/*
* Determine size of a large object
*
* NOTE: LOs can contain gaps, just like Unix files. We actually return
* the offset of the last byte + 1.
*/
static uint32
inv_getsize(LargeObjectDesc *obj_desc)
{
uint32 lastbyte = 0;
ScanKeyData skey[1];
SysScanDesc sd;
HeapTuple tuple;
Assert(PointerIsValid(obj_desc));
open_lo_relation();
ScanKeyInit(&skey[0],
Anum_pg_largeobject_loid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(obj_desc->id));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
obj_desc->snapshot, 1, skey);
/*
* Because the pg_largeobject index is on both loid and pageno, but we
* constrain only loid, a backwards scan should visit all pages of the
* large object in reverse pageno order. So, it's sufficient to examine
* the first valid tuple (== last valid page).
*/
tuple = systable_getnext_ordered(sd, BackwardScanDirection);
if (HeapTupleIsValid(tuple))
{
Form_pg_largeobject data;
bytea *datafield;
bool pfreeit;
if (HeapTupleHasNulls(tuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
data = (Form_pg_largeobject) GETSTRUCT(tuple);
datafield = &(data->data); /* see note at top of file */
pfreeit = false;
if (VARATT_IS_EXTENDED(datafield))
{
datafield = (bytea *)
heap_tuple_untoast_attr((struct varlena *) datafield);
pfreeit = true;
}
lastbyte = data->pageno * LOBLKSIZE + getbytealen(datafield);
if (pfreeit)
pfree(datafield);
}
systable_endscan_ordered(sd);
return lastbyte;
}
/* ----------
* toast_delete_datum -
*
* Delete a single external stored value.
* ----------
*/
static void
toast_delete_datum(Relation rel, Datum value)
{
struct varlena *attr = (struct varlena *) DatumGetPointer(value);
struct varatt_external toast_pointer;
Relation toastrel;
Relation toastidx;
ScanKeyData toastkey;
SysScanDesc toastscan;
HeapTuple toasttup;
if (!VARATT_IS_EXTERNAL(attr))
return;
/* Must copy to access aligned fields */
VARATT_EXTERNAL_GET_POINTER(toast_pointer, attr);
/*
* Open the toast relation and its index
*/
toastrel = heap_open(toast_pointer.va_toastrelid, RowExclusiveLock);
toastidx = index_open(toastrel->rd_rel->reltoastidxid, RowExclusiveLock);
/*
* Setup a scan key to find chunks with matching va_valueid
*/
ScanKeyInit(&toastkey,
(AttrNumber) 1,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(toast_pointer.va_valueid));
/*
* Find all the chunks. (We don't actually care whether we see them in
* sequence or not, but since we've already locked the index we might as
* well use systable_beginscan_ordered.)
*/
toastscan = systable_beginscan_ordered(toastrel, toastidx,
SnapshotToast, 1, &toastkey);
while ((toasttup = systable_getnext_ordered(toastscan, ForwardScanDirection)) != NULL)
{
/*
* Have a chunk, delete it
*/
simple_heap_delete(toastrel, &toasttup->t_self);
}
/*
* End scan and close relations
*/
systable_endscan_ordered(toastscan);
index_close(toastidx, RowExclusiveLock);
heap_close(toastrel, RowExclusiveLock);
}
/*
* Determine size of a large object
*
* NOTE: LOs can contain gaps, just like Unix files. We actually return
* the offset of the last byte + 1.
*/
static uint32
inv_getsize(struct lobj *obj_desc)
{
uint32 lastbyte = 0;
struct scankey skey[1];
struct sys_scan* sd;
struct heap_tuple* tuple;
ASSERT(PTR_VALID(obj_desc));
open_lo_relation();
scankey_init(&skey[0], Anum_pg_largeobject_loid, BT_EQ_STRAT_NR, F_OIDEQ, OID_TO_D(obj_desc->id));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r, obj_desc->snapshot, 1, skey);
/*
* Because the pg_largeobject index is on both loid and pageno, but we
* constrain only loid, a backwards scan should visit all pages of the
* large object in reverse pageno order. So, it's sufficient to examine
* the first valid tuple (== last valid page).
*/
tuple = systable_getnext_ordered(sd, BACKWARD_SCANDIR);
if (HT_VALID(tuple)) {
Form_pg_largeobject data;
bytea* datafield;
bool pfreeit;
if (HT_HAS_NULLS(tuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
data = (Form_pg_largeobject) GET_STRUCT(tuple);
datafield = &(data->data); /* see note at top of file */
pfreeit = false;
if (VLA_EXTENDED(datafield)) {
datafield = (bytea *) heap_tuple_untoast_attr((struct vla *) datafield);
pfreeit = true;
}
lastbyte = data->pageno * LO_BLK_SIZE + getbytealen(datafield);
if (pfreeit)
pfree(datafield);
}
systable_endscan_ordered(sd);
return lastbyte;
}
/*
* enum_endpoint: common code for enum_first/enum_last
*/
static Oid
enum_endpoint(Oid enumtypoid, ScanDirection direction)
{
Relation enum_rel;
Relation enum_idx;
SysScanDesc enum_scan;
HeapTuple enum_tuple;
ScanKeyData skey;
Oid minmax;
/*
* Find the first/last enum member using pg_enum_typid_sortorder_index.
* Note we must not use the syscache. See comments for RenumberEnumType
* in catalog/pg_enum.c for more info.
*/
ScanKeyInit(&skey,
Anum_pg_enum_enumtypid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(enumtypoid));
enum_rel = heap_open(EnumRelationId, AccessShareLock);
enum_idx = index_open(EnumTypIdSortOrderIndexId, AccessShareLock);
enum_scan = systable_beginscan_ordered(enum_rel, enum_idx, NULL,
1, &skey);
enum_tuple = systable_getnext_ordered(enum_scan, direction);
if (HeapTupleIsValid(enum_tuple))
{
/* check it's safe to use in SQL */
check_safe_enum_use(enum_tuple);
minmax = HeapTupleGetOid(enum_tuple);
}
else
{
/* should only happen with an empty enum */
minmax = InvalidOid;
}
systable_endscan_ordered(enum_scan);
index_close(enum_idx, AccessShareLock);
heap_close(enum_rel, AccessShareLock);
return minmax;
}
/*
* enum_endpoint: common code for enum_first/enum_last
*/
static Oid
enum_endpoint(Oid enumtypoid, ScanDirection direction)
{
Relation enum_rel;
Relation enum_idx;
SysScanDesc enum_scan;
HeapTuple enum_tuple;
ScanKeyData skey;
Oid minmax;
/*
* Find the first/last enum member using pg_enum_typid_sortorder_index.
* Note we must not use the syscache, and must use an MVCC snapshot here.
* See comments for RenumberEnumType in catalog/pg_enum.c for more info.
*/
ScanKeyInit(&skey,
Anum_pg_enum_enumtypid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(enumtypoid));
enum_rel = heap_open(EnumRelationId, AccessShareLock);
enum_idx = index_open(EnumTypIdSortOrderIndexId, AccessShareLock);
enum_scan = systable_beginscan_ordered(enum_rel, enum_idx,
GetTransactionSnapshot(),
1, &skey);
enum_tuple = systable_getnext_ordered(enum_scan, direction);
if (HeapTupleIsValid(enum_tuple))
minmax = HeapTupleGetOid(enum_tuple);
else
minmax = InvalidOid;
systable_endscan_ordered(enum_scan);
index_close(enum_idx, AccessShareLock);
heap_close(enum_rel, AccessShareLock);
return minmax;
}
static ArrayType *
enum_range_internal(Oid enumtypoid, Oid lower, Oid upper)
{
ArrayType *result;
Relation enum_rel;
Relation enum_idx;
SysScanDesc enum_scan;
HeapTuple enum_tuple;
ScanKeyData skey;
Datum *elems;
int max,
cnt;
bool left_found;
/*
* Scan the enum members in order using pg_enum_typid_sortorder_index.
* Note we must not use the syscache. See comments for RenumberEnumType
* in catalog/pg_enum.c for more info.
*/
ScanKeyInit(&skey,
Anum_pg_enum_enumtypid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(enumtypoid));
enum_rel = heap_open(EnumRelationId, AccessShareLock);
enum_idx = index_open(EnumTypIdSortOrderIndexId, AccessShareLock);
enum_scan = systable_beginscan_ordered(enum_rel, enum_idx, NULL, 1, &skey);
max = 64;
elems = (Datum *) palloc(max * sizeof(Datum));
cnt = 0;
left_found = !OidIsValid(lower);
while (HeapTupleIsValid(enum_tuple = systable_getnext_ordered(enum_scan, ForwardScanDirection)))
{
Oid enum_oid = HeapTupleGetOid(enum_tuple);
if (!left_found && lower == enum_oid)
left_found = true;
if (left_found)
{
/* check it's safe to use in SQL */
check_safe_enum_use(enum_tuple);
if (cnt >= max)
{
max *= 2;
elems = (Datum *) repalloc(elems, max * sizeof(Datum));
}
elems[cnt++] = ObjectIdGetDatum(enum_oid);
}
if (OidIsValid(upper) && upper == enum_oid)
break;
}
systable_endscan_ordered(enum_scan);
index_close(enum_idx, AccessShareLock);
heap_close(enum_rel, AccessShareLock);
/* and build the result array */
/* note this hardwires some details about the representation of Oid */
result = construct_array(elems, cnt, enumtypoid, sizeof(Oid), true, 'i');
pfree(elems);
return result;
}
void
inv_truncate(LargeObjectDesc *obj_desc, int len)
{
int32 pageno = (int32) (len / LOBLKSIZE);
int off;
ScanKeyData skey[2];
SysScanDesc sd;
HeapTuple oldtuple;
Form_pg_largeobject olddata;
struct
{
bytea hdr;
char data[LOBLKSIZE]; /* make struct big enough */
int32 align_it; /* ensure struct is aligned well enough */
} workbuf;
char *workb = VARDATA(&workbuf.hdr);
HeapTuple newtup;
Datum values[Natts_pg_largeobject];
bool nulls[Natts_pg_largeobject];
bool replace[Natts_pg_largeobject];
CatalogIndexState indstate;
Assert(PointerIsValid(obj_desc));
/* enforce writability because snapshot is probably wrong otherwise */
if ((obj_desc->flags & IFS_WRLOCK) == 0)
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("large object %u was not opened for writing",
obj_desc->id)));
/* check existence of the target largeobject */
if (!LargeObjectExists(obj_desc->id))
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("large object %u was already dropped", obj_desc->id)));
open_lo_relation();
indstate = CatalogOpenIndexes(lo_heap_r);
/*
* Set up to find all pages with desired loid and pageno >= target
*/
ScanKeyInit(&skey[0],
Anum_pg_largeobject_loid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(obj_desc->id));
ScanKeyInit(&skey[1],
Anum_pg_largeobject_pageno,
BTGreaterEqualStrategyNumber, F_INT4GE,
Int32GetDatum(pageno));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
obj_desc->snapshot, 2, skey);
/*
* If possible, get the page the truncation point is in. The truncation
* point may be beyond the end of the LO or in a hole.
*/
olddata = NULL;
if ((oldtuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
{
if (HeapTupleHasNulls(oldtuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
olddata = (Form_pg_largeobject) GETSTRUCT(oldtuple);
Assert(olddata->pageno >= pageno);
}
/*
* If we found the page of the truncation point we need to truncate the
* data in it. Otherwise if we're in a hole, we need to create a page to
* mark the end of data.
*/
if (olddata != NULL && olddata->pageno == pageno)
{
/* First, load old data into workbuf */
bytea *datafield = &(olddata->data); /* see note at top of
* file */
bool pfreeit = false;
int pagelen;
if (VARATT_IS_EXTENDED(datafield))
{
datafield = (bytea *)
heap_tuple_untoast_attr((struct varlena *) datafield);
pfreeit = true;
}
pagelen = getbytealen(datafield);
Assert(pagelen <= LOBLKSIZE);
memcpy(workb, VARDATA(datafield), pagelen);
if (pfreeit)
pfree(datafield);
/*
* Fill any hole
*/
off = len % LOBLKSIZE;
if (off > pagelen)
MemSet(workb + pagelen, 0, off - pagelen);
/* compute length of new page */
SET_VARSIZE(&workbuf.hdr, off + VARHDRSZ);
/*
* Form and insert updated tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
memset(replace, false, sizeof(replace));
values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
replace[Anum_pg_largeobject_data - 1] = true;
newtup = heap_modify_tuple(oldtuple, RelationGetDescr(lo_heap_r),
values, nulls, replace);
simple_heap_update(lo_heap_r, &newtup->t_self, newtup);
CatalogIndexInsert(indstate, newtup);
heap_freetuple(newtup);
}
else
{
/*
* If the first page we found was after the truncation point, we're in
* a hole that we'll fill, but we need to delete the later page
* because the loop below won't visit it again.
*/
if (olddata != NULL)
{
Assert(olddata->pageno > pageno);
simple_heap_delete(lo_heap_r, &oldtuple->t_self);
}
/*
* Write a brand new page.
*
* Fill the hole up to the truncation point
*/
off = len % LOBLKSIZE;
if (off > 0)
MemSet(workb, 0, off);
/* compute length of new page */
SET_VARSIZE(&workbuf.hdr, off + VARHDRSZ);
/*
* Form and insert new tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
values[Anum_pg_largeobject_loid - 1] = ObjectIdGetDatum(obj_desc->id);
values[Anum_pg_largeobject_pageno - 1] = Int32GetDatum(pageno);
values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
newtup = heap_form_tuple(lo_heap_r->rd_att, values, nulls);
simple_heap_insert(lo_heap_r, newtup);
CatalogIndexInsert(indstate, newtup);
heap_freetuple(newtup);
}
/*
* Delete any pages after the truncation point. If the initial search
* didn't find a page, then of course there's nothing more to do.
*/
if (olddata != NULL)
{
while ((oldtuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
{
simple_heap_delete(lo_heap_r, &oldtuple->t_self);
}
}
systable_endscan_ordered(sd);
CatalogCloseIndexes(indstate);
/*
* Advance command counter so that tuple updates will be seen by later
* large-object operations in this transaction.
*/
CommandCounterIncrement();
}
int
inv_write(LargeObjectDesc *obj_desc, const char *buf, int nbytes)
{
int nwritten = 0;
int n;
int off;
int len;
int32 pageno = (int32) (obj_desc->offset / LOBLKSIZE);
ScanKeyData skey[2];
SysScanDesc sd;
HeapTuple oldtuple;
Form_pg_largeobject olddata;
bool neednextpage;
bytea *datafield;
bool pfreeit;
struct
{
bytea hdr;
char data[LOBLKSIZE]; /* make struct big enough */
int32 align_it; /* ensure struct is aligned well enough */
} workbuf;
char *workb = VARDATA(&workbuf.hdr);
HeapTuple newtup;
Datum values[Natts_pg_largeobject];
bool nulls[Natts_pg_largeobject];
bool replace[Natts_pg_largeobject];
CatalogIndexState indstate;
Assert(PointerIsValid(obj_desc));
Assert(buf != NULL);
/* enforce writability because snapshot is probably wrong otherwise */
if ((obj_desc->flags & IFS_WRLOCK) == 0)
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("large object %u was not opened for writing",
obj_desc->id)));
/* check existence of the target largeobject */
if (!LargeObjectExists(obj_desc->id))
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("large object %u was already dropped", obj_desc->id)));
if (nbytes <= 0)
return 0;
open_lo_relation();
indstate = CatalogOpenIndexes(lo_heap_r);
ScanKeyInit(&skey[0],
Anum_pg_largeobject_loid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(obj_desc->id));
ScanKeyInit(&skey[1],
Anum_pg_largeobject_pageno,
BTGreaterEqualStrategyNumber, F_INT4GE,
Int32GetDatum(pageno));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
obj_desc->snapshot, 2, skey);
oldtuple = NULL;
olddata = NULL;
neednextpage = true;
while (nwritten < nbytes)
{
/*
* If possible, get next pre-existing page of the LO. We expect the
* indexscan will deliver these in order --- but there may be holes.
*/
if (neednextpage)
{
if ((oldtuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
{
if (HeapTupleHasNulls(oldtuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
olddata = (Form_pg_largeobject) GETSTRUCT(oldtuple);
Assert(olddata->pageno >= pageno);
}
neednextpage = false;
}
/*
* If we have a pre-existing page, see if it is the page we want to
* write, or a later one.
*/
if (olddata != NULL && olddata->pageno == pageno)
{
/*
* Update an existing page with fresh data.
*
* First, load old data into workbuf
*/
datafield = &(olddata->data); /* see note at top of file */
pfreeit = false;
if (VARATT_IS_EXTENDED(datafield))
{
datafield = (bytea *)
heap_tuple_untoast_attr((struct varlena *) datafield);
pfreeit = true;
}
len = getbytealen(datafield);
Assert(len <= LOBLKSIZE);
memcpy(workb, VARDATA(datafield), len);
if (pfreeit)
pfree(datafield);
/*
* Fill any hole
*/
off = (int) (obj_desc->offset % LOBLKSIZE);
if (off > len)
MemSet(workb + len, 0, off - len);
/*
* Insert appropriate portion of new data
*/
n = LOBLKSIZE - off;
n = (n <= (nbytes - nwritten)) ? n : (nbytes - nwritten);
memcpy(workb + off, buf + nwritten, n);
nwritten += n;
obj_desc->offset += n;
off += n;
/* compute valid length of new page */
len = (len >= off) ? len : off;
SET_VARSIZE(&workbuf.hdr, len + VARHDRSZ);
/*
* Form and insert updated tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
memset(replace, false, sizeof(replace));
values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
replace[Anum_pg_largeobject_data - 1] = true;
newtup = heap_modify_tuple(oldtuple, RelationGetDescr(lo_heap_r),
values, nulls, replace);
simple_heap_update(lo_heap_r, &newtup->t_self, newtup);
CatalogIndexInsert(indstate, newtup);
heap_freetuple(newtup);
/*
* We're done with this old page.
*/
oldtuple = NULL;
olddata = NULL;
neednextpage = true;
}
else
{
/*
* Write a brand new page.
*
* First, fill any hole
*/
off = (int) (obj_desc->offset % LOBLKSIZE);
if (off > 0)
MemSet(workb, 0, off);
/*
* Insert appropriate portion of new data
*/
n = LOBLKSIZE - off;
n = (n <= (nbytes - nwritten)) ? n : (nbytes - nwritten);
memcpy(workb + off, buf + nwritten, n);
nwritten += n;
obj_desc->offset += n;
/* compute valid length of new page */
len = off + n;
SET_VARSIZE(&workbuf.hdr, len + VARHDRSZ);
/*
* Form and insert updated tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
values[Anum_pg_largeobject_loid - 1] = ObjectIdGetDatum(obj_desc->id);
values[Anum_pg_largeobject_pageno - 1] = Int32GetDatum(pageno);
values[Anum_pg_largeobject_data - 1] = PointerGetDatum(&workbuf);
newtup = heap_form_tuple(lo_heap_r->rd_att, values, nulls);
simple_heap_insert(lo_heap_r, newtup);
CatalogIndexInsert(indstate, newtup);
heap_freetuple(newtup);
}
pageno++;
}
systable_endscan_ordered(sd);
CatalogCloseIndexes(indstate);
/*
* Advance command counter so that my tuple updates will be seen by later
* large-object operations in this transaction.
*/
CommandCounterIncrement();
return nwritten;
}
/*
* Fetch configuration cache entry
*/
TSConfigCacheEntry *
lookup_ts_config_cache(Oid cfgId)
{
TSConfigCacheEntry *entry;
if (TSConfigCacheHash == NULL)
{
/* First time through: initialize the hash table */
init_ts_config_cache();
}
/* Check single-entry cache */
if (lastUsedConfig && lastUsedConfig->cfgId == cfgId &&
lastUsedConfig->isvalid)
return lastUsedConfig;
/* Try to look up an existing entry */
entry = (TSConfigCacheEntry *) hash_search(TSConfigCacheHash,
(void *) &cfgId,
HASH_FIND, NULL);
if (entry == NULL || !entry->isvalid)
{
/*
* If we didn't find one, we want to make one. But first look up the
* object to be sure the OID is real.
*/
HeapTuple tp;
Form_pg_ts_config cfg;
Relation maprel;
Relation mapidx;
ScanKeyData mapskey;
SysScanDesc mapscan;
HeapTuple maptup;
ListDictionary maplists[MAXTOKENTYPE + 1];
Oid mapdicts[MAXDICTSPERTT];
int maxtokentype;
int ndicts;
int i;
tp = SearchSysCache1(TSCONFIGOID, ObjectIdGetDatum(cfgId));
if (!HeapTupleIsValid(tp))
elog(ERROR, "cache lookup failed for text search configuration %u",
cfgId);
cfg = (Form_pg_ts_config) GETSTRUCT(tp);
/*
* Sanity checks
*/
if (!OidIsValid(cfg->cfgparser))
elog(ERROR, "text search configuration %u has no parser", cfgId);
if (entry == NULL)
{
bool found;
/* Now make the cache entry */
entry = (TSConfigCacheEntry *)
hash_search(TSConfigCacheHash,
(void *) &cfgId,
HASH_ENTER, &found);
Assert(!found); /* it wasn't there a moment ago */
}
else
{
/* Cleanup old contents */
if (entry->map)
{
for (i = 0; i < entry->lenmap; i++)
if (entry->map[i].dictIds)
pfree(entry->map[i].dictIds);
pfree(entry->map);
}
}
MemSet(entry, 0, sizeof(TSConfigCacheEntry));
entry->cfgId = cfgId;
entry->prsId = cfg->cfgparser;
ReleaseSysCache(tp);
/*
* Scan pg_ts_config_map to gather dictionary list for each token type
*
* Because the index is on (mapcfg, maptokentype, mapseqno), we will
* see the entries in maptokentype order, and in mapseqno order for
* each token type, even though we didn't explicitly ask for that.
*/
MemSet(maplists, 0, sizeof(maplists));
maxtokentype = 0;
ndicts = 0;
ScanKeyInit(&mapskey,
Anum_pg_ts_config_map_mapcfg,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(cfgId));
maprel = heap_open(TSConfigMapRelationId, AccessShareLock);
mapidx = index_open(TSConfigMapIndexId, AccessShareLock);
mapscan = systable_beginscan_ordered(maprel, mapidx,
NULL, 1, &mapskey);
while ((maptup = systable_getnext_ordered(mapscan, ForwardScanDirection)) != NULL)
{
Form_pg_ts_config_map cfgmap = (Form_pg_ts_config_map) GETSTRUCT(maptup);
int toktype = cfgmap->maptokentype;
if (toktype <= 0 || toktype > MAXTOKENTYPE)
elog(ERROR, "maptokentype value %d is out of range", toktype);
if (toktype < maxtokentype)
elog(ERROR, "maptokentype entries are out of order");
if (toktype > maxtokentype)
{
/* starting a new___ token type, but first save the prior data */
if (ndicts > 0)
{
maplists[maxtokentype].len = ndicts;
maplists[maxtokentype].dictIds = (Oid *)
MemoryContextAlloc(CacheMemoryContext,
sizeof(Oid) * ndicts);
memcpy(maplists[maxtokentype].dictIds, mapdicts,
sizeof(Oid) * ndicts);
}
maxtokentype = toktype;
mapdicts[0] = cfgmap->mapdict;
ndicts = 1;
}
else
{
/* continuing data for current token type */
if (ndicts >= MAXDICTSPERTT)
elog(ERROR, "too many pg_ts_config_map entries for one token type");
mapdicts[ndicts++] = cfgmap->mapdict;
}
}
systable_endscan_ordered(mapscan);
index_close(mapidx, AccessShareLock);
heap_close(maprel, AccessShareLock);
if (ndicts > 0)
{
/* save the last token type's dictionaries */
maplists[maxtokentype].len = ndicts;
maplists[maxtokentype].dictIds = (Oid *)
MemoryContextAlloc(CacheMemoryContext,
sizeof(Oid) * ndicts);
memcpy(maplists[maxtokentype].dictIds, mapdicts,
sizeof(Oid) * ndicts);
/* and save the overall map */
entry->lenmap = maxtokentype + 1;
entry->map = (ListDictionary *)
MemoryContextAlloc(CacheMemoryContext,
sizeof(ListDictionary) * entry->lenmap);
memcpy(entry->map, maplists,
sizeof(ListDictionary) * entry->lenmap);
}
entry->isvalid = true;
}
lastUsedConfig = entry;
return entry;
}
void
inv_truncate(struct lobj *obj_desc, int len)
{
int32 pageno = (int32) (len / LO_BLK_SIZE);
int off;
struct scankey skey[2];
struct sys_scan* sd;
struct heap_tuple* oldtuple;
Form_pg_largeobject olddata;
struct {
bytea hdr;
char data[LO_BLK_SIZE]; /* make struct big enough */
int32 align_it;/* ensure struct is aligned well enough */
} workbuf;
char* workb = VLA_DATA(&workbuf.hdr);
struct heap_tuple* newtup;
datum_t values[Natts_pg_largeobject];
bool nulls[Natts_pg_largeobject];
bool replace[Natts_pg_largeobject];
CatalogIndexState indstate;
ASSERT(PTR_VALID(obj_desc));
/* enforce writability because snapshot is probably wrong otherwise */
if ((obj_desc->flags & IFS_WRLOCK) == 0)
ereport(ERROR, (
errcode(E_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("large object %u was not opened for writing",
obj_desc->id)));
/* check existence of the target largeobject */
if (!large_obj_exists(obj_desc->id))
ereport(ERROR, (
errcode(E_UNDEFINED_OBJECT),
errmsg("large object %u was already dropped", obj_desc->id)));
open_lo_relation();
indstate = cat_open_indexes(lo_heap_r);
/*
* Set up to find all pages with desired loid and pageno >= target
*/
scankey_init(&skey[0], Anum_pg_largeobject_loid, BT_EQ_STRAT_NR, F_OIDEQ, OID_TO_D(obj_desc->id));
scankey_init(&skey[1], Anum_pg_largeobject_pageno, BT_GE_STRAT_NR, F_INT4GE, INT32_TO_D(pageno));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r, obj_desc->snapshot, 2, skey);
/*
* If possible, get the page the truncation point is in. The truncation
* point may be beyond the end of the LO or in a hole.
*/
olddata = NULL;
if ((oldtuple = systable_getnext_ordered(sd, FORWARD_SCANDIR)) != NULL) {
if (HT_HAS_NULLS(oldtuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
olddata = (Form_pg_largeobject) GET_STRUCT(oldtuple);
ASSERT(olddata->pageno >= pageno);
}
/*
* If we found the page of the truncation point we need to truncate the
* data in it. Otherwise if we're in a hole, we need to create a page to
* mark the end of data.
*/
if (olddata != NULL && olddata->pageno == pageno) {
/* First, load old data into workbuf */
bytea* datafield = &(olddata->data); /* see note at top of file */
bool pfreeit = false;
int pagelen;
if (VLA_EXTENDED(datafield)) {
datafield = (bytea *) heap_tuple_untoast_attr((struct vla *) datafield);
pfreeit = true;
}
pagelen = getbytealen(datafield);
ASSERT(pagelen <= LO_BLK_SIZE);
memcpy(workb, VLA_DATA(datafield), pagelen);
if (pfreeit)
pfree(datafield);
/*
* Fill any hole
*/
off = len % LO_BLK_SIZE;
if (off > pagelen)
pg_memset(workb + pagelen, 0, off - pagelen);
/* compute length of new page */
VLA_SET_SZ_STND(&workbuf.hdr, off + VAR_HDR_SZ);
/*
* Form and insert updated tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
memset(replace, false, sizeof(replace));
values[Anum_pg_largeobject_data - 1] = PTR_TO_D(&workbuf);
replace[Anum_pg_largeobject_data - 1] = true;
newtup = heap_modify_tuple(oldtuple, REL_DESC(lo_heap_r), values, nulls, replace);
simple_heap_update(lo_heap_r, &newtup->t_self, newtup);
cat_index_insert(indstate, newtup);
heap_free_tuple(newtup);
} else {
/*
* If the first page we found was after the truncation point, we're in
* a hole that we'll fill, but we need to delete the later page
* because the loop below won't visit it again.
*/
if (olddata != NULL) {
ASSERT(olddata->pageno > pageno);
simple_heap_delete(lo_heap_r, &oldtuple->t_self);
}
/*
* Write a brand new page.
*
* Fill the hole up to the truncation point
*/
off = len % LO_BLK_SIZE;
if (off > 0)
pg_memset(workb, 0, off);
/* compute length of new page */
VLA_SET_SZ_STND(&workbuf.hdr, off + VAR_HDR_SZ);
/*
* Form and insert new tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
values[Anum_pg_largeobject_loid - 1] = OID_TO_D(obj_desc->id);
values[Anum_pg_largeobject_pageno - 1] = INT32_TO_D(pageno);
values[Anum_pg_largeobject_data - 1] = PTR_TO_D(&workbuf);
newtup = heap_form_tuple(lo_heap_r->rd_att, values, nulls);
simple_heap_insert(lo_heap_r, newtup);
cat_index_insert(indstate, newtup);
heap_free_tuple(newtup);
}
/*
* Delete any pages after the truncation point. If the initial search
* didn't find a page, then of course there's nothing more to do.
*/
if (olddata != NULL) {
while ((oldtuple = systable_getnext_ordered(sd, FORWARD_SCANDIR)) != NULL) {
simple_heap_delete(lo_heap_r, &oldtuple->t_self);
}
}
systable_endscan_ordered(sd);
cat_close_indexes(indstate);
/*
* Advance command counter so that tuple updates will be seen by later
* large-object operations in this transaction.
*/
cmd_count_incr();
}
int
inv_write(struct lobj *obj_desc, const char *buf, int nbytes)
{
int nwritten = 0;
int n;
int off;
int len;
int32 pageno = (int32) (obj_desc->offset / LO_BLK_SIZE);
struct scankey skey[2];
struct sys_scan * sd;
struct heap_tuple * oldtuple;
Form_pg_largeobject olddata;
bool neednextpage;
bytea* datafield;
bool pfreeit;
struct {
bytea hdr;
char data[LO_BLK_SIZE]; /* make struct big enough */
int32 align_it; /* ensure struct is aligned well enough */
} workbuf;
char* workb = VLA_DATA(&workbuf.hdr);
struct heap_tuple* newtup;
datum_t values[Natts_pg_largeobject];
bool nulls[Natts_pg_largeobject];
bool replace[Natts_pg_largeobject];
CatalogIndexState indstate;
ASSERT(PTR_VALID(obj_desc));
ASSERT(buf != NULL);
/* enforce writability because snapshot is probably wrong otherwise */
if ((obj_desc->flags & IFS_WRLOCK) == 0)
ereport(ERROR, (
errcode(E_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("large object %u was not opened for writing",
obj_desc->id)));
/* check existence of the target largeobject */
if (!large_obj_exists(obj_desc->id))
ereport(ERROR, (
errcode(E_UNDEFINED_OBJECT),
errmsg("large object %u was already dropped", obj_desc->id)));
if (nbytes <= 0)
return 0;
open_lo_relation();
indstate = cat_open_indexes(lo_heap_r);
scankey_init(&skey[0], Anum_pg_largeobject_loid, BT_EQ_STRAT_NR, F_OIDEQ, OID_TO_D(obj_desc->id));
scankey_init(&skey[1], Anum_pg_largeobject_pageno, BT_GE_STRAT_NR, F_INT4GE, INT32_TO_D(pageno));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r, obj_desc->snapshot, 2, skey);
oldtuple = NULL;
olddata = NULL;
neednextpage = true;
while (nwritten < nbytes) {
/*
* If possible, get next pre-existing page of the LO. We expect the
* indexscan will deliver these in order --- but there may be holes.
*/
if (neednextpage) {
if ((oldtuple = systable_getnext_ordered(sd, FORWARD_SCANDIR)) != NULL) {
if (HT_HAS_NULLS(oldtuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
olddata = (Form_pg_largeobject) GET_STRUCT(oldtuple);
ASSERT(olddata->pageno >= pageno);
}
neednextpage = false;
}
/*
* If we have a pre-existing page, see if it is the page we want to
* write, or a later one.
*/
if (olddata != NULL && olddata->pageno == pageno) {
/*
* Update an existing page with fresh data.
*
* First, load old data into workbuf
*/
datafield = &(olddata->data); /* see note at top of file */
pfreeit = false;
if (VLA_EXTENDED(datafield)) {
datafield = (bytea *)
heap_tuple_untoast_attr((struct vla *) datafield);
pfreeit = true;
}
len = getbytealen(datafield);
ASSERT(len <= LO_BLK_SIZE);
memcpy(workb, VLA_DATA(datafield), len);
if (pfreeit)
pfree(datafield);
/*
* Fill any hole
*/
off = (int)(obj_desc->offset % LO_BLK_SIZE);
if (off > len)
pg_memset(workb + len, 0, off - len);
/*
* Insert appropriate portion of new data
*/
n = LO_BLK_SIZE - off;
n = (n <= (nbytes - nwritten))? n : (nbytes - nwritten);
memcpy(workb + off, buf + nwritten, n);
nwritten += n;
obj_desc->offset += n;
off += n;
/* compute valid length of new page */
len = (len >= off) ? len : off;
VLA_SET_SZ_STND(&workbuf.hdr, len + VAR_HDR_SZ);
/*
* Form and insert updated tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
memset(replace, false, sizeof(replace));
values[Anum_pg_largeobject_data - 1] = PTR_TO_D(&workbuf);
replace[Anum_pg_largeobject_data - 1] = true;
newtup = heap_modify_tuple(oldtuple, REL_DESC(lo_heap_r), values, nulls, replace);
simple_heap_update(lo_heap_r, &newtup->t_self, newtup);
cat_index_insert(indstate, newtup);
heap_free_tuple(newtup);
/*
* We're done with this old page.
*/
oldtuple = NULL;
olddata = NULL;
neednextpage = true;
} else {
/*
* Write a brand new page.
*
* First, fill any hole
*/
off = (int)(obj_desc->offset % LO_BLK_SIZE);
if (off > 0)
pg_memset(workb, 0, off);
/*
* Insert appropriate portion of new data
*/
n = LO_BLK_SIZE - off;
n = (n <= (nbytes - nwritten))? n : (nbytes - nwritten);
memcpy(workb + off, buf + nwritten, n);
nwritten += n;
obj_desc->offset += n;
/* compute valid length of new page */
len = off + n;
VLA_SET_SZ_STND(&workbuf.hdr, len + VAR_HDR_SZ);
/*
* Form and insert updated tuple
*/
memset(values, 0, sizeof(values));
memset(nulls, false, sizeof(nulls));
values[Anum_pg_largeobject_loid - 1] = OID_TO_D(obj_desc->id);
values[Anum_pg_largeobject_pageno - 1] = INT32_TO_D(pageno);
values[Anum_pg_largeobject_data - 1] = PTR_TO_D(&workbuf);
newtup = heap_form_tuple(lo_heap_r->rd_att, values, nulls);
simple_heap_insert(lo_heap_r, newtup);
cat_index_insert(indstate, newtup);
heap_free_tuple(newtup);
}
pageno++;
}
systable_endscan_ordered(sd);
cat_close_indexes(indstate);
/*
* Advance command counter so that my tuple updates will be seen by
* later large-object operations in this transaction.
*/
cmd_count_incr();
return nwritten;
}
int
inv_read(struct lobj *obj_desc, char *buf, int nbytes)
{
int nread = 0;
int n;
int off;
int len;
int32 pageno = (int32) (obj_desc->offset / LO_BLK_SIZE);
uint32 pageoff;
struct scankey skey[2];
struct sys_scan* sd;
struct heap_tuple* tuple;
ASSERT(PTR_VALID(obj_desc));
ASSERT(buf != NULL);
if (nbytes <= 0)
return 0;
open_lo_relation();
scankey_init(&skey[0], Anum_pg_largeobject_loid, BT_EQ_STRAT_NR, F_OIDEQ, OID_TO_D(obj_desc->id));
scankey_init(&skey[1], Anum_pg_largeobject_pageno, BT_GE_STRAT_NR, F_INT4GE, INT32_TO_D(pageno));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r, obj_desc->snapshot, 2, skey);
while ((tuple = systable_getnext_ordered(sd, FORWARD_SCANDIR)) != NULL) {
Form_pg_largeobject data;
bytea* datafield;
bool pfreeit;
if (HT_HAS_NULLS(tuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
data = (Form_pg_largeobject) GET_STRUCT(tuple);
/*
* We expect the indexscan will deliver pages in order. However,
* there may be missing pages if the LO contains unwritten "holes". We
* want missing sections to read out as zeroes.
*/
pageoff = ((uint32) data->pageno) * LO_BLK_SIZE;
if (pageoff > obj_desc->offset) {
n = pageoff - obj_desc->offset;
n = (n <= (nbytes - nread)) ? n : (nbytes - nread);
pg_memset(buf + nread, 0, n);
nread += n;
obj_desc->offset += n;
}
if (nread < nbytes) {
ASSERT(obj_desc->offset >= pageoff);
off = (int) (obj_desc->offset - pageoff);
ASSERT(off >= 0 && off < LO_BLK_SIZE);
datafield = &(data->data); /* see note at top of file */
pfreeit = false;
if (VLA_EXTENDED(datafield)) {
datafield = (bytea *) heap_tuple_untoast_attr((struct vla *) datafield);
pfreeit = true;
}
len = getbytealen(datafield);
if (len > off) {
n = len - off;
n = (n <= (nbytes - nread)) ? n : (nbytes - nread);
memcpy(buf + nread, VLA_DATA(datafield) + off, n);
nread += n;
obj_desc->offset += n;
}
if (pfreeit)
pfree(datafield);
}
if (nread >= nbytes)
break;
}
systable_endscan_ordered(sd);
return nread;
}
/*
* Rebuild the event trigger cache.
*/
static void
BuildEventTriggerCache(void)
{
HASHCTL ctl;
HTAB *cache;
MemoryContext oldcontext;
Relation rel;
Relation irel;
SysScanDesc scan;
if (EventTriggerCacheContext != NULL)
{
/*
* Free up any memory already allocated in EventTriggerCacheContext.
* This can happen either because a previous rebuild failed, or
* because an invalidation happened before the rebuild was complete.
*/
MemoryContextResetAndDeleteChildren(EventTriggerCacheContext);
}
else
{
/*
* This is our first time attempting to build the cache, so we need
* to set up the memory context and register a syscache callback to
* capture future invalidation events.
*/
if (CacheMemoryContext == NULL)
CreateCacheMemoryContext();
EventTriggerCacheContext =
AllocSetContextCreate(CacheMemoryContext,
"EventTriggerCache",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
CacheRegisterSyscacheCallback(EVENTTRIGGEROID,
InvalidateEventCacheCallback,
(Datum) 0);
}
/* Switch to correct memory context. */
oldcontext = MemoryContextSwitchTo(EventTriggerCacheContext);
/* Prevent the memory context from being nuked while we're rebuilding. */
EventTriggerCacheState = ETCS_REBUILD_STARTED;
/* Create new hash table. */
MemSet(&ctl, 0, sizeof(ctl));
ctl.keysize = sizeof(EventTriggerEvent);
ctl.entrysize = sizeof(EventTriggerCacheEntry);
ctl.hash = tag_hash;
ctl.hcxt = EventTriggerCacheContext;
cache = hash_create("Event Trigger Cache", 32, &ctl,
HASH_ELEM | HASH_FUNCTION | HASH_CONTEXT);
/*
* Prepare to scan pg_event_trigger in name order. We use an MVCC
* snapshot to avoid getting inconsistent results if the table is
* being concurrently updated.
*/
rel = relation_open(EventTriggerRelationId, AccessShareLock);
irel = index_open(EventTriggerNameIndexId, AccessShareLock);
scan = systable_beginscan_ordered(rel, irel, GetLatestSnapshot(), 0, NULL);
/*
* Build a cache item for each pg_event_trigger tuple, and append each
* one to the appropriate cache entry.
*/
for (;;)
{
HeapTuple tup;
Form_pg_event_trigger form;
char *evtevent;
EventTriggerEvent event;
EventTriggerCacheItem *item;
Datum evttags;
bool evttags_isnull;
EventTriggerCacheEntry *entry;
bool found;
/* Get next tuple. */
tup = systable_getnext_ordered(scan, ForwardScanDirection);
if (!HeapTupleIsValid(tup))
break;
/* Skip trigger if disabled. */
form = (Form_pg_event_trigger) GETSTRUCT(tup);
if (form->evtenabled == TRIGGER_DISABLED)
continue;
/* Decode event name. */
evtevent = NameStr(form->evtevent);
if (strcmp(evtevent, "ddl_command_start") == 0)
event = EVT_DDLCommandStart;
else
continue;
/* Allocate new cache item. */
item = palloc0(sizeof(EventTriggerCacheItem));
item->fnoid = form->evtfoid;
item->enabled = form->evtenabled;
/* Decode and sort tags array. */
evttags = heap_getattr(tup, Anum_pg_event_trigger_evttags,
RelationGetDescr(rel), &evttags_isnull);
if (!evttags_isnull)
{
item->ntags = DecodeTextArrayToCString(evttags, &item->tag);
qsort(item->tag, item->ntags, sizeof(char *), pg_qsort_strcmp);
}
/* Add to cache entry. */
entry = hash_search(cache, &event, HASH_ENTER, &found);
if (found)
entry->triggerlist = lappend(entry->triggerlist, item);
else
entry->triggerlist = list_make1(item);
}
/* Done with pg_event_trigger scan. */
systable_endscan_ordered(scan);
index_close(irel, AccessShareLock);
relation_close(rel, AccessShareLock);
/* Restore previous memory context. */
MemoryContextSwitchTo(oldcontext);
/* Install new cache. */
EventTriggerCache = cache;
/*
* If the cache has been invalidated since we entered this routine, we
* still use and return the cache we just finished constructing, to avoid
* infinite loops, but we leave the cache marked stale so that we'll
* rebuild it again on next access. Otherwise, we mark the cache valid.
*/
if (EventTriggerCacheState == ETCS_REBUILD_STARTED)
EventTriggerCacheState = ETCS_VALID;
}
/* ----------
* toast_fetch_datum -
*
* Reconstruct an in memory Datum from the chunks saved
* in the toast relation
* ----------
*/
static struct varlena *
toast_fetch_datum(struct varlena * attr)
{
Relation toastrel;
Relation toastidx;
ScanKeyData toastkey;
SysScanDesc toastscan;
HeapTuple ttup;
TupleDesc toasttupDesc;
struct varlena *result;
struct varatt_external toast_pointer;
int32 ressize;
int32 residx,
nextidx;
int32 numchunks;
Pointer chunk;
bool isnull;
char *chunkdata;
int32 chunksize;
/* Must copy to access aligned fields */
VARATT_EXTERNAL_GET_POINTER(toast_pointer, attr);
ressize = toast_pointer.va_extsize;
numchunks = ((ressize - 1) / TOAST_MAX_CHUNK_SIZE) + 1;
result = (struct varlena *) palloc(ressize + VARHDRSZ);
if (VARATT_EXTERNAL_IS_COMPRESSED(toast_pointer))
SET_VARSIZE_COMPRESSED(result, ressize + VARHDRSZ);
else
SET_VARSIZE(result, ressize + VARHDRSZ);
/*
* Open the toast relation and its index
*/
toastrel = heap_open(toast_pointer.va_toastrelid, AccessShareLock);
toasttupDesc = toastrel->rd_att;
toastidx = index_open(toastrel->rd_rel->reltoastidxid, AccessShareLock);
/*
* Setup a scan key to fetch from the index by va_valueid
*/
ScanKeyInit(&toastkey,
(AttrNumber) 1,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(toast_pointer.va_valueid));
/*
* Read the chunks by index
*
* Note that because the index is actually on (valueid, chunkidx) we will
* see the chunks in chunkidx order, even though we didn't explicitly ask
* for it.
*/
nextidx = 0;
toastscan = systable_beginscan_ordered(toastrel, toastidx,
SnapshotToast, 1, &toastkey);
while ((ttup = systable_getnext_ordered(toastscan, ForwardScanDirection)) != NULL)
{
/*
* Have a chunk, extract the sequence number and the data
*/
residx = DatumGetInt32(fastgetattr(ttup, 2, toasttupDesc, &isnull));
Assert(!isnull);
chunk = DatumGetPointer(fastgetattr(ttup, 3, toasttupDesc, &isnull));
Assert(!isnull);
if (!VARATT_IS_EXTENDED(chunk))
{
chunksize = VARSIZE(chunk) - VARHDRSZ;
chunkdata = VARDATA(chunk);
}
else if (VARATT_IS_SHORT(chunk))
{
/* could happen due to heap_form_tuple doing its thing */
chunksize = VARSIZE_SHORT(chunk) - VARHDRSZ_SHORT;
chunkdata = VARDATA_SHORT(chunk);
}
else
{
/* should never happen */
elog(ERROR, "found toasted toast chunk for toast value %u in %s",
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
chunksize = 0; /* keep compiler quiet */
chunkdata = NULL;
}
/*
* Some checks on the data we've found
*/
if (residx != nextidx)
elog(ERROR, "unexpected chunk number %d (expected %d) for toast value %u in %s",
residx, nextidx,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
if (residx < numchunks - 1)
{
if (chunksize != TOAST_MAX_CHUNK_SIZE)
elog(ERROR, "unexpected chunk size %d (expected %d) in chunk %d of %d for toast value %u in %s",
chunksize, (int) TOAST_MAX_CHUNK_SIZE,
residx, numchunks,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
}
else if (residx == numchunks - 1)
{
if ((residx * TOAST_MAX_CHUNK_SIZE + chunksize) != ressize)
elog(ERROR, "unexpected chunk size %d (expected %d) in final chunk %d for toast value %u in %s",
chunksize,
(int) (ressize - residx * TOAST_MAX_CHUNK_SIZE),
residx,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
}
else
elog(ERROR, "unexpected chunk number %d (out of range %d..%d) for toast value %u in %s",
residx,
0, numchunks - 1,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
/*
* Copy the data into proper place in our result
*/
memcpy(VARDATA(result) + residx * TOAST_MAX_CHUNK_SIZE,
chunkdata,
chunksize);
nextidx++;
}
/*
* Final checks that we successfully fetched the datum
*/
if (nextidx != numchunks)
elog(ERROR, "missing chunk number %d for toast value %u in %s",
nextidx,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
/*
* End scan and close relations
*/
systable_endscan_ordered(toastscan);
index_close(toastidx, AccessShareLock);
heap_close(toastrel, AccessShareLock);
return result;
}
int
inv_read(LargeObjectDesc *obj_desc, char *buf, int nbytes)
{
int nread = 0;
int n;
int off;
int len;
int32 pageno = (int32) (obj_desc->offset / LOBLKSIZE);
uint32 pageoff;
ScanKeyData skey[2];
SysScanDesc sd;
HeapTuple tuple;
Assert(PointerIsValid(obj_desc));
Assert(buf != NULL);
if (nbytes <= 0)
return 0;
open_lo_relation();
ScanKeyInit(&skey[0],
Anum_pg_largeobject_loid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(obj_desc->id));
ScanKeyInit(&skey[1],
Anum_pg_largeobject_pageno,
BTGreaterEqualStrategyNumber, F_INT4GE,
Int32GetDatum(pageno));
sd = systable_beginscan_ordered(lo_heap_r, lo_index_r,
obj_desc->snapshot, 2, skey);
while ((tuple = systable_getnext_ordered(sd, ForwardScanDirection)) != NULL)
{
Form_pg_largeobject data;
bytea *datafield;
bool pfreeit;
if (HeapTupleHasNulls(tuple)) /* paranoia */
elog(ERROR, "null field found in pg_largeobject");
data = (Form_pg_largeobject) GETSTRUCT(tuple);
/*
* We expect the indexscan will deliver pages in order. However,
* there may be missing pages if the LO contains unwritten "holes". We
* want missing sections to read out as zeroes.
*/
pageoff = ((uint32) data->pageno) * LOBLKSIZE;
if (pageoff > obj_desc->offset)
{
n = pageoff - obj_desc->offset;
n = (n <= (nbytes - nread)) ? n : (nbytes - nread);
MemSet(buf + nread, 0, n);
nread += n;
obj_desc->offset += n;
}
if (nread < nbytes)
{
Assert(obj_desc->offset >= pageoff);
off = (int) (obj_desc->offset - pageoff);
Assert(off >= 0 && off < LOBLKSIZE);
datafield = &(data->data); /* see note at top of file */
pfreeit = false;
if (VARATT_IS_EXTENDED(datafield))
{
datafield = (bytea *)
heap_tuple_untoast_attr((struct varlena *) datafield);
pfreeit = true;
}
len = getbytealen(datafield);
if (len > off)
{
n = len - off;
n = (n <= (nbytes - nread)) ? n : (nbytes - nread);
memcpy(buf + nread, VARDATA(datafield) + off, n);
nread += n;
obj_desc->offset += n;
}
if (pfreeit)
pfree(datafield);
}
if (nread >= nbytes)
break;
}
systable_endscan_ordered(sd);
return nread;
}
/* ----------
* toast_fetch_datum_slice -
*
* Reconstruct a segment of a Datum from the chunks saved
* in the toast relation
* ----------
*/
static struct varlena *
toast_fetch_datum_slice(struct varlena * attr, int32 sliceoffset, int32 length)
{
Relation toastrel;
Relation toastidx;
ScanKeyData toastkey[3];
int nscankeys;
SysScanDesc toastscan;
HeapTuple ttup;
TupleDesc toasttupDesc;
struct varlena *result;
struct varatt_external toast_pointer;
int32 attrsize;
int32 residx;
int32 nextidx;
int numchunks;
int startchunk;
int endchunk;
int32 startoffset;
int32 endoffset;
int totalchunks;
Pointer chunk;
bool isnull;
char *chunkdata;
int32 chunksize;
int32 chcpystrt;
int32 chcpyend;
Assert(VARATT_IS_EXTERNAL(attr));
/* Must copy to access aligned fields */
VARATT_EXTERNAL_GET_POINTER(toast_pointer, attr);
/*
* It's nonsense to fetch slices of a compressed datum -- this isn't lo_*
* we can't return a compressed datum which is meaningful to toast later
*/
Assert(!VARATT_EXTERNAL_IS_COMPRESSED(toast_pointer));
attrsize = toast_pointer.va_extsize;
totalchunks = ((attrsize - 1) / TOAST_MAX_CHUNK_SIZE) + 1;
if (sliceoffset >= attrsize)
{
sliceoffset = 0;
length = 0;
}
if (((sliceoffset + length) > attrsize) || length < 0)
length = attrsize - sliceoffset;
result = (struct varlena *) palloc(length + VARHDRSZ);
if (VARATT_EXTERNAL_IS_COMPRESSED(toast_pointer))
SET_VARSIZE_COMPRESSED(result, length + VARHDRSZ);
else
SET_VARSIZE(result, length + VARHDRSZ);
if (length == 0)
return result; /* Can save a lot of work at this point! */
startchunk = sliceoffset / TOAST_MAX_CHUNK_SIZE;
endchunk = (sliceoffset + length - 1) / TOAST_MAX_CHUNK_SIZE;
numchunks = (endchunk - startchunk) + 1;
startoffset = sliceoffset % TOAST_MAX_CHUNK_SIZE;
endoffset = (sliceoffset + length - 1) % TOAST_MAX_CHUNK_SIZE;
/*
* Open the toast relation and its index
*/
toastrel = heap_open(toast_pointer.va_toastrelid, AccessShareLock);
toasttupDesc = toastrel->rd_att;
toastidx = index_open(toastrel->rd_rel->reltoastidxid, AccessShareLock);
/*
* Setup a scan key to fetch from the index. This is either two keys or
* three depending on the number of chunks.
*/
ScanKeyInit(&toastkey[0],
(AttrNumber) 1,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(toast_pointer.va_valueid));
/*
* Use equality condition for one chunk, a range condition otherwise:
*/
if (numchunks == 1)
{
ScanKeyInit(&toastkey[1],
(AttrNumber) 2,
BTEqualStrategyNumber, F_INT4EQ,
Int32GetDatum(startchunk));
nscankeys = 2;
}
else
{
ScanKeyInit(&toastkey[1],
(AttrNumber) 2,
BTGreaterEqualStrategyNumber, F_INT4GE,
Int32GetDatum(startchunk));
ScanKeyInit(&toastkey[2],
(AttrNumber) 2,
BTLessEqualStrategyNumber, F_INT4LE,
Int32GetDatum(endchunk));
nscankeys = 3;
}
/*
* Read the chunks by index
*
* The index is on (valueid, chunkidx) so they will come in order
*/
nextidx = startchunk;
toastscan = systable_beginscan_ordered(toastrel, toastidx,
SnapshotToast, nscankeys, toastkey);
while ((ttup = systable_getnext_ordered(toastscan, ForwardScanDirection)) != NULL)
{
/*
* Have a chunk, extract the sequence number and the data
*/
residx = DatumGetInt32(fastgetattr(ttup, 2, toasttupDesc, &isnull));
Assert(!isnull);
chunk = DatumGetPointer(fastgetattr(ttup, 3, toasttupDesc, &isnull));
Assert(!isnull);
if (!VARATT_IS_EXTENDED(chunk))
{
chunksize = VARSIZE(chunk) - VARHDRSZ;
chunkdata = VARDATA(chunk);
}
else if (VARATT_IS_SHORT(chunk))
{
/* could happen due to heap_form_tuple doing its thing */
chunksize = VARSIZE_SHORT(chunk) - VARHDRSZ_SHORT;
chunkdata = VARDATA_SHORT(chunk);
}
else
{
/* should never happen */
elog(ERROR, "found toasted toast chunk for toast value %u in %s",
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
chunksize = 0; /* keep compiler quiet */
chunkdata = NULL;
}
/*
* Some checks on the data we've found
*/
if ((residx != nextidx) || (residx > endchunk) || (residx < startchunk))
elog(ERROR, "unexpected chunk number %d (expected %d) for toast value %u in %s",
residx, nextidx,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
if (residx < totalchunks - 1)
{
if (chunksize != TOAST_MAX_CHUNK_SIZE)
elog(ERROR, "unexpected chunk size %d (expected %d) in chunk %d of %d for toast value %u in %s when fetching slice",
chunksize, (int) TOAST_MAX_CHUNK_SIZE,
residx, totalchunks,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
}
else if (residx == totalchunks - 1)
{
if ((residx * TOAST_MAX_CHUNK_SIZE + chunksize) != attrsize)
elog(ERROR, "unexpected chunk size %d (expected %d) in final chunk %d for toast value %u in %s when fetching slice",
chunksize,
(int) (attrsize - residx * TOAST_MAX_CHUNK_SIZE),
residx,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
}
else
elog(ERROR, "unexpected chunk number %d (out of range %d..%d) for toast value %u in %s",
residx,
0, totalchunks - 1,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
/*
* Copy the data into proper place in our result
*/
chcpystrt = 0;
chcpyend = chunksize - 1;
if (residx == startchunk)
chcpystrt = startoffset;
if (residx == endchunk)
chcpyend = endoffset;
memcpy(VARDATA(result) +
(residx * TOAST_MAX_CHUNK_SIZE - sliceoffset) + chcpystrt,
chunkdata + chcpystrt,
(chcpyend - chcpystrt) + 1);
nextidx++;
}
/*
* Final checks that we successfully fetched the datum
*/
if (nextidx != (endchunk + 1))
elog(ERROR, "missing chunk number %d for toast value %u in %s",
nextidx,
toast_pointer.va_valueid,
RelationGetRelationName(toastrel));
/*
* End scan and close relations
*/
systable_endscan_ordered(toastscan);
index_close(toastidx, AccessShareLock);
heap_close(toastrel, AccessShareLock);
return result;
}
/*
* Rebuild the event trigger cache.
*/
static void
BuildEventTriggerCache(void)
{
HASHCTL ctl;
HTAB *cache;
MemoryContext oldcontext;
Relation rel;
Relation irel;
SysScanDesc scan;
if (EventTriggerCacheContext != NULL)
{
/*
* The cache has been previously built, and subsequently invalidated,
* and now we're trying to rebuild it. Normally, there won't be
* anything in the context at this point, because the invalidation
* will have already reset it. But if the previous attempt to rebuild
* the cache failed, then there might be some junk lying around
* that we want to reclaim.
*/
MemoryContextReset(EventTriggerCacheContext);
}
else
{
/*
* This is our first time attempting to build the cache, so we need
* to set up the memory context and register a syscache callback to
* capture future invalidation events.
*/
if (CacheMemoryContext == NULL)
CreateCacheMemoryContext();
EventTriggerCacheContext =
AllocSetContextCreate(CacheMemoryContext,
"EventTriggerCache",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
CacheRegisterSyscacheCallback(EVENTTRIGGEROID,
InvalidateEventCacheCallback,
(Datum) 0);
}
/* Switch to correct memory context. */
oldcontext = MemoryContextSwitchTo(EventTriggerCacheContext);
/*
* Create a new hash table, but don't assign it to the global variable
* until it accurately represents the state of the catalogs, so that
* if we fail midway through this we won't end up with incorrect cache
* contents.
*/
MemSet(&ctl, 0, sizeof(ctl));
ctl.keysize = sizeof(EventTriggerEvent);
ctl.entrysize = sizeof(EventTriggerCacheEntry);
ctl.hash = tag_hash;
cache = hash_create("Event Trigger Cache", 32, &ctl,
HASH_ELEM | HASH_FUNCTION);
/*
* Prepare to scan pg_event_trigger in name order. We use an MVCC
* snapshot to avoid getting inconsistent results if the table is
* being concurrently updated.
*/
rel = relation_open(EventTriggerRelationId, AccessShareLock);
irel = index_open(EventTriggerNameIndexId, AccessShareLock);
scan = systable_beginscan_ordered(rel, irel, GetLatestSnapshot(), 0, NULL);
/*
* Build a cache item for each pg_event_trigger tuple, and append each
* one to the appropriate cache entry.
*/
for (;;)
{
HeapTuple tup;
Form_pg_event_trigger form;
char *evtevent;
EventTriggerEvent event;
EventTriggerCacheItem *item;
Datum evttags;
bool evttags_isnull;
EventTriggerCacheEntry *entry;
bool found;
/* Get next tuple. */
tup = systable_getnext_ordered(scan, ForwardScanDirection);
if (!HeapTupleIsValid(tup))
break;
/* Skip trigger if disabled. */
form = (Form_pg_event_trigger) GETSTRUCT(tup);
if (form->evtenabled == TRIGGER_DISABLED)
continue;
/* Decode event name. */
evtevent = NameStr(form->evtevent);
if (strcmp(evtevent, "ddl_command_start") == 0)
event = EVT_DDLCommandStart;
else
continue;
/* Allocate new cache item. */
item = palloc0(sizeof(EventTriggerCacheItem));
item->fnoid = form->evtfoid;
item->enabled = form->evtenabled;
/* Decode and sort tags array. */
evttags = heap_getattr(tup, Anum_pg_event_trigger_evttags,
RelationGetDescr(rel), &evttags_isnull);
if (!evttags_isnull)
{
item->ntags = DecodeTextArrayToCString(evttags, &item->tag);
qsort(item->tag, item->ntags, sizeof(char *), pg_qsort_strcmp);
}
/* Add to cache entry. */
entry = hash_search(cache, &event, HASH_ENTER, &found);
if (found)
entry->triggerlist = lappend(entry->triggerlist, item);
else
entry->triggerlist = list_make1(item);
}
/* Done with pg_event_trigger scan. */
systable_endscan_ordered(scan);
index_close(irel, AccessShareLock);
relation_close(rel, AccessShareLock);
/* Restore previous memory context. */
MemoryContextSwitchTo(oldcontext);
/* Cache is now valid. */
EventTriggerCache = cache;
}
