(). The previous coding presented a significant bottleneck when dumping databases containing many thousands of schemas, since the total time spent searching would increase roughly as O(N^2) in the number of objects. Noted by Jeff Janes, though I rewrote his proposed patch to use the existing findObjectByOid infrastructure. Since this is a longstanding performance bug, backpatch to all supported versions. --- src/bin/pg_dump/common.c | 18 ++++++++++++++++-- src/bin/pg_dump/pg_dump.c | 35 ++++++++++++++--------------------- src/bin/pg_dump/pg_dump.h | 1 + 3 files changed, 31 insertions(+), 23 deletions(-) diff --git a/src/bin/pg_dump/common.c b/src/bin/pg_dump/common.c index 9ec180decad..42d3645a6d4 100644 --- a/src/bin/pg_dump/common.c +++ b/src/bin/pg_dump/common.c @@ -48,16 +48,19 @@ static TableInfo *tblinfo; static TypeInfo *typinfo; static FuncInfo *funinfo; static OprInfo *oprinfo; +static NamespaceInfo *nspinfo; static int numTables; static int numTypes; static int numFuncs; static int numOperators; static int numCollations; +static int numNamespaces; static DumpableObject **tblinfoindex; static DumpableObject **typinfoindex; static DumpableObject **funinfoindex; static DumpableObject **oprinfoindex; static DumpableObject **collinfoindex; +static DumpableObject **nspinfoindex; static void flagInhTables(TableInfo *tbinfo, int numTables, @@ -81,7 +84,6 @@ getSchemaData(Archive *fout, int *numTablesPtr) ExtensionInfo *extinfo; InhInfo *inhinfo; CollInfo *collinfo; - int numNamespaces; int numExtensions; int numAggregates; int numInherits; @@ -101,7 +103,8 @@ getSchemaData(Archive *fout, int *numTablesPtr) if (g_verbose) write_msg(NULL, "reading schemas\n"); - getNamespaces(fout, &numNamespaces); + nspinfo = getNamespaces(fout, &numNamespaces); + nspinfoindex = buildIndexArray(nspinfo, numNamespaces, sizeof(NamespaceInfo)); /* * getTables should be done as soon as possible, so as to minimize the @@ -732,6 +735,17 @@ findCollationByOid(Oid oid) return (CollInfo *) findObjectByOid(oid, collinfoindex, numCollations); } +/* + * findNamespaceByOid + * finds the entry (in nspinfo) of the namespace with the given oid + * returns NULL if not found + */ +NamespaceInfo * +findNamespaceByOid(Oid oid) +{ + return (NamespaceInfo *) findObjectByOid(oid, nspinfoindex, numNamespaces); +} + /* * findParentsByOid diff --git a/src/bin/pg_dump/pg_dump.c b/src/bin/pg_dump/pg_dump.c index 6e13b746289..46ab6ee2bbc 100644 --- a/src/bin/pg_dump/pg_dump.c +++ b/src/bin/pg_dump/pg_dump.c @@ -129,10 +129,6 @@ char g_comment_end[10]; static const CatalogId nilCatalogId = {0, 0}; -/* these are to avoid passing around info for findNamespace() */ -static NamespaceInfo *g_namespaces; -static int g_numNamespaces; - /* flags for various command-line long options */ static int binary_upgrade = 0; static int disable_dollar_quoting = 0; @@ -2595,8 +2591,7 @@ getNamespaces(Archive *fout, int *numNamespaces) selectDumpableNamespace(&nsinfo[1]); - g_namespaces = nsinfo; - g_numNamespaces = *numNamespaces = 2; + *numNamespaces = 2; return nsinfo; } @@ -2648,8 +2643,7 @@ getNamespaces(Archive *fout, int *numNamespaces) PQclear(res); destroyPQExpBuffer(query); - g_namespaces = nsinfo; - g_numNamespaces = *numNamespaces = ntups; + *numNamespaces = ntups; return nsinfo; } @@ -2660,35 +2654,34 @@ getNamespaces(Archive *fout, int *numNamespaces) * getNamespaces * * NB: for pre-7.3 source database, we use object OID to guess whether it's - * a system object or not. In 7.3 and later there is no guessing. + * a system object or not. In 7.3 and later there is no guessing, and we + * don't use objoid at all. */ static NamespaceInfo * findNamespace(Archive *fout, Oid nsoid, Oid objoid) { - int i; + NamespaceInfo *nsinfo; if (fout->remoteVersion >= 70300) { - for (i = 0; i < g_numNamespaces; i++) - { - NamespaceInfo *nsinfo = &g_namespaces[i]; - - if (nsoid == nsinfo->dobj.catId.oid) - return nsinfo; - } - exit_horribly(NULL, "schema with OID %u does not exist\n", nsoid); + nsinfo = findNamespaceByOid(nsoid); } else { - /* This code depends on the layout set up by getNamespaces. */ + /* This code depends on the dummy objects set up by getNamespaces. */ + Oid i; + if (objoid > g_last_builtin_oid) i = 0; /* user object */ else i = 1; /* system object */ - return &g_namespaces[i]; + nsinfo = findNamespaceByOid(i); } - return NULL; /* keep compiler quiet */ + if (nsinfo == NULL) + exit_horribly(NULL, "schema with OID %u does not exist\n", nsoid); + + return nsinfo; } /* diff --git a/src/bin/pg_dump/pg_dump.h b/src/bin/pg_dump/pg_dump.h index 5d6125c4259..f26d8d3ab66 100644 --- a/src/bin/pg_dump/pg_dump.h +++ b/src/bin/pg_dump/pg_dump.h @@ -511,6 +511,7 @@ extern TypeInfo *findTypeByOid(Oid oid); extern FuncInfo *findFuncByOid(Oid oid); extern OprInfo *findOprByOid(Oid oid); extern CollInfo *findCollationByOid(Oid oid); +extern NamespaceInfo *findNamespaceByOid(Oid oid); extern void simple_oid_list_append(SimpleOidList *list, Oid val); extern void simple_string_list_append(SimpleStringList *list, const char *val);