Changeset: a9b777c3b1ce for MonetDB
URL: https://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=a9b777c3b1ce
Modified Files:
gdk/gdk_bat.c
gdk/gdk_batop.c
gdk/gdk_bbp.c
gdk/gdk_firstn.c
gdk/gdk_group.c
gdk/gdk_join.c
gdk/gdk_logger.c
gdk/gdk_sample.c
gdk/gdk_select.c
gdk/gdk_unique.c
sql/backends/monet5/generator/generator.c
sql/backends/monet5/sql.c
sql/storage/bat/bat_table.c
sql/storage/bat/bat_utils.c
Branch: default
Log Message:
Merge with Jul2017 branch.
diffs (truncated from 600 to 300 lines):
diff --git a/gdk/gdk_bat.c b/gdk/gdk_bat.c
--- a/gdk/gdk_bat.c
+++ b/gdk/gdk_bat.c
@@ -246,10 +246,10 @@ BATdense(oid hseq, oid tseq, BUN cnt)
BAT *bn;
bn = COLnew(hseq, TYPE_void, 0, TRANSIENT);
- if (bn == NULL)
- return NULL;
- BATtseqbase(bn, tseq);
- BATsetcount(bn, cnt);
+ if (bn != NULL) {
+ BATtseqbase(bn, tseq);
+ BATsetcount(bn, cnt);
+ }
return bn;
}
@@ -1380,6 +1380,7 @@ BATsetcount(BAT *b, BUN cnt)
{
/* head column is always VOID, and some head properties never change */
assert(b->hseqbase != oid_nil);
+ assert(cnt <= BUN_MAX);
b->batCount = cnt;
b->batDirtydesc = TRUE;
diff --git a/gdk/gdk_batop.c b/gdk/gdk_batop.c
--- a/gdk/gdk_batop.c
+++ b/gdk/gdk_batop.c
@@ -1306,21 +1306,17 @@ BATsort(BAT **sorted, BAT **order, BAT *
*sorted = bn;
}
if (order) {
- on = COLnew(b->hseqbase, TYPE_void, BATcount(b),
TRANSIENT);
+ on = BATdense(b->hseqbase, b->hseqbase, BATcount(b));
if (on == NULL)
goto error;
- BATsetcount(on, BATcount(b));
- BATtseqbase(on, b->hseqbase);
*order = on;
}
if (groups) {
if (BATtkey(b)) {
/* singleton groups */
- gn = COLnew(0, TYPE_void, BATcount(b),
TRANSIENT);
+ gn = BATdense(0, 0, BATcount(b));
if (gn == NULL)
goto error;
- BATsetcount(gn, BATcount(b));
- BATtseqbase(gn, 0);
} else {
/* single group */
const oid *o = 0;
@@ -1822,15 +1818,9 @@ BATcount_no_nil(BAT *b)
static BAT *
newdensecand(oid first, oid last)
{
- BAT *bn;
-
- if ((bn = COLnew(0, TYPE_void, 0, TRANSIENT)) == NULL)
- return NULL;
if (last < first)
first = last = 0; /* empty range */
- BATsetcount(bn, last - first);
- BATtseqbase(bn, first);
- return bn;
+ return BATdense(0, first, last - first);
}
/* merge two candidate lists and produce a new one
diff --git a/gdk/gdk_bbp.c b/gdk/gdk_bbp.c
--- a/gdk/gdk_bbp.c
+++ b/gdk/gdk_bbp.c
@@ -321,9 +321,18 @@ BBPselectfarm(int role, int type, enum h
{
int i;
- assert(role >= 0 && role < 32);
(void) type; /* may use in future */
(void) hptype; /* may use in future */
+
+ assert(role >= 0 && role < 32);
+#ifndef PERSISTENTHASH
+ if (hptype == hashheap)
+ role = TRANSIENT;
+#endif
+#ifndef PERSISTENTIDX
+ if (hptype == orderidxheap)
+ role = TRANSIENT;
+#endif
for (i = 0; i < MAXFARMS; i++)
if (BBPfarms[i].dirname && BBPfarms[i].roles & (1 << role))
return i;
@@ -1909,7 +1918,9 @@ BBPdump(void)
vm += HEAPvmsize(b->thash->heap);
}
}
- fprintf(stderr, "\n");
+ fprintf(stderr, " role: %s, persistence: %s\n",
+ b->batRole == PERSISTENT ? "persistent" : "transient",
+ b->batPersistence == PERSISTENT ? "persistent" :
"transient");
}
fprintf(stderr,
"# %d bats: mem=" SZFMT ", vm=" SZFMT " %d cached bats: mem="
SZFMT ", vm=" SZFMT "\n",
diff --git a/gdk/gdk_firstn.c b/gdk/gdk_firstn.c
--- a/gdk/gdk_firstn.c
+++ b/gdk/gdk_firstn.c
@@ -119,6 +119,10 @@
* refer to the N smallest/largest (depending on asc) tail values of b
* (taking the optional candidate list s into account). If there are
* multiple equal values to take us past N, we return a subset of those.
+ *
+ * If lastp is non-NULL, it is filled in with the oid of the "last"
+ * value, i.e. the value of which there may be multiple occurrences
+ * that are not all included in the first N.
*/
static BAT *
BATfirstn_unique(BAT *b, BAT *s, BUN n, int asc, oid *lastp)
@@ -148,11 +152,9 @@ BATfirstn_unique(BAT *b, BAT *s, BUN n,
}
} else if (n >= cnt) {
/* trivial: return everything */
- bn = COLnew(0, TYPE_void, cnt, TRANSIENT);
+ bn = BATdense(0, start + b->hseqbase, cnt);
if (bn == NULL)
return NULL;
- BATsetcount(bn, cnt);
- BATtseqbase(bn, start + b->hseqbase);
if (lastp)
*lastp = 0;
return bn;
@@ -178,18 +180,14 @@ BATfirstn_unique(BAT *b, BAT *s, BUN n,
*lastp = candend[-(ssize_t)n];
return BATslice(s, i - n, i);
}
- bn = COLnew(0, TYPE_void, n, TRANSIENT);
- if (bn == NULL)
- return NULL;
- BATsetcount(bn, n);
if (asc ? b->tsorted : b->trevsorted) {
/* first n entries from b */
- BATtseqbase(bn, start + b->hseqbase);
+ bn = BATdense(0, start + b->hseqbase, n);
if (lastp)
*lastp = start + b->hseqbase + n - 1;
} else {
/* last n entries from b */
- BATtseqbase(bn, start + cnt + b->hseqbase - n);
+ bn = BATdense(0, start + cnt + b->hseqbase - n, n);
if (lastp)
*lastp = start + cnt + b->hseqbase - n;
}
@@ -370,6 +368,17 @@ BATfirstn_unique(BAT *b, BAT *s, BUN n,
} \
} while (0)
+/* This version of BATfirstn is like the one above, except that it
+ * also looks at groups. The values of the group IDs are important:
+ * we return only the smallest N (i.e., not dependent on asc which
+ * refers only to the values in the BAT b).
+ *
+ * If lastp is non-NULL, it is filled in with the oid of the "last"
+ * value, i.e. the value of which there may be multiple occurrences
+ * that are not all included in the first N. If lastgp is non-NULL,
+ * it is filled with the group ID (not the oid of the group ID) for
+ * that same value.
+ */
static BAT *
BATfirstn_unique_with_groups(BAT *b, BAT *s, BAT *g, BUN n, int asc, oid
*lastp, oid *lastgp)
{
@@ -401,11 +410,7 @@ BATfirstn_unique_with_groups(BAT *b, BAT
if (n == 0) {
/* candidate list might refer only to values outside
* of the bat and hence be effectively empty */
- bn = COLnew(0, TYPE_void, 0, TRANSIENT);
- if (bn == NULL)
- return NULL;
- BATtseqbase(bn, 0);
- return bn;
+ return BATdense(0, 0, 0);
}
bn = COLnew(0, TYPE_oid, n, TRANSIENT);
@@ -581,7 +586,7 @@ BATfirstn_grouped(BAT **topn, BAT **gids
return GDK_FAIL;
if (BATcount(bn) == 0) {
if (gids) {
- gn = COLnew(0, TYPE_void, 0, TRANSIENT);
+ gn = BATdense(0, 0, 0);
if (gn == NULL) {
BBPunfix(bn->batCacheid);
return GDK_FAIL;
@@ -707,7 +712,7 @@ BATfirstn_grouped_with_groups(BAT **topn
}
if (BATcount(bn) == 0) {
if (gids) {
- gn = COLnew(0, TYPE_void, 0, TRANSIENT);
+ gn = BATdense(0, 0, 0);
if (gn == NULL) {
BBPunfix(bn->batCacheid);
return GDK_FAIL;
@@ -816,17 +821,15 @@ BATfirstn(BAT **topn, BAT **gids, BAT *b
if (n == 0 || BATcount(b) == 0 || (s != NULL && BATcount(s) == 0)) {
/* trivial: empty result */
- *topn = COLnew(0, TYPE_void, 0, TRANSIENT);
+ *topn = BATdense(0, 0, 0);
if (*topn == NULL)
return GDK_FAIL;
- BATtseqbase(*topn, 0);
if (gids) {
- *gids = COLnew(0, TYPE_void, 0, TRANSIENT);
+ *gids = BATdense(0, 0, 0);
if (*gids == NULL) {
BBPreclaim(*topn);
return GDK_FAIL;
}
- BATtseqbase(*gids, 0);
}
return GDK_SUCCEED;
}
diff --git a/gdk/gdk_group.c b/gdk/gdk_group.c
--- a/gdk/gdk_group.c
+++ b/gdk/gdk_group.c
@@ -599,17 +599,16 @@ BATgroup_internal(BAT **groups, BAT **ex
h ? BATgetId(h) : "NULL", h ? BATcount(h) : 0,
subsorted);
ngrp = cnt == 0 ? 0 : cand ? s->hseqbase + (cand - (const oid
*) Tloc(s, 0)) : s ? s->hseqbase + start - s->tseqbase : b->hseqbase;
- gn = COLnew(hseqb, TYPE_void, BATcount(b), TRANSIENT);
+ gn = BATdense(hseqb, 0, BATcount(b));
if (gn == NULL)
goto error;
- BATsetcount(gn, BATcount(b));
- BATtseqbase(gn, 0);
*groups = gn;
if (extents) {
if (cand) {
en = COLnew(0, TYPE_oid, cnt, TRANSIENT);
if (en == NULL)
goto error;
+ BATsetcount(en, cnt);
memcpy(Tloc(en, 0), cand, cnt * sizeof(oid));
en->tsorted = 1;
en->trevsorted = cnt <= 1;
@@ -618,12 +617,10 @@ BATgroup_internal(BAT **groups, BAT **ex
en->tnonil = 1;
en->tdense = 0;
} else {
- en = COLnew(0, TYPE_void, BATcount(b),
TRANSIENT);
+ en = BATdense(0, b->hseqbase, cnt);
if (en == NULL)
goto error;
- BATtseqbase(en, b->hseqbase + start);
}
- BATsetcount(en, cnt);
*extents = en;
}
if (histo) {
@@ -672,11 +669,9 @@ BATgroup_internal(BAT **groups, BAT **ex
*groups = gn;
if (extents) {
ngrp = b->hseqbase + start;
- en = COLnew(0, TYPE_void, 1, TRANSIENT);
+ en = BATdense(0, b->hseqbase + start, 1);
if (en == NULL)
goto error;
- BATtseqbase(en, b->hseqbase + start);
- BATsetcount(en, 1);
*extents = en;
}
if (histo) {
diff --git a/gdk/gdk_join.c b/gdk/gdk_join.c
--- a/gdk/gdk_join.c
+++ b/gdk/gdk_join.c
@@ -157,18 +157,16 @@ joininitresults(BAT **r1p, BAT **r2p, BU
}
if (maxsize == 0) {
- r1 = COLnew(0, TYPE_void, 0, TRANSIENT);
+ r1 = BATdense(0, 0, 0);
if (r1 == NULL) {
return BUN_NONE;
}
- BATtseqbase(r1, 0);
if (r2p) {
- r2 = COLnew(0, TYPE_void, 0, TRANSIENT);
+ r2 = BATdense(0, 0, 0);
if (r2 == NULL) {
BBPreclaim(r1);
return BUN_NONE;
}
- BATtseqbase(r2, 0);
*r2p = r2;
}
*r1p = r1;
@@ -3755,15 +3753,13 @@ BATjoin(BAT **r1p, BAT **r2p, BAT *l, BA
if (sr)
rcount = MIN(rcount, BATcount(sr));
if (lcount == 0 || rcount == 0) {
- r1 = COLnew(0, TYPE_void, 0, TRANSIENT);
_______________________________________________
checkin-list mailing list
[email protected]
https://www.monetdb.org/mailman/listinfo/checkin-list