/* ----------------------------------------------------------------
* ForeignNext
*
* This is a workhorse for ExecForeignScan
* ----------------------------------------------------------------
*/
static struct tupslot *
ForeignNext(foreign_ss *node)
{
struct tupslot *slot;
foreign_scan_sc *plan = (foreign_scan_sc *) node->ss.ps.plan;
expr_ctx_n *econtext = node->ss.ps.ps_ExprContext;
struct mctx * oldcontext;
/* Call the Iterate function in short-lived context */
oldcontext = mctx_switch(econtext->ecxt_per_tuple_memory);
slot = node->fdwroutine->IterateForeignScan(node);
mctx_switch(oldcontext);
/*
* If any system columns are requested, we have to force the tuple into
* physical-tuple form to avoid "cannot extract system attribute from
* virtual tuple" errors later. We also insert a valid value for
* tableoid, which is the only actually-useful system column.
*/
if (plan->fsSystemCol
&& !TUPSLOT_NULL(slot)) {
struct heap_tuple* tup;
tup = exec_materialize_slot(slot);
tup->t_tableOid = REL_ID(node->ss.ss_currentRelation);
}
return slot;
}
/* --------------------------------
* exec_materialize_slot
* Force a slot into the "materialized" state.
*
* This causes the slot's tuple to be a local copy not dependent on
* any external storage. A pointer to the contained tuple is returned.
*
* A typical use for this operation is to prepare a computed tuple
* for being stored on disk. The original data may or may not be
* virtual, but in any case we need a private copy for heap_insert
* to scribble on.
* --------------------------------
*/
struct heap_tuple*
exec_materialize_slot(struct tupslot *slot)
{
struct mctx* oldContext;
/*
* sanity checks
*/
ASSERT(slot != NULL);
ASSERT(!slot->tts_isempty);
/*
* If we have a regular physical tuple, and it's locally palloc'd, we have
* nothing to do.
*/
if (slot->tts_tuple && slot->tts_shouldFree)
return slot->tts_tuple;
/*
* Otherwise, copy or build a physical tuple, and store it into the slot.
*
* We may be called in a context that is shorter-lived than the tuple
* slot, but we have to ensure that the materialized tuple will survive
* anyway.
*/
oldContext = mctx_switch(slot->tts_mcxt);
slot->tts_tuple = exec_copy_slot_tuple(slot);
slot->tts_shouldFree = true;
mctx_switch(oldContext);
/*
* Drop the pin on the referenced buffer, if there is one.
*/
if (BUF_VALID(slot->tts_buffer))
release_buf(slot->tts_buffer);
slot->tts_buffer = INVALID_BUF;
/*
* Mark extracted state invalid. This is important because the slot is
* not supposed to depend any more on the previous external data; we
* mustn't leave any dangling pass-by-reference datums in tts_values.
* However, we have not actually invalidated any such datums, if there
* happen to be any previously fetched from the slot. (Note in particular
* that we have not pfree'd tts_mintuple, if there is one.)
*/
slot->tts_nvalid = 0;
/*
* On the same principle of not depending on previous remote storage,
* forget the mintuple if it's not local storage. (If it is local
* storage, we must not pfree it now, since callers might have already
* fetched datum pointers referencing it.)
*/
if (!slot->tts_shouldFreeMin)
slot->tts_mintuple = NULL;
return slot->tts_tuple;
}
/*
* connectby - does the real work for connectby_text()
*/
static struct tupstore *
connectby(char *relname,
char *key_fld,
char *parent_key_fld,
char *orderby_fld,
char *branch_delim,
char *start_with,
int max_depth,
bool show_branch,
bool show_serial,
struct mctx * per_query_ctx,
bool randomAccess,
AttInMetadata *attinmeta)
{
struct tupstore *tupstore = NULL;
int ret;
struct mctx * oldcontext;
int serial = 1;
/* Connect to SPI manager */
if ((ret = SPI_connect()) < 0)
/* internal error */
elog(ERROR, "connectby: SPI_connect returned %d", ret);
/* switch to longer term context to create the tuple store */
oldcontext = mctx_switch(per_query_ctx);
/* initialize our tuplestore */
tupstore = tts_begin_heap(randomAccess, false, work_mem);
mctx_switch(oldcontext);
/* now go get the whole tree */
tupstore = build_tuplestore_recursively(key_fld,
parent_key_fld,
relname,
orderby_fld,
branch_delim,
start_with,
start_with, /* current_branch */
0, /* initial level is 0 */
&serial, /* initial serial is 1 */
max_depth,
show_branch,
show_serial,
per_query_ctx,
attinmeta,
tupstore);
SPI_finish();
return tupstore;
}
/*
* ExecIndexEvalRuntimeKeys
* Evaluate any runtime key values, and update the scankeys.
*/
void
ExecIndexEvalRuntimeKeys(
expr_ctx_n *econtext,
index_runtime_key_info_s *runtimeKeys,
int numRuntimeKeys)
{
int j;
struct mctx* oldContext;
/* We want to keep the key values in per-tuple memory */
oldContext = mctx_switch(econtext->ecxt_per_tuple_memory);
for (j = 0; j < numRuntimeKeys; j++) {
struct scankey* scan_key = runtimeKeys[j].scan_key;
expr_state_n* key_expr = runtimeKeys[j].key_expr;
datum_t scanvalue;
bool isNull;
/*
* For each run-time key, extract the run-time expression and evaluate
* it with respect to the current context. We then stick the result
* into the proper scan key.
*
* Note: the result of the eval could be a pass-by-ref value that's
* stored in some outer scan's tuple, not in
* econtext->ecxt_per_tuple_memory. We assume that the outer tuple
* will stay put throughout our scan. If this is wrong, we could copy
* the result into our context explicitly, but I think that's not
* necessary.
*
* It's also entirely possible that the result of the eval is a
* toasted value. In this case we should forcibly detoast it, to
* avoid repeat detoastings each time the value is examined by an
* index support function.
*/
scanvalue = EXEC_EVAL_EXPR(key_expr, econtext, &isNull, NULL);
if (isNull) {
scan_key->sk_argument = scanvalue;
scan_key->sk_flags |= SK_ISNULL;
} else {
if (runtimeKeys[j].key_toastable)
scanvalue = PTR_TO_D(PG_DETOAST_DATUM(scanvalue));
scan_key->sk_argument = scanvalue;
scan_key->sk_flags &= ~SK_ISNULL;
}
}
mctx_switch(oldContext);
}
/**
* Worker loop
*/
static void WorkerLoop( workerctx_t *wc)
{
lpel_task_t *t = NULL;
do {
/* before executing a task, handle all pending requests! */
LpelSpmdHandleRequests(wc->wid);
t = LpelSchedFetchReady( wc->sched);
if (t != NULL) {
/* execute task */
wc->current_task = t;
mctx_switch(&wc->mctx, &t->mctx);
WORKER_DBGMSG(wc, "Back on worker %d context.\n", wc->wid);
/* check if there are any contexts marked for deletion. */
LpelCollectTask(wc, NULL);
} else {
/* no ready tasks */
WaitForNewMessage( wc);
}
/* fetch (remaining) messages */
FetchAllMessages( wc);
} while ( !( 0==wc->num_tasks && wc->terminate) );
}
/* --------------------------------
* exec_copy_slot
* Copy the source slot's contents into the destination slot.
*
* The destination acquires a private copy that will not go away
* if the source is cleared.
*
* The caller must ensure the slots have compatible tupdescs.
* --------------------------------
*/
struct tupslot*
exec_copy_slot(struct tupslot* dstslot, struct tupslot* srcslot)
{
struct heap_tuple* newTuple;
struct mctx * oldContext;
/*
* There might be ways to optimize this when the source is virtual, but
* for now just always build a physical copy. Make sure it is in the
* right context.
*/
oldContext = mctx_switch(dstslot->tts_mcxt);
newTuple = exec_copy_slot_tuple(srcslot);
mctx_switch(oldContext);
return exec_store_tuple(newTuple, dstslot, INVALID_BUF, true);
}
static void WrapperLoop(workerctx_t *wp)
{
lpel_task_t *t = NULL;
workermsg_t msg;
do {
t = wp->current_task;
if (t != NULL) {
/* execute task */
mctx_switch(&wp->mctx, &t->mctx);
} else {
/* no ready tasks */
LpelMailboxRecv(wp->mailbox, &msg);
switch(msg.type) {
case WORKER_MSG_ASSIGN:
t = msg.body.task;
WORKER_DBG("wrapper: get task %d\n", t->uid);
assert(t->state == TASK_CREATED);
t->state = TASK_READY;
wp->current_task = t;
#ifdef USE_LOGGING
if (t->mon) {
if (MON_CB(worker_create_wrapper)) {
wp->mon = MON_CB(worker_create_wrapper)(t->mon);
} else {
wp->mon = NULL;
}
}
if (t->mon && MON_CB(task_assign)) {
MON_CB(task_assign)(t->mon, wp->mon);
}
#endif
break;
case WORKER_MSG_WAKEUP:
t = msg.body.task;
WORKER_DBG("wrapper: unblock task %d\n", t->uid);
assert (t->state == TASK_BLOCKED);
t->state = TASK_READY;
wp->current_task = t;
#ifdef USE_LOGGING
if (t->mon && MON_CB(task_assign)) {
MON_CB(task_assign)(t->mon, wp->mon);
}
#endif
break;
default:
assert(0);
break;
}
}
} while (!wp->terminate);
LpelTaskDestroy(wp->current_task);
/* cleanup task context marked for deletion */
}
/**
* Dispatch next ready task
*
* This dispatcher function is called upon blocking a task.
* It executes on the same exec-stack as the task itself, calls the FetchReady
* function from the scheduler module, and, if there is a ready task, makes
* a continuation to that ready task. If no task is ready, execution returns to the
* worker loop (wc->mctx). If the task runs on a wrapper, execution returns to the
* wrapper loop in either case.
* This way, in the optimal case, task switching only requires a single context switch
* instead of two.
*
* @param t the current task context
*/
void LpelWorkerDispatcher( lpel_task_t *t)
{
workerctx_t *wc = t->worker_context;
/* dependent of worker or wrapper */
if (wc->wid != -1) {
lpel_task_t *next;
/* before picking the next task, process messages to consider
* also newly arrived READY tasks
*/
FetchAllMessages( wc);
/* before executing a task, handle all pending requests! */
LpelSpmdHandleRequests(wc->wid);
next = LpelSchedFetchReady( wc->sched);
if (next != NULL) {
/* short circuit */
if (next==t) { return; }
/* execute task */
wc->current_task = next;
mctx_switch(&t->mctx, &next->mctx); /*SWITCH*/
} else {
/* no ready task! -> back to worker context */
wc->current_task = NULL;
mctx_switch(&t->mctx, &wc->mctx); /*SWITCH*/
}
} else {
/* we are on a wrapper.
* back to wrapper context
*/
wc->current_task = NULL;
mctx_switch(&t->mctx, &wc->mctx); /*SWITCH*/
/* nothing to finalize on a wrapper */
}
/*********************************
* ... CTX SWITCH ...
*********************************/
/* let task continue its business ... */
}
/* ----------------
* create_expr_ctx
*
* Create a context for expression evaluation within an EState.
*
* An executor run may require multiple ExprContexts (we usually make one
* for each plan_n node, and a separate one for per-output-tuple processing
* such as constraint checking). Each expr_ctx_n has its own "per-tuple"
* memory context.
*
* Note we make no assumption about the caller's memory context.
* ----------------
*/
expr_ctx_n*
create_expr_ctx(exec_state_n *estate)
{
expr_ctx_n *econtext;
struct mctx * oldcontext;
/* Create the expr_ctx_n node within the per-query memory context */
oldcontext = mctx_switch(estate->es_query_cxt);
econtext = MK_N(ExprContext,expr_ctx_n);
/* Initialize fields of expr_ctx_n */
econtext->ecxt_scantuple = NULL;
econtext->ecxt_innertuple = NULL;
econtext->ecxt_outertuple = NULL;
econtext->ecxt_per_query_memory = estate->es_query_cxt;
/*
* Create working memory for expression evaluation in this context.
*/
econtext->ecxt_per_tuple_memory = aset_create_normal(estate->es_query_cxt, "ExprContext");
econtext->ecxt_param_exec_vals = estate->es_param_exec_vals;
econtext->ecxt_param_list_info = estate->es_param_list_info;
econtext->ecxt_aggvalues = NULL;
econtext->ecxt_aggnulls = NULL;
econtext->caseValue_datum = (datum_t) 0;
econtext->caseValue_isNull = true;
econtext->domainValue_datum = (datum_t) 0;
econtext->domainValue_isNull = true;
econtext->ecxt_estate = estate;
econtext->ecxt_callbacks = NULL;
/*
* Link the expr_ctx_n into the exec_state_n to ensure it is shut down when the
* exec_state_n is freed. Because we use lcons(), shutdowns will occur in
* reverse order of creation, which may not be essential but can't hurt.
*/
estate->es_exprcontexts = lcons(econtext, estate->es_exprcontexts);
mctx_switch(oldcontext);
return econtext;
}
/*
* regexp_matches()
* Return a table of matches of a pattern within a string.
*/
datum_t regexp_matches(PG_FUNC_ARGS)
{
struct fcall_ctx *funcctx;
regexp_matches_ctx *matchctx;
if (SRF_IS_FIRSTCALL()) {
text *pattern = ARG_TEXT_PP(1);
text *flags = ARG_TEXT_PP_IF_EXISTS(2);
struct mctx * oldcontext;
funcctx = SRF_FIRSTCALL_INIT();
oldcontext = mctx_switch(funcctx->multi_call_memory_ctx);
/* be sure to copy the input string into the multi-call ctx */
matchctx = setup_regexp_matches(ARG_TEXT_P_COPY(0), pattern,
flags, PG_COLLATION(), false, true, false);
/* Pre-create workspace that build_regexp_matches_result needs */
matchctx->elems = (datum_t *) palloc(sizeof(datum_t) * matchctx->npatterns);
matchctx->nulls = (bool *) palloc(sizeof(bool) * matchctx->npatterns);
mctx_switch(oldcontext);
funcctx->user_fctx = (void *)matchctx;
}
funcctx = SRF_PERCALL_SETUP();
matchctx = (regexp_matches_ctx *) funcctx->user_fctx;
if (matchctx->next_match < matchctx->nmatches) {
array_s *result_ary;
result_ary = build_regexp_matches_result(matchctx);
matchctx->next_match++;
SRF_RETURN_NEXT(funcctx, PTR_TO_D(result_ary));
}
/* release space in multi-call ctx to avoid intraquery memory leak */
cleanup_regexp_matches(matchctx);
SRF_RETURN_DONE(funcctx);
}
void LpelWorkerTaskBlock(lpel_task_t *t){
workerctx_t *wc = t->worker_context;
if (wc->wid < 0) { //wrapper
wc->current_task = NULL;
} else {
WORKER_DBG("worker %d: block task %d\n", wc->wid, t->uid);
//sendUpdatePrior(t); //update prior for neighbor
requestTask(wc);
}
wc->current_task = NULL;
mctx_switch(&t->mctx, &wc->mctx); // switch back to the worker/wrapper
}
void LpelWorkerTaskExit(lpel_task_t *t) {
workerctx_t *wc = t->worker_context;
WORKER_DBG("worker %d: task %d exit\n", wc->wid, t->uid);
if (wc->wid >= 0) {
requestTask(wc); // FIXME: should have requested before
wc->current_task = NULL;
}
else
wc->terminate = 1; // wrapper: terminate
mctx_switch(&t->mctx, &wc->mctx); // switch back to the worker
}
void LpelWorkerTaskYield(lpel_task_t *t){
workerctx_t *wc = t->worker_context;
if (wc->wid < 0) { //wrapper
WORKER_DBG("wrapper: task %d yields\n", t->uid);
}
else {
//sendUpdatePrior(t); //update prior for neighbor
requestTask(wc);
WORKER_DBG("worker %d: return task %d\n", wc->wid, t->uid);
wc->current_task = NULL;
}
mctx_switch(&t->mctx, &wc->mctx); // switch back to the worker/wrapper
}
/* --------------------------------
* exec_fetch_slot_min_tuple
* Fetch the slot's minimal physical tuple.
*
* If the slot contains a virtual tuple, we convert it to minimal
* physical form. The slot retains ownership of the minimal tuple.
* If it contains a regular tuple we convert to minimal form and store
* that in addition to the regular tuple (not instead of, because
* callers may hold pointers to Datums within the regular tuple).
*
* As above, the result must be treated as read-only.
* --------------------------------
*/
struct min_tuple*
exec_fetch_slot_min_tuple(struct tupslot *slot)
{
struct mctx* oldContext;
/*
* sanity checks
*/
ASSERT(slot != NULL);
ASSERT(!slot->tts_isempty);
/*
* If we have a minimal physical tuple (local or not) then just return it.
*/
if (slot->tts_mintuple)
return slot->tts_mintuple;
/*
* Otherwise, copy or build a minimal tuple, and store it into the slot.
*
* We may be called in a context that is shorter-lived than the tuple
* slot, but we have to ensure that the materialized tuple will survive
* anyway.
*/
oldContext = mctx_switch(slot->tts_mcxt);
slot->tts_mintuple = exec_copy_slot_mintup(slot);
slot->tts_shouldFreeMin = true;
mctx_switch(oldContext);
/*
* Note: we may now have a situation where we have a local minimal tuple
* attached to a virtual or non-local physical tuple. There seems no harm
* in that at the moment, but if any materializes, we should change this
* function to force the slot into minimal-tuple-only state.
*/
return slot->tts_mintuple;
}
/*
* regexp_split_to_table()
* Split the string at matches of the pattern, returning the
* split-out substrings as a table.
*/
datum_t regexp_split_to_table(PG_FUNC_ARGS)
{
struct fcall_ctx *funcctx;
regexp_matches_ctx *splitctx;
if (SRF_IS_FIRSTCALL()) {
text *pattern = ARG_TEXT_PP(1);
text *flags = ARG_TEXT_PP_IF_EXISTS(2);
struct mctx * oldcontext;
funcctx = SRF_FIRSTCALL_INIT();
oldcontext = mctx_switch(funcctx->multi_call_memory_ctx);
/* be sure to copy the input string into the multi-call ctx */
splitctx = setup_regexp_matches(ARG_TEXT_P_COPY(0), pattern,
flags, PG_COLLATION(), true, false, true);
mctx_switch(oldcontext);
funcctx->user_fctx = (void*)splitctx;
}
funcctx = SRF_PERCALL_SETUP();
splitctx = (regexp_matches_ctx*) funcctx->user_fctx;
if (splitctx->next_match <= splitctx->nmatches) {
datum_t result = build_regexp_split_result(splitctx);
splitctx->next_match++;
SRF_RETURN_NEXT(funcctx, result);
}
/* release space in multi-call ctx to avoid intraquery memory leak */
cleanup_regexp_matches(splitctx);
SRF_RETURN_DONE(funcctx);
}
/*
* executor_start hook: start up logging if needed
*/
static void
explain_executor_start(struct qry_desc *queryDesc, int eflags)
{
if (auto_explain_enabled())
{
/* Enable per-node instrumentation iff log_analyze is required. */
if (auto_explain_log_analyze && (eflags & EXEC_FLAG_EXPLAIN_ONLY) == 0)
{
queryDesc->instrument_options |= INSTRUMENT_TIMER;
if (auto_explain_log_buffers)
queryDesc->instrument_options |= INSTRUMENT_BUFFERS;
}
}
if (prev_executor_start)
prev_executor_start(queryDesc, eflags);
else
standard_executor_start(queryDesc, eflags);
if (auto_explain_enabled())
{
/*
* Set up to track total elapsed time in executor_run. Make sure the
* space is allocated in the per-query context so it will go away at
* executor_end.
*/
if (queryDesc->totaltime == NULL)
{
struct mctx * oldcxt;
oldcxt = mctx_switch(queryDesc->estate->es_query_cxt);
queryDesc->totaltime = instr_alloc(1, INSTRUMENT_ALL);
mctx_switch(oldcxt);
}
}
}
static void WorkerLoop(workerctx_t *wc)
{
WORKER_DBG("start worker %d\n", wc->wid);
lpel_task_t *t = NULL;
requestTask(wc); // ask for the first time
workermsg_t msg;
do {
LpelMailboxRecv(wc->mailbox, &msg);
switch(msg.type) {
case WORKER_MSG_ASSIGN:
t = msg.body.task;
WORKER_DBG("worker %d: get task %d\n", wc->wid, t->uid);
assert(t->state == TASK_READY);
t->worker_context = wc;
wc->current_task = t;
#ifdef USE_LOGGING
if (wc->mon && MON_CB(worker_waitstop)) {
MON_CB(worker_waitstop)(wc->mon);
}
if (t->mon && MON_CB(task_assign)) {
MON_CB(task_assign)(t->mon, wc->mon);
}
#endif
mctx_switch(&wc->mctx, &t->mctx);
//task return here
assert(t->state != TASK_RUNNING);
// if (t->state != TASK_ZOMBIE) {
wc->current_task = NULL;
t->worker_context = NULL;
returnTask(t);
// } else
// LpelTaskDestroy(t); // if task finish, destroy it and not return to master
break;
case WORKER_MSG_TERMINATE:
wc->terminate = 1;
break;
default:
assert(0);
break;
}
// reach here --> message request for task has been sent
} while (!(wc->terminate) );
}
static void WrapperLoop( workerctx_t *wc)
{
lpel_task_t *t = NULL;
do {
t = wc->wraptask;
if (t != NULL) {
/* execute task */
wc->current_task = t;
wc->wraptask = NULL;
mctx_switch(&wc->mctx, &t->mctx);
} else {
/* no ready tasks */
WaitForNewMessage( wc);
}
/* fetch (remaining) messages */
FetchAllMessages( wc);
} while ( !wc->terminate);
}
void switch_to_scheduler()
{
th_queue_t* foundQueueTh;
th_queue_t* queueSche;
int tid = getpid();
if((queueSche = find_thread_by_tid(sched_queueHead, tid)) == NULL)
{
fprintf(stderr, "Can't find the scheduler, kernel id = %d\n", getpid());
abort();
}
if((foundQueueTh = find_thread_by_tid(ready_queueHead, queueSche->thread->current_tid)) == NULL)
{
fprintf(stderr, "Can't find the thread, thread id = %d\n", queueSche->thread->current_tid);
abort();
}
//printf("%d is on duty, from %d\n", tid, foundQueueTh->thread->tid);fflush(stdout);
mctx_switch(&(foundQueueTh->thread->mctx), &(queueSche->thread->mctx));
}
datum_t
heap_page_items(PG_FUNC_ARGS)
{
bytea *raw_page = ARG_BYTEA_P(0);
heap_page_items_state *inter_call_data = NULL;
struct fcall_ctx *fctx;
int raw_page_size;
if (!superuser())
ereport(ERROR,
(errcode(E_INSUFFICIENT_PRIVILEGE),
(errmsg("must be superuser to use raw page functions"))));
raw_page_size = VLA_SZ(raw_page) - VAR_HDR_SZ;
if (SRF_IS_FIRSTCALL())
{
struct tuple * tupdesc;
struct mctx * mctx;
if (raw_page_size < PAGE_HDR_SZ)
ereport(ERROR,
(errcode(E_INVALID_PARAMETER_VALUE),
errmsg("input page too small (%d bytes)", raw_page_size)));
fctx = SRF_FIRSTCALL_INIT();
mctx = mctx_switch(fctx->multi_call_memory_ctx);
inter_call_data = palloc(sizeof(heap_page_items_state));
/* Build a tuple descriptor for our result type */
if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
inter_call_data->tupd = tupdesc;
inter_call_data->offset = FIRST_ITEM_ID;
inter_call_data->page = VLA_DATA(raw_page);
fctx->max_calls = PAGE_MAX_ITEM_ID(inter_call_data->page);
fctx->user_fctx = inter_call_data;
mctx_switch(mctx);
}
fctx = SRF_PERCALL_SETUP();
inter_call_data = fctx->user_fctx;
if (fctx->call_cntr < fctx->max_calls)
{
page_p page = inter_call_data->page;
struct heap_tuple * resultTuple;
datum_t result;
struct item_id * id;
datum_t values[13];
bool nulls[13];
uint16 lp_offset;
uint16 lp_flags;
uint16 lp_len;
memset(nulls, 0, sizeof(nulls));
/* Extract information from the line pointer */
id = PAGE_ITEM_ID(page, inter_call_data->offset);
lp_offset = ITEMID_OFFSET(id);
lp_flags = ITEMID_FLAGS(id);
lp_len = ITEMID_LENGTH(id);
values[0] = UINT16_TO_D(inter_call_data->offset);
values[1] = UINT16_TO_D(lp_offset);
values[2] = UINT16_TO_D(lp_flags);
values[3] = UINT16_TO_D(lp_len);
/*
* We do just enough validity checking to make sure we don't reference
* data outside the page passed to us. The page could be corrupt in
* many other ways, but at least we won't crash.
*/
if (ITEMID_HAS_STORAGE(id) &&
lp_len >= sizeof(struct htup_header *) &&
lp_offset == MAX_ALIGN(lp_offset) &&
lp_offset + lp_len <= raw_page_size)
{
struct htup_header * tuphdr;
int bits_len;
/* Extract information from the tuple header */
tuphdr = (struct htup_header *) PAGE_GET_ITEM(page, id);
values[4] = UINT32_TO_D(HTH_GET_XMIN(tuphdr));
values[5] = UINT32_TO_D(HTH_GET_XMAX(tuphdr));
values[6] = UINT32_TO_D(HTH_GET_RAW_CMDID(tuphdr)); /* shared with xvac */
values[7] = PTR_TO_D(&tuphdr->t_ctid);
values[8] = UINT32_TO_D(tuphdr->t_infomask2);
values[9] = UINT32_TO_D(tuphdr->t_infomask);
values[10] = UINT8_TO_D(tuphdr->t_hoff);
/*
* We already checked that the item as is completely within the
* raw page passed to us, with the length given in the line
* pointer.. Let's check that t_hoff doesn't point over lp_len,
* before using it to access t_bits and oid.
*/
if (tuphdr->t_hoff >= sizeof(struct htup_header *) &&
tuphdr->t_hoff <= lp_len)
{
if (tuphdr->t_infomask & HEAP_HAS_NULL)
{
bits_len = tuphdr->t_hoff -
(((char *) tuphdr->t_bits) -((char *) tuphdr));
values[11] = CStringGetTextDatum(
bits_to_text(tuphdr->t_bits, bits_len * 8));
}
else
nulls[11] = true;
if (tuphdr->t_infomask & HEAP_HAS_OID)
values[12] = HTH_GET_OID(tuphdr);
else
nulls[12] = true;
}
else
{
nulls[11] = true;
nulls[12] = true;
}
}
else
{
/*
* The line pointer is not used, or it's invalid. Set the rest of
* the fields to NULL
*/
int i;
for (i = 4; i <= 12; i++)
nulls[i] = true;
}
/* Build and return the result tuple. */
resultTuple = heap_form_tuple(inter_call_data->tupd, values, nulls);
result = HeapTupleGetDatum(resultTuple);
inter_call_data->offset++;
SRF_RETURN_NEXT(fctx, result);
}
else
SRF_RETURN_DONE(fctx);
}
datum_t
bt_page_items(PG_FUNC_ARGS)
{
text *relname = ARG_TEXT_P(0);
uint32 blkno = ARG_UINT32(1);
datum_t result;
char *values[6];
struct heap_tuple * tuple;
struct fcall_ctx *fctx;
struct mctx * mctx;
struct user_args *uargs;
if (!superuser())
ereport(ERROR,
(errcode(E_INSUFFICIENT_PRIVILEGE),
(errmsg("must be superuser to use pageinspect functions"))));
if (SRF_IS_FIRSTCALL())
{
range_var_n *relrv;
struct relation * rel;
buf_id_t buffer;
struct bt_page_opaque * opaque;
struct tuple * tupleDesc;
fctx = SRF_FIRSTCALL_INIT();
relrv = nl_to_range_var(textToQualifiedNameList(relname));
rel = relation_openrv(relrv, ACCESS_SHR_LOCK);
if (!IS_INDEX(rel) || !IS_BTREE(rel))
elog(ERROR, "relation \"%s\" is not a btree index",
REL_NAME(rel));
/*
* Reject attempts to read non-local temporary relations; we would be
* likely to get wrong data since we have no visibility into the
* owning session's local buffers.
*/
if (REL_IS_OTHER_TMP(rel))
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("cannot access temporary tables of other sessions")));
if (blkno == 0)
elog(ERROR, "block 0 is a meta page");
CHECK_RELATION_BLOCK_RANGE(rel, blkno);
buffer = read_buf(rel, blkno);
/*
* We copy the page into local storage to avoid holding pin on the
* buffer longer than we must, and possibly failing to release it at
* all if the calling query doesn't fetch all rows.
*/
mctx = mctx_switch(fctx->multi_call_memory_ctx);
uargs = palloc(sizeof(struct user_args));
uargs->page = palloc(BLK_SZ);
memcpy(uargs->page, BUF_PAGE(buffer), BLK_SZ);
release_buf(buffer);
relation_close(rel, ACCESS_SHR_LOCK);
uargs->offset = FIRST_ITEM_ID;
opaque = (struct bt_page_opaque *) PAGE_SPECIAL_PTR(uargs->page);
if (P_ISDELETED(opaque))
elog(NOTICE, "page is deleted");
fctx->max_calls = PAGE_MAX_ITEM_ID(uargs->page);
/* Build a tuple descriptor for our result type */
if (get_call_result_type(fcinfo, NULL, &tupleDesc) != TYPEFUNC_COMPOSITE)
elog(ERROR, "return type must be a row type");
fctx->attinmeta = TupleDescGetAttInMetadata(tupleDesc);
fctx->user_fctx = uargs;
mctx_switch(mctx);
}
fctx = SRF_PERCALL_SETUP();
uargs = fctx->user_fctx;
if (fctx->call_cntr < fctx->max_calls)
{
struct item_id * id;
struct index_tuple * itup;
int j;
int off;
int dlen;
char *dump;
char *ptr;
id = PAGE_ITEM_ID(uargs->page, uargs->offset);
if (!ITEMID_VALID(id))
elog(ERROR, "invalid ItemId");
itup = (struct index_tuple *) PAGE_GET_ITEM(uargs->page, id);
j = 0;
values[j] = palloc(32);
snprintf(values[j++], 32, "%d", uargs->offset);
values[j] = palloc(32);
snprintf(values[j++], 32, "(%u,%u)",
BLK_ID_TO_BLK_NR(&(itup->t_tid.ip_blkid)),
itup->t_tid.ip_posid);
values[j] = palloc(32);
snprintf(values[j++], 32, "%d", (int) INDEX_TUPLE_SZ(itup));
values[j] = palloc(32);
snprintf(values[j++], 32, "%c", INDEX_TUPLE_HAS_NULLS(itup) ? 't' : 'f');
values[j] = palloc(32);
snprintf(values[j++], 32, "%c", INDEX_TUPLE_HAS_VAR(itup) ? 't' : 'f');
ptr = (char *) itup + INDEX_TUPLE_DATA_OFFSET(itup->t_info);
dlen = INDEX_TUPLE_SZ(itup) - INDEX_TUPLE_DATA_OFFSET(itup->t_info);
dump = pzalloc(dlen * 3 + 1);
values[j] = dump;
for (off = 0; off < dlen; off++)
{
if (off > 0)
*dump++ = ' ';
sprintf(dump, "%02x", *(ptr + off) & 0xff);
dump += 2;
}
tuple = build_tuple_from_cstrings(fctx->attinmeta, values);
result = HeapTupleGetDatum(tuple);
uargs->offset = uargs->offset + 1;
SRF_RETURN_NEXT(fctx, result);
}
else
{
pfree(uargs->page);
pfree(uargs);
SRF_RETURN_DONE(fctx);
}
}
void* scheduler(void* p)
{
//這裡還要加上signal mask防止interrupt過來
th_queue_t* queueSche;
th_queue_t* nextQueueTh;
// queueSche = th_queue_head(ready_queueHead);
// srand(time(NULL)); //seed rand()
//disable_timer();
pthread_mutex_lock(&output_lock);
pthread_mutex_lock(&output_lock);
printf("I`m in %d, %d\n", getpid(), sched_queueHead==NULL);fflush(stdout);
pthread_mutex_unlock(&output_lock);
for(;;) //infinite loop
{
if((queueSche = find_thread_by_tid(sched_queueHead, getpid())) == NULL)
abort();
pthread_mutex_lock(&kill_lock);
if(wait_for_kill>0)
{
if(getpid() != main_kernel_id)
{
th_queue_delete(sched_queueHead, queueSche->thread);
wait_for_kill--;
pthread_mutex_unlock(&kill_lock);
pthread_mutex_lock(&output_lock);
printf("kill %d, wait = %d\n", getpid(),wait_for_kill );fflush(stdout);
pthread_mutex_unlock(&output_lock);
exit(0);
abort();
}
}
pthread_mutex_unlock(&kill_lock);
pthread_mutex_lock(&findthread_lock);
for(;;)
{
if((nextQueueTh = find_next_thread(ready_queueHead, queueSche->thread->current_tid)) == NULL)
abort();
queueSche->thread->current_tid = nextQueueTh->thread->tid;
if(nextQueueTh->thread->mctx.status == TH_WAITING)
{
break;
}
}
pthread_mutex_lock(&output_lock);
printf("I`m %d find out %d, %dn", getpid(), nextQueueTh->thread->tid, nextQueueTh->thread->mctx.status);fflush(stdout);
pthread_mutex_unlock(&output_lock);
nextQueueTh->thread->mctx.status = TH_RUNNING;
queueSche->thread->current_tid = nextQueueTh->thread->tid;
pthread_mutex_unlock(&findthread_lock);
enable_timer();
mctx_switch(&(queueSche->thread->mctx), &(nextQueueTh->thread->mctx));
if(nextQueueTh->thread->mctx.status !=TH_EXITED && \
nextQueueTh->thread->mctx.status !=TH_KILLED)
nextQueueTh->thread->mctx.status = TH_WAITING;
pthread_mutex_lock(&output_lock);
//printf("I`m %d comeback from %d\n", getpid(), nextQueueTh->thread->tid);fflush(stdout);
pthread_mutex_unlock(&output_lock);
/*
mctx_list[currentTid].status=TH_RUNNING;
mctx_switch(&mctx_list[0],&mctx_list[currentTid]);
*/
}
return NULL;
}
datum_t
connectby_text(PG_FUNC_ARGS)
{
char *relname = text_to_cstring(ARG_TEXT_PP(0));
char *key_fld = text_to_cstring(ARG_TEXT_PP(1));
char *parent_key_fld = text_to_cstring(ARG_TEXT_PP(2));
char *start_with = text_to_cstring(ARG_TEXT_PP(3));
int max_depth = ARG_INT32(4);
char *branch_delim = NULL;
bool show_branch = false;
bool show_serial = false;
return_set_info_n *rsinfo = (return_set_info_n *) fcinfo->resultinfo;
struct tuple * tupdesc;
AttInMetadata *attinmeta;
struct mctx * per_query_ctx;
struct mctx * oldcontext;
/* check to see if caller supports us returning a tuplestore */
if (rsinfo == NULL || !IS_A(rsinfo, ReturnSetInfo))
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("set-valued function called in context that cannot accept a set")));
if (!(rsinfo->allowedModes & SFRM_Materialize) ||
rsinfo->expectedDesc == NULL)
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("materialize mode required, but it is not " \
"allowed in this context")));
if (fcinfo->nargs == 6)
{
branch_delim = text_to_cstring(ARG_TEXT_PP(5));
show_branch = true;
}
else
/* default is no show, tilde for the delimiter */
branch_delim = pstrdup("~");
per_query_ctx = rsinfo->econtext->ecxt_per_query_memory;
oldcontext = mctx_switch(per_query_ctx);
/* get the requested return tuple description */
tupdesc = tupdesc_copy(rsinfo->expectedDesc);
/* does it meet our needs */
validateConnectbyTupleDesc(tupdesc, show_branch, show_serial);
/* OK, use it then */
attinmeta = TupleDescGetAttInMetadata(tupdesc);
/* OK, go to work */
rsinfo->returnMode = SFRM_Materialize;
rsinfo->setResult = connectby(relname,
key_fld,
parent_key_fld,
NULL,
branch_delim,
start_with,
max_depth,
show_branch,
show_serial,
per_query_ctx,
rsinfo->allowedModes & SFRM_Materialize_Random,
attinmeta);
rsinfo->setDesc = tupdesc;
mctx_switch(oldcontext);
/*
* SFRM_Materialize mode expects us to return a NULL Datum. The actual
* tuples are in our tuplestore and passed back through rsinfo->setResult.
* rsinfo->setDesc is set to the tuple description that we actually used
* to build our tuples with, so the caller can verify we did what it was
* expecting.
*/
return (datum_t) 0;
}
/*
* load up the categories hash table
*/
static struct htab *
load_categories_hash(char *cats_sql, struct mctx * per_query_ctx)
{
struct htab *crosstab_hash;
struct hash_ctl ctl;
int ret;
int proc;
struct mctx * SPIcontext;
/* initialize the category hash table */
pg_memset(&ctl, 0, sizeof(ctl));
ctl.keysize = MAX_CATNAME_LEN;
ctl.entrysize = sizeof(crosstab_HashEnt);
ctl.hcxt = per_query_ctx;
/*
* use INIT_CATS, defined above as a guess of how many hash table entries
* to create, initially
*/
crosstab_hash = hcreate("crosstab hash",
INIT_CATS,
&ctl,
H_ELEM | HASH_CONTEXT);
/* Connect to SPI manager */
if ((ret = SPI_connect()) < 0)
/* internal error */
elog(ERROR, "load_categories_hash: SPI_connect returned %d", ret);
/* Retrieve the category name rows */
ret = SPI_execute(cats_sql, true, 0);
proc = SPI_processed;
/* Check for qualifying tuples */
if ((ret == SPI_OK_SELECT) && (proc > 0))
{
struct SPI_tuple_table *spi_tuptable = SPI_tuptable;
struct tuple * spi_tupdesc = spi_tuptable->tupdesc;
int i;
/*
* The provided categories SQL query must always return one column:
* category - the label or identifier for each column
*/
if (spi_tupdesc->natts != 1)
ereport(ERROR,
(errcode(E_SYNTAX_ERROR),
errmsg("provided \"categories\" SQL must " \
"return 1 column of at least one row")));
for (i = 0; i < proc; i++)
{
crosstab_cat_desc *catdesc;
char *catname;
struct heap_tuple * spi_tuple;
/* get the next sql result tuple */
spi_tuple = spi_tuptable->vals[i];
/* get the category from the current sql result tuple */
catname = SPI_getvalue(spi_tuple, spi_tupdesc, 1);
SPIcontext = mctx_switch(per_query_ctx);
catdesc = (crosstab_cat_desc *) palloc(sizeof(crosstab_cat_desc));
catdesc->catname = catname;
catdesc->attidx = i;
/* Add the proc description block to the hashtable */
crosstab_HashTableInsert(crosstab_hash, catdesc);
mctx_switch(SPIcontext);
}
}
if (SPI_finish() != SPI_OK_FINISH)
/* internal error */
elog(ERROR, "load_categories_hash: SPI_finish() failed");
return crosstab_hash;
}
datum_t
crosstab_hash(PG_FUNC_ARGS)
{
char *sql = text_to_cstring(ARG_TEXT_PP(0));
char *cats_sql = text_to_cstring(ARG_TEXT_PP(1));
return_set_info_n *rsinfo = (return_set_info_n *) fcinfo->resultinfo;
struct tuple * tupdesc;
struct mctx * per_query_ctx;
struct mctx * oldcontext;
struct htab *crosstab_hash;
/* check to see if caller supports us returning a tuplestore */
if (rsinfo == NULL || !IS_A(rsinfo, ReturnSetInfo))
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("set-valued function called in context that cannot accept a set")));
if (!(rsinfo->allowedModes & SFRM_Materialize) ||
rsinfo->expectedDesc == NULL)
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("materialize mode required, but it is not " \
"allowed in this context")));
per_query_ctx = rsinfo->econtext->ecxt_per_query_memory;
oldcontext = mctx_switch(per_query_ctx);
/* get the requested return tuple description */
tupdesc = tupdesc_copy(rsinfo->expectedDesc);
/*
* Check to make sure we have a reasonable tuple descriptor
*
* Note we will attempt to coerce the values into whatever the return
* attribute type is and depend on the "in" function to complain if
* needed.
*/
if (tupdesc->natts < 2)
ereport(ERROR,
(errcode(E_SYNTAX_ERROR),
errmsg("query-specified return tuple and " \
"crosstab function are not compatible")));
/* load up the categories hash table */
crosstab_hash = load_categories_hash(cats_sql, per_query_ctx);
/* let the caller know we're sending back a tuplestore */
rsinfo->returnMode = SFRM_Materialize;
/* now go build it */
rsinfo->setResult = get_crosstab_tuplestore(sql,
crosstab_hash,
tupdesc,
per_query_ctx,
rsinfo->allowedModes & SFRM_Materialize_Random);
/*
* SFRM_Materialize mode expects us to return a NULL Datum. The actual
* tuples are in our tuplestore and passed back through rsinfo->setResult.
* rsinfo->setDesc is set to the tuple description that we actually used
* to build our tuples with, so the caller can verify we did what it was
* expecting.
*/
rsinfo->setDesc = tupdesc;
mctx_switch(oldcontext);
return (datum_t) 0;
}
datum_t
crosstab(PG_FUNC_ARGS)
{
char *sql = text_to_cstring(ARG_TEXT_PP(0));
return_set_info_n *rsinfo = (return_set_info_n *) fcinfo->resultinfo;
struct tupstore *tupstore;
struct tuple * tupdesc;
int call_cntr;
int max_calls;
AttInMetadata *attinmeta;
struct SPI_tuple_table *spi_tuptable;
struct tuple * spi_tupdesc;
bool firstpass;
char *lastrowid;
int i;
int num_categories;
struct mctx * per_query_ctx;
struct mctx * oldcontext;
int ret;
int proc;
/* check to see if caller supports us returning a tuplestore */
if (rsinfo == NULL || !IS_A(rsinfo, ReturnSetInfo))
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("set-valued function called in context that cannot accept a set")));
if (!(rsinfo->allowedModes & SFRM_Materialize))
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("materialize mode required, but it is not " \
"allowed in this context")));
per_query_ctx = rsinfo->econtext->ecxt_per_query_memory;
/* Connect to SPI manager */
if ((ret = SPI_connect()) < 0)
/* internal error */
elog(ERROR, "crosstab: SPI_connect returned %d", ret);
/* Retrieve the desired rows */
ret = SPI_execute(sql, true, 0);
proc = SPI_processed;
/* If no qualifying tuples, fall out early */
if (ret != SPI_OK_SELECT || proc <= 0)
{
SPI_finish();
rsinfo->isDone = ExprEndResult;
RET_NULL();
}
spi_tuptable = SPI_tuptable;
spi_tupdesc = spi_tuptable->tupdesc;
/*----------
* The provided SQL query must always return three columns.
*
* 1. rowname
* the label or identifier for each row in the final result
* 2. category
* the label or identifier for each column in the final result
* 3. values
* the value for each column in the final result
*----------
*/
if (spi_tupdesc->natts != 3)
ereport(ERROR,
(errcode(E_INVALID_PARAMETER_VALUE),
errmsg("invalid source data SQL statement"),
errdetail("The provided SQL must return 3 "
"columns: rowid, category, and values.")));
/* get a tuple descriptor for our result type */
switch (get_call_result_type(fcinfo, NULL, &tupdesc))
{
case TYPEFUNC_COMPOSITE:
/* success */
break;
case TYPEFUNC_RECORD:
/* failed to determine actual type of RECORD */
ereport(ERROR,
(errcode(E_FEATURE_NOT_SUPPORTED),
errmsg("function returning record called in context "
"that cannot accept type record")));
break;
default:
/* result type isn't composite */
elog(ERROR, "return type must be a row type");
break;
}
/*
* Check that return tupdesc is compatible with the data we got from SPI,
* at least based on number and type of attributes
*/
if (!compatCrosstabTupleDescs(tupdesc, spi_tupdesc))
ereport(ERROR,
(errcode(E_SYNTAX_ERROR),
errmsg("return and sql tuple descriptions are " \
"incompatible")));
/*
* switch to long-lived memory context
*/
oldcontext = mctx_switch(per_query_ctx);
/* make sure we have a persistent copy of the result tupdesc */
tupdesc = tupdesc_copy(tupdesc);
/* initialize our tuplestore in long-lived context */
tupstore =
tts_begin_heap(rsinfo->allowedModes & SFRM_Materialize_Random,
false, work_mem);
mctx_switch(oldcontext);
/*
* Generate attribute metadata needed later to produce tuples from raw C
* strings
*/
attinmeta = TupleDescGetAttInMetadata(tupdesc);
/* total number of tuples to be examined */
max_calls = proc;
/* the return tuple always must have 1 rowid + num_categories columns */
num_categories = tupdesc->natts - 1;
firstpass = true;
lastrowid = NULL;
for (call_cntr = 0; call_cntr < max_calls; call_cntr++)
{
bool skip_tuple = false;
char **values;
/* allocate and zero space */
values = (char **) pzalloc((1 + num_categories) * sizeof(char *));
/*
* now loop through the sql results and assign each value in sequence
* to the next category
*/
for (i = 0; i < num_categories; i++)
{
struct heap_tuple * spi_tuple;
char *rowid;
/* see if we've gone too far already */
if (call_cntr >= max_calls)
break;
/* get the next sql result tuple */
spi_tuple = spi_tuptable->vals[call_cntr];
/* get the rowid from the current sql result tuple */
rowid = SPI_getvalue(spi_tuple, spi_tupdesc, 1);
/*
* If this is the first pass through the values for this rowid,
* set the first column to rowid
*/
if (i == 0)
{
xpstrdup(values[0], rowid);
/*
* Check to see if the rowid is the same as that of the last
* tuple sent -- if so, skip this tuple entirely
*/
if (!firstpass && xstreq(lastrowid, rowid))
{
xpfree(rowid);
skip_tuple = true;
break;
}
}
/*
* If rowid hasn't changed on us, continue building the output
* tuple.
*/
if (xstreq(rowid, values[0]))
{
/*
* Get the next category item value, which is always attribute
* number three.
*
* Be careful to assign the value to the array index based on
* which category we are presently processing.
*/
values[1 + i] = SPI_getvalue(spi_tuple, spi_tupdesc, 3);
/*
* increment the counter since we consume a row for each
* category, but not for last pass because the outer loop will
* do that for us
*/
if (i < (num_categories - 1))
call_cntr++;
xpfree(rowid);
}
else
{
/*
* We'll fill in NULLs for the missing values, but we need to
* decrement the counter since this sql result row doesn't
* belong to the current output tuple.
*/
call_cntr--;
xpfree(rowid);
break;
}
}
if (!skip_tuple)
{
struct heap_tuple * tuple;
/* build the tuple and store it */
tuple = build_tuple_from_cstrings(attinmeta, values);
tts_put_tuple(tupstore, tuple);
heap_free_tuple(tuple);
}
/* Remember current rowid */
xpfree(lastrowid);
xpstrdup(lastrowid, values[0]);
firstpass = false;
/* Clean up */
for (i = 0; i < num_categories + 1; i++)
if (values[i] != NULL)
pfree(values[i]);
pfree(values);
}
/* let the caller know we're sending back a tuplestore */
rsinfo->returnMode = SFRM_Materialize;
rsinfo->setResult = tupstore;
rsinfo->setDesc = tupdesc;
/* release SPI related resources (and return to caller's context) */
SPI_finish();
return (datum_t) 0;
}
datum_t
normal_rand(PG_FUNC_ARGS)
{
struct fcall_ctx *funcctx;
int call_cntr;
int max_calls;
normal_rand_fctx *fctx;
float8 mean;
float8 stddev;
float8 carry_val;
bool use_carry;
struct mctx * oldcontext;
/* stuff done only on the first call of the function */
if (SRF_IS_FIRSTCALL())
{
/* create a function context for cross-call persistence */
funcctx = SRF_FIRSTCALL_INIT();
/*
* switch to memory context appropriate for multiple function calls
*/
oldcontext = mctx_switch(funcctx->multi_call_memory_ctx);
/* total number of tuples to be returned */
funcctx->max_calls = ARG_UINT32(0);
/* allocate memory for user context */
fctx = (normal_rand_fctx *) palloc(sizeof(normal_rand_fctx));
/*
* Use fctx to keep track of upper and lower bounds from call to call.
* It will also be used to carry over the spare value we get from the
* Box-Muller algorithm so that we only actually calculate a new value
* every other call.
*/
fctx->mean = ARG_FLOAT8(1);
fctx->stddev = ARG_FLOAT8(2);
fctx->carry_val = 0;
fctx->use_carry = false;
funcctx->user_fctx = fctx;
mctx_switch(oldcontext);
}
/* stuff done on every call of the function */
funcctx = SRF_PERCALL_SETUP();
call_cntr = funcctx->call_cntr;
max_calls = funcctx->max_calls;
fctx = funcctx->user_fctx;
mean = fctx->mean;
stddev = fctx->stddev;
carry_val = fctx->carry_val;
use_carry = fctx->use_carry;
if (call_cntr < max_calls) /* do when there is more left to send */
{
float8 result;
if (use_carry)
{
/*
* reset use_carry and use second value obtained on last pass
*/
fctx->use_carry = false;
result = carry_val;
}
else
{
float8 normval_1;
float8 normval_2;
/* Get the next two normal values */
get_normal_pair(&normval_1, &normval_2);
/* use the first */
result = mean + (stddev * normval_1);
/* and save the second */
fctx->carry_val = mean + (stddev * normval_2);
fctx->use_carry = true;
}
/* send the result */
SRF_RETURN_NEXT(funcctx, FLOAT8_TO_D(result));
}
else
/* do when there is no more left */
SRF_RETURN_DONE(funcctx);
}
datum_t gistrescan(PG_FUNC_ARGS)
{
struct index_scan *scan;
struct scankey *key;
struct scankey *orderbys;
/* nkeys and norderbys arguments are ignored */
struct gist_scan_opaque *so;
int i;
struct mctx *oldCxt;
size_t size;
scan = (struct index_scan *)ARG_POINTER(0);
key = (struct scankey *)ARG_POINTER(1);
orderbys = (struct scankey *)ARG_POINTER(3);
so = (struct gist_scan_opaque *)scan->opaque;
/* rescan an existing indexscan --- reset state */
mctx_reset(so->queueCxt);
so->curTreeItem = NULL;
/* create new, empty RBTree for search queue */
oldCxt = mctx_switch(so->queueCxt);
size = GIST_ITEM_HDR_SZ + sizeof(double) * scan->numberOfOrderBys;
so->queue = rb_create(
size,
GISTSearchTreeItemComparator,
GISTSearchTreeItemCombiner,
GISTSearchTreeItemAllocator,
GISTSearchTreeItemDeleter,
scan);
mctx_switch(oldCxt);
so->firstCall = true;
/* Update scan key, if a new one is given */
if (key && scan->numberOfKeys > 0) {
size = scan->numberOfKeys * sizeof(struct scankey);
memmove(scan->keyData, key, size);
/*
* Modify the scan key so that the Consistent method is called for all
* comparisons. The original operator is passed to the Consistent
* function in the form of its strategy number, which is available
* from the sk_strategy field, and its subtype from the sk_subtype
* field.
*
* Next, if any of keys is a NULL and that key is not marked with
* SK_SEARCHNULL/SK_SEARCHNOTNULL then nothing can be found (ie, we
* assume all indexable operators are strict).
*/
so->qual_ok = true;
for (i = 0; i < scan->numberOfKeys; i++) {
struct scankey *skey;
skey = scan->keyData + i;
skey->sk_func = so->giststate->consistentFn[skey->sk_attno - 1];
if (skey->sk_flags & SK_ISNULL) {
if (!(skey->sk_flags & (SK_SEARCHNULL | SK_SEARCHNOTNULL)))
so->qual_ok = false;
}
}
}
/* Update order-by key, if a new one is given */
if (orderbys
&& scan->numberOfOrderBys > 0) {
size = scan->numberOfOrderBys * sizeof(struct scankey);
memmove(scan->orderByData, orderbys, size);
/*
* Modify the order-by key so that the Distance method is called for
* all comparisons. The original operator is passed to the Distance
* function in the form of its strategy number, which is available
* from the sk_strategy field, and its subtype from the sk_subtype
* field.
*/
for (i = 0; i < scan->numberOfOrderBys; i++) {
struct scankey *skey;
skey = scan->orderByData + i;
skey->sk_func = so->giststate->distanceFn[skey->sk_attno - 1];
/* Check we actually have a distance function ... */
if (!OID_VALID(skey->sk_func.fn_oid))
elog(ERROR,
"missing support function %d for attribute %d"
" of index \"%s\"",
GIST_DISTANCE_PROC,
skey->sk_attno,
REL_NAME(scan->indexRelation));
}
}
RET_VOID();
}
datum_t
tsa_rewrite_accum(PG_FUNC_ARGS)
{
TSQuery acc;
array_s *qa;
TSQuery q;
QTNode *qex = NULL,
*subs = NULL,
*acctree = NULL;
bool isfind = false;
datum_t *elemsp;
int nelemsp;
struct mctx * aggcontext;
struct mctx * oldcontext;
if (!AggCheckCallContext(fcinfo, &aggcontext))
elog(ERROR, "tsa_rewrite_accum called in non-aggregate context");
if (PG_ARG_ISNULL(0) || ARG_POINTER(0) == NULL)
{
acc = (TSQuery) mctx_alloc(aggcontext, HDRSIZETQ);
VLA_SET_SZ_STND(acc, HDRSIZETQ);
acc->size = 0;
}
else
acc = ARG_TSQUERY(0);
if (PG_ARG_ISNULL(1) || ARG_POINTER(1) == NULL)
RET_TSQUERY(acc);
else
qa = ARG_ARRAY_P_COPY(1);
if (ARR_NDIM(qa) != 1)
elog(ERROR, "array must be one-dimensional, not %d dimensions",
ARR_NDIM(qa));
if (ArrayGetNItems(ARR_NDIM(qa), ARR_DIMS(qa)) != 3)
elog(ERROR, "array must have three elements");
if (ARR_ELEMTYPE(qa) != TSQUERYOID)
elog(ERROR, "array must contain tsquery elements");
deconstruct_array(qa, TSQUERYOID, -1, false, 'i', &elemsp, NULL, &nelemsp);
q = DatumGetTSQuery(elemsp[0]);
if (q->size == 0)
{
pfree(elemsp);
RET_POINTER(acc);
}
if (!acc->size)
{
if (VLA_SZ(acc) > HDRSIZETQ)
{
pfree(elemsp);
RET_POINTER(acc);
}
else
acctree = QT2QTN(GETQUERY(q), GETOPERAND(q));
}
else
acctree = QT2QTN(GETQUERY(acc), GETOPERAND(acc));
QTNTernary(acctree);
QTNSort(acctree);
q = DatumGetTSQuery(elemsp[1]);
if (q->size == 0)
{
pfree(elemsp);
RET_POINTER(acc);
}
qex = QT2QTN(GETQUERY(q), GETOPERAND(q));
QTNTernary(qex);
QTNSort(qex);
q = DatumGetTSQuery(elemsp[2]);
if (q->size)
subs = QT2QTN(GETQUERY(q), GETOPERAND(q));
acctree = findsubquery(acctree, qex, subs, &isfind);
if (isfind || !acc->size)
{
/* pfree( acc ); do not pfree(p), because nodeAgg.c will */
if (acctree)
{
QTNBinary(acctree);
oldcontext = mctx_switch(aggcontext);
acc = QTN2QT(acctree);
mctx_switch(oldcontext);
}
else
{
acc = (TSQuery) mctx_alloc(aggcontext, HDRSIZETQ);
VLA_SET_SZ_STND(acc, HDRSIZETQ);
acc->size = 0;
}
}
pfree(elemsp);
QTNFree(qex);
QTNFree(subs);
QTNFree(acctree);
RET_TSQUERY(acc);
}
/* ----------------
* create_exec_state
*
* Create and initialize an exec_state_n node, which is the root of
* working storage for an entire Executor invocation.
*
* Principally, this creates the per-query memory context that will be
* used to hold all working data that lives till the end of the query.
* Note that the per-query context will become a child of the caller's
* current_mctx.
* ----------------
*/
exec_state_n *
create_exec_state(void)
{
exec_state_n* estate;
struct mctx* qcontext;
struct mctx* oldcontext;
/*
* Create the per-query context for this Executor run.
*/
qcontext = aset_create_normal(current_mctx, "ExecutorState");
/*
* Make the exec_state_n node within the per-query context. This way, we don't
* need a separate pfree() operation for it at shutdown.
*/
oldcontext = mctx_switch(qcontext);
estate = MK_N(EState,exec_state_n);
/*
* Initialize all fields of the Executor State structure
*/
estate->es_direction = FORWARD_SCANDIR;
estate->es_snapshot = snap_now;
estate->es_crosscheck_snapshot = INVALID_SNAPSHOT; /* no crosscheck */
estate->es_range_table = NIL;
estate->es_plannedstmt = NULL;
estate->es_junkFilter = NULL;
estate->es_output_cid = (cmd_id_t) 0;
estate->es_result_relations = NULL;
estate->es_num_result_relations = 0;
estate->es_result_relation_info = NULL;
estate->es_trig_target_relations = NIL;
estate->es_trig_tuple_slot = NULL;
estate->es_trig_oldtup_slot = NULL;
estate->es_trig_newtup_slot = NULL;
estate->es_param_list_info = NULL;
estate->es_param_exec_vals = NULL;
estate->es_query_cxt = qcontext;
estate->es_tupleTable = NIL;
estate->es_rowMarks = NIL;
estate->es_processed = 0;
estate->es_lastoid = INVALID_OID;
estate->es_top_eflags = 0;
estate->es_instrument = 0;
estate->es_select_into = false;
estate->es_into_oids = false;
estate->es_finished = false;
estate->es_exprcontexts = NIL;
estate->es_subplanstates = NIL;
estate->es_auxmodifytables = NIL;
estate->es_per_tuple_exprcontext = NULL;
estate->es_epqTuple = NULL;
estate->es_epqTupleSet = NULL;
estate->es_epqScanDone = NULL;
/*
* Return the executor state structure
*/
mctx_switch(oldcontext);
return estate;
}
