/*------------------------------------------------------------------------- * * be-dumpdata.c-- * support for collection of returned tuples from an internal * PQ call into a backend buffer. * * Copyright (c) 1994, Regents of the University of California * * * IDENTIFICATION * $Header: /usr/local/cvsroot/pgsql/src/backend/libpq/be-dumpdata.c,v 1.9 1997/09/12 04:07:50 momjian Exp $ * *------------------------------------------------------------------------- */ /* * INTERFACE ROUTINES * be_portalinit - initialize backend portal administration * be_portalpush - add a portal to the top of the portal stack * be_portalpop - remove portal on the top of the stack & return it * be_currentportal - return the top portal on the portal stack * be_newportal - return a new portal. * be_portalinit - initialize backend portal expected to hold results. * be_printtup - add a tuple to a backend portal * * NOTES * Since backend user-defined operators can call queries * which in turn call user-defined operators can call queries... * we have to keep track of portals on a stack. BeginCommand() * puts portals on the stack and the PQ functions remove them. * */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* ---------------- * backend portal stack for recursive PQexec calls * ---------------- */ static Dllist *be_portalstack; /* ---------------- * be_portalinit - initialize backend portal administration * * This is called once from InitPostgres() to initialize * the portal stack. * ---------------- */ void be_portalinit(void) { be_portalstack = DLNewList(); } /* ---------------- * be_portalpush - add a portal to the top of the portal stack * * used by BeginCommand() * ---------------- */ void be_portalpush(PortalEntry *entry) { DLAddTail(be_portalstack, DLNewElem(entry)); } /* ---------------- * be_portalpop - remove the portal on the top of the stack & return it * * used by PQexec() * ---------------- */ PortalEntry * be_portalpop(void) { PortalEntry *p; Dlelem *elt; elt = DLRemTail(be_portalstack); p = (elt ? (PortalEntry *) DLE_VAL(elt) : NULL); DLFreeElem(elt); return p; } /* ---------------- * be_currentportal - return the top portal on the portal stack * * used by be_printtup() * ---------------- */ PortalEntry * be_currentportal(void) { Dlelem *elt; elt = DLGetTail(be_portalstack); return (elt ? (PortalEntry *) DLE_VAL(elt) : NULL); } /* ---------------- * be_newportal - return a new portal. * * If the user-defined function does not specify a portal name, * we generate a unique one. Names are generated from a combination * of a postgres oid and an integer counter which is incremented * every time we ask for a local portal. * * used by BeginCommand() * ---------------- */ static Oid be_portaloid; static u_int be_portalcnt = 0; PortalEntry * be_newportal(void) { PortalEntry *entry; char buf[PortalNameLength]; /* ---------------- * generate a new name * ---------------- */ if (be_portalcnt == 0) be_portaloid = newoid(); be_portalcnt++; sprintf(buf, "be_%d_%d", be_portaloid, be_portalcnt); /* ---------------- * initialize the new portal entry and keep track * of the current memory context for be_printtup(). * This is important - otherwise whatever we allocate * will go away and the contents of the portal after * PQexec() returns will be meaningless. * ---------------- */ entry = pbuf_setup(buf); entry->portalcxt = (Pointer) CurrentMemoryContext; return entry; } /* ---------------- * be_typeinit - initialize backend portal expected to hold * query results. * * used by BeginCommand() * ---------------- */ void be_typeinit(PortalEntry *entry, TupleDesc tupDesc, int natts) { PortalBuffer *portal; GroupBuffer *group; int i; AttributeTupleForm *attrs = tupDesc->attrs; /* ---------------- * add a new portal group to the portal * ---------------- */ portal = entry->portal; portal->no_groups++; portal->groups = group = pbuf_addGroup(portal); group->no_fields = natts; /* ---------------- * initialize portal group type info * ---------------- */ if (natts > 0) { group->types = pbuf_addTypes(natts); for (i = 0; i < natts; ++i) { strncpy(group->types[i].name, attrs[i]->attname.data, NAMEDATALEN); group->types[i].adtid = attrs[i]->atttypid; group->types[i].adtsize = attrs[i]->attlen; } } } /* ---------------- * be_printtup - add a tuple to a backend portal * * used indirectly by ExecRetrieve() * * This code is pretty much copied from printtup(), dump_type() * and dump_data(). -cim 2/12/91 * ---------------- */ void be_printtup(HeapTuple tuple, TupleDesc typeinfo) { int i; Datum attr; bool isnull; Oid typoutput; PortalEntry *entry = NULL; PortalBuffer *portal = NULL; GroupBuffer *group = NULL; TupleBlock *tuples = NULL; char **values; int *lengths; MemoryContext savecxt; /* ---------------- * get the current portal and group * ---------------- */ entry = be_currentportal(); portal = entry->portal; group = portal->groups; /* ---------------- * switch to the portal's memory context so that * the tuples we allocate are returned to the user. * ---------------- */ savecxt = MemoryContextSwitchTo((MemoryContext) entry->portalcxt); /* ---------------- * If no tuple block yet, allocate one. * If the current block is full, allocate another one. * ---------------- */ if (group->tuples == NULL) { tuples = group->tuples = pbuf_addTuples(); tuples->tuple_index = 0; } else { tuples = group->tuples; /* walk to the end of the linked list of TupleBlocks */ while (tuples->next) tuples = tuples->next; /* * now, tuples is the last TupleBlock, check to see if it is full. * If so, allocate a new TupleBlock and add it to the end of the * chain */ if (tuples->tuple_index == TupleBlockSize) { tuples->next = pbuf_addTuples(); tuples = tuples->next; tuples->tuple_index = 0; } } /* ---------------- * Allocate space for a tuple. * ---------------- */ tuples->values[tuples->tuple_index] = pbuf_addTuple(tuple->t_natts); tuples->lengths[tuples->tuple_index] = pbuf_addTupleValueLengths(tuple->t_natts); /* ---------------- * copy printable representations of the tuple's attributes * to the portal. * * This seems silly, because the user's function which is calling * PQexec() or PQfn() will probably just convert this back into the * internal form anyways, but the point here is to provide a uniform * libpq interface and this is how the fe libpq interface currently * works. Pretty soon we'll have to add code to let the fe or be * select the desired data representation and then deal with that. * This should not be too hard, as there already exist typrecieve() * and typsend() procedures for user-defined types (see pg_type.h) * -cim 2/11/91 * ---------------- */ values = tuples->values[tuples->tuple_index]; lengths = tuples->lengths[tuples->tuple_index]; for (i = 0; i < tuple->t_natts; i++) { attr = heap_getattr(tuple, InvalidBuffer, i + 1, typeinfo, &isnull); typoutput = typtoout((Oid) typeinfo->attrs[i]->atttypid); lengths[i] = typeinfo->attrs[i]->attlen; if (lengths[i] == -1) /* variable length attribute */ if (!isnull) lengths[i] = VARSIZE(attr) - VARHDRSZ; else lengths[i] = 0; if (!isnull && OidIsValid(typoutput)) { values[i] = fmgr(typoutput, attr, gettypelem(typeinfo->attrs[i]->atttypid)); } else values[i] = NULL; } /* ---------------- * increment tuple group counters * ---------------- */ portal->no_tuples++; group->no_tuples++; tuples->tuple_index++; /* ---------------- * return to the original memory context * ---------------- */ MemoryContextSwitchTo(savecxt); }