Changeset: 55e394c7e78d for MonetDB
URL: https://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=55e394c7e78d
Added Files:
sql/test/merge-partitions/Tests/mergepart11.stable.err
sql/test/merge-partitions/Tests/mergepart11.stable.out
Modified Files:
sql/server/rel_propagate.c
sql/server/rel_updates.c
sql/test/merge-partitions/Tests/mergepart11.sql
Branch: merge-partitions
Log Message:
At this moment we won't be supporting updates on the partitioned column.
diffs (truncated from 548 to 300 lines):
diff --git a/sql/server/rel_propagate.c b/sql/server/rel_propagate.c
--- a/sql/server/rel_propagate.c
+++ b/sql/server/rel_propagate.c
@@ -211,7 +211,7 @@ rel_ddl_distribute(sql_allocator *sa, sq
}
static sql_rel*
-rel_propagate_delete(mvc *sql, sql_rel *rel, sql_table *t, int *changes)
+rel_generate_subdeletes(mvc *sql, sql_rel *rel, sql_table *t, int *changes)
{
int just_one = 1;
sql_rel *sel = NULL;
@@ -239,21 +239,57 @@ rel_propagate_delete(mvc *sql, sql_rel *
}
(*changes)++;
}
- return rel_ddl_distribute(sql->sa, sel, NULL, NULL);
+ return sel;
}
static sql_rel*
-rel_propagate_insert(mvc *sql, sql_rel *rel, sql_table *t, int *changes)
+rel_generate_subupdates(mvc *sql, sql_rel *rel, sql_table *t, int *changes)
+{
+ int just_one = 1;
+ sql_rel *sel = NULL;
+
+ for (node *n = t->members.set->h; n; n = n->next) {
+ sql_part *pt = (sql_part *) n->data;
+ sql_table *sub = find_sql_table(t->s, pt->base.name);
+ sql_rel *s1, *dup = NULL;
+ list *uexps = exps_copy(sql->sa, rel->exps);
+
+ if(rel->r) {
+ dup = rel_copy(sql->sa, rel->r, 1);
+ dup = rel_change_base_table(sql, dup, t, sub);
+ }
+
+ for(node *ne = uexps->h ; ne ; ne = ne->next)
+ ne->data = exp_change_column_table(sql, (sql_exp*)
ne->data, t, sub);
+
+ s1 = rel_update(sql, rel_basetable(sql, sub, sub->base.name),
dup, NULL, uexps);
+ if (just_one == 0) {
+ sel = rel_list(sql->sa, sel, s1);
+ } else {
+ sel = s1;
+ just_one = 0;
+ }
+ (*changes)++;
+ }
+
+ return sel;
+}
+
+static sql_rel*
+rel_generate_subinserts(mvc *sql, sql_rel *rel, sql_rel **anti_rel, sql_exp
**exception, sql_table *t, int *changes,
+ const char *operation, const
char *desc)
{
int colr = t->pcol->colnr, just_one = 1, found_nils = 0;
- sql_rel *anti_dup = rel_dup(rel->r) /* the anti relation */, *new_table
= NULL, *sel = NULL;
- sql_exp *anti_exp = NULL, *anti_le = list_fetch(anti_dup->exps, colr),
*accum = NULL, *aggr = NULL,
- *exception = NULL;
+ sql_rel *new_table = NULL, *sel = NULL;
+ sql_exp *anti_exp = NULL, *anti_le = NULL, *accum = NULL, *aggr = NULL;
list *anti_exps = new_exp_list(sql->sa);
sql_subaggr *cf = sql_bind_aggr(sql->sa, sql->session->schema, "count",
NULL);
+ char buf[BUFSIZ];
+
+ *anti_rel = rel_dup(rel->r);
+ anti_le = list_fetch((*anti_rel)->exps, colr);
anti_le = exp_column(sql->sa, exp_relname(anti_le), exp_name(anti_le),
exp_subtype(anti_le),
anti_le->card,
has_nil(anti_le), is_intern(anti_le));
- char buf[BUFSIZ];
for (node *n = t->members.set->h; n; n = n->next) {
sql_part *pt = (sql_part *) n->data;
@@ -331,6 +367,7 @@ rel_propagate_insert(mvc *sql, sql_rel *
(*changes)++;
}
+ //generate the exception
if(isRangePartitionTable(t)) {
if (list_length(t->members.set) == 1) //when there is just one
partition must set the anti_exp
set_anti(accum);
@@ -347,63 +384,67 @@ rel_propagate_insert(mvc *sql, sql_rel *
assert(0);
}
//generate a count aggregation for the values not present in any of the
partitions
- anti_dup = rel_select(sql->sa, anti_dup, anti_exp);
- anti_dup = rel_project(sql->sa, anti_dup, rel_projections(sql,
anti_dup, NULL, 1, 1));
- anti_dup = rel_groupby(sql, anti_dup, NULL);
- aggr = exp_aggr(sql->sa, NULL, cf, 0, 0, anti_dup->card, 0);
- (void) rel_groupby_add_aggr(sql, anti_dup, aggr);
+ *anti_rel = rel_select(sql->sa, *anti_rel, anti_exp);
+ *anti_rel = rel_project(sql->sa, *anti_rel, rel_projections(sql,
*anti_rel, NULL, 1, 1));
+ *anti_rel = rel_groupby(sql, *anti_rel, NULL);
+ aggr = exp_aggr(sql->sa, NULL, cf, 0, 0, (*anti_rel)->card, 0);
+ (void) rel_groupby_add_aggr(sql, *anti_rel, aggr);
exp_label(sql->sa, aggr, ++sql->label);
- //generate the exception
aggr = exp_column(sql->sa, exp_relname(aggr), exp_name(aggr),
exp_subtype(aggr), aggr->card,
has_nil(aggr), is_intern(aggr));
- snprintf(buf, BUFSIZ, "INSERT: the insert violates the partition %s of
values",
+ snprintf(buf, BUFSIZ, "%s: the %s violates the partition %s of values",
operation, desc,
isRangePartitionTable(t) ? "range" : "list");
- exception = exp_exception(sql->sa, aggr, buf);
- return rel_ddl_distribute(sql->sa, sel, anti_dup,
list_append(new_exp_list(sql->sa), exception));
+ *exception = exp_exception(sql->sa, aggr, buf);
+
+ return sel;
+}
+
+static sql_rel*
+rel_propagate_insert(mvc *sql, sql_rel *rel, sql_table *t, int *changes)
+{
+ sql_exp* exception = NULL;
+ sql_rel* anti_rel = NULL;
+ sql_rel* res = rel_generate_subinserts(sql, rel, &anti_rel, &exception,
t, changes, "INSERT", "insert");
+
+ return rel_ddl_distribute(sql->sa, res, anti_rel,
list_append(new_exp_list(sql->sa), exception));
+}
+
+static sql_rel*
+rel_propagate_delete(mvc *sql, sql_rel *rel, sql_table *t, int *changes)
+{
+ rel = rel_generate_subdeletes(sql, rel, t, changes);
+ return rel_ddl_distribute(sql->sa, rel, NULL, NULL);
}
static sql_rel*
rel_propagate_update(mvc *sql, sql_rel *rel, sql_table *t, int *changes)
{
- int just_one = 1;
+ int found_partition_col = 0;
sql_rel *sel = NULL;
- /*int found_part_col = 0;
for (node *n = ((sql_rel*)rel->r)->exps->h; n; n = n->next) {
sql_exp* exp = (sql_exp*) n->data;
- if(exp->type == e_column) {
- sql_exp* l = (sql_exp*) exp->l;
- if(!strcmp((char*)l->l, t->base.name) &&
!strcmp((char*)l->r, t->pcol->base.name))
- found_part_col = 1;
+ if(exp->type == e_column && exp->l && exp->r &&
!strcmp((char*)exp->l, t->base.name) &&
+ !strcmp((char*)exp->r, t->pcol->base.name)) {
+ found_partition_col = 1;
}
- }*/
-
- for (node *n = t->members.set->h; n; n = n->next) {
- sql_part *pt = (sql_part *) n->data;
- sql_table *sub = find_sql_table(t->s, pt->base.name);
- sql_rel *s1, *dup = NULL;
- list *uexps = exps_copy(sql->sa, rel->exps);
+ }
- if(rel->r) {
- dup = rel_copy(sql->sa, rel->r, 1);
- dup = rel_change_base_table(sql, dup, t, sub);
- }
-
- for(node *ne = uexps->h ; ne ; ne = ne->next)
- ne->data = exp_change_column_table(sql, (sql_exp*)
ne->data, t, sub);
+ if(!found_partition_col) { //easy scenario where the partitioned column
is not being updated, just propagate
+ sel = rel_generate_subupdates(sql, rel, t, changes);
+ return rel_ddl_distribute(sql->sa, sel, NULL, NULL);
+ } else { //harder scenario, has to insert and delete across partitions.
+ /*sql_exp *exception = NULL;
+ sql_rel *inserts = NULL, *deletes = NULL, *anti_rel = NULL;
- //easy scenario where the partitioned column is not being
updated
- s1 = rel_update(sql, rel_basetable(sql, sub, sub->base.name),
dup, NULL, uexps);
- if (just_one == 0) {
- sel = rel_list(sql->sa, sel, s1);
- } else {
- sel = s1;
- just_one = 0;
- }
- (*changes)++;
+ deletes = rel_generate_subdeletes(sql, rel, t, changes);
+ deletes = rel_ddl_distribute(sql->sa, deletes, NULL, NULL);
+ inserts = rel_generate_subinserts(sql, rel, &anti_rel,
&exception, t, changes, "UPDATE", "update");
+ inserts = rel_ddl_distribute(sql->sa, inserts, anti_rel,
list_append(new_exp_list(sql->sa), exception));
+ return rel_list(sql->sa, deletes, inserts);*/
+ assert(0);
}
- return rel_ddl_distribute(sql->sa, sel, NULL, NULL);
}
static sql_rel*
@@ -412,14 +453,12 @@ rel_subtable_insert(mvc *sql, sql_rel *r
sql_table *upper = t->p; //is part of a partition table and not been
used yet
int colr = upper->pcol->colnr;
sql_part *pt = find_sql_part(upper, t->base.name);
- sql_rel *anti_dup = rel_dup(rel->r) /* the anti relation */, *right =
rel->r, *left = rel->l;
- sql_exp *le = list_fetch(right->exps, colr), *anti_exp = NULL,
- *anti_le = list_fetch(anti_dup->exps, colr), *aggr =
NULL, *exception = NULL;
+ sql_rel *anti_dup = rel_dup(rel->r) /* the anti relation */, *left =
rel->l;
+ sql_exp *anti_exp = NULL, *anti_le = list_fetch(anti_dup->exps, colr),
*aggr = NULL, *exception = NULL;
list *anti_exps = new_exp_list(sql->sa);
sql_subaggr *cf = sql_bind_aggr(sql->sa, sql->session->schema, "count",
NULL);
char buf[BUFSIZ];
- le = exp_column(sql->sa, exp_relname(le), exp_name(le),
exp_subtype(le), le->card, has_nil(le), is_intern(le));
anti_le = exp_column(sql->sa, exp_relname(anti_le), exp_name(anti_le),
exp_subtype(anti_le),
anti_le->card,
has_nil(anti_le), is_intern(anti_le));
@@ -474,7 +513,7 @@ rel_propagate(mvc *sql, sql_rel *rel, in
if(l->op == op_basetable) {
sql_table *t = l->l;
- if(isRangePartitionTable(t) || isListPartitionTable(t)) {
+ if((isRangePartitionTable(t) || isListPartitionTable(t)) &&
!find_prop(l->p, PROP_USED)) {
assert(list_length(t->members.set) > 0);
if(is_delete(rel->op) || is_truncate(rel->op)) {
//propagate deletions to the partitions
return rel_propagate_delete(sql, rel, t,
changes);
@@ -488,10 +527,6 @@ rel_propagate(mvc *sql, sql_rel *rel, in
} else if(t->p && (isRangePartitionTable(t->p) ||
isListPartitionTable(t->p)) && !find_prop(l->p, PROP_USED)) {
if(is_insert(rel->op)) { //insertion directly to
sub-table (must do validation)
return rel_subtable_insert(sql, rel, t,
changes);
- } else if(is_update(rel->op)) { //do the proper
validation
- //TODO
- } else if(!(is_delete(rel->op) ||
is_truncate(rel->op))) { //no validation needed in deletes
- assert(0);
}
}
}
diff --git a/sql/server/rel_updates.c b/sql/server/rel_updates.c
--- a/sql/server/rel_updates.c
+++ b/sql/server/rel_updates.c
@@ -900,6 +900,13 @@ update_table(mvc *sql, dlist *qname, dli
dnode *n;
const char *rname = NULL;
sql_rel *res = NULL, *bt = rel_basetable(sql, t, t->base.name);
+ int partitioned_column = -1;
+
+ if(isRangePartitionTable(t) || isListPartitionTable(t)) {
+ partitioned_column = t->pcol->colnr;
+ } else if(t->p && (isRangePartitionTable(t->p) ||
isListPartitionTable(t->p))) {
+ partitioned_column = t->p->pcol->colnr;
+ }
res = bt;
#if 0
@@ -1075,6 +1082,8 @@ update_table(mvc *sql, dlist *qname, dli
sql_column *c = mvc_bind_column(sql, t,
cname);
sql_exp *v = n->data;
+ if(partitioned_column == c->colnr)
+ return sql_error(sql, 02,
SQLSTATE(42000) "UPDATE: Update on the partitioned column not possible at the
moment");
if (!exp_name(v))
exp_label(sql->sa, v,
++sql->label);
v = exp_column(sql->sa, exp_relname(v),
exp_name(v), exp_subtype(v), v->card, has_nil(v), is_intern(v));
@@ -1092,6 +1101,8 @@ update_table(mvc *sql, dlist *qname, dli
char *cname = assignment->h->next->data.sval;
sql_column *c = mvc_bind_column(sql, t, cname);
+ if(partitioned_column == c->colnr)
+ return sql_error(sql, 02,
SQLSTATE(42000) "UPDATE: Update on the partitioned column not possible at the
moment");
if (!v) {
v = exp_atom(sql->sa,
atom_general(sql->sa, &c->type, NULL));
} else if ((v = update_check_column(sql, t, c,
v, r, cname)) == NULL) {
diff --git a/sql/test/merge-partitions/Tests/mergepart11.sql
b/sql/test/merge-partitions/Tests/mergepart11.sql
--- a/sql/test/merge-partitions/Tests/mergepart11.sql
+++ b/sql/test/merge-partitions/Tests/mergepart11.sql
@@ -1,11 +1,11 @@
-CREATE MERGE TABLE moveaccrosspartitions (a int, b varchar(32)) PARTITION BY
VALUES (a);
+CREATE MERGE TABLE moveaccrosspartitions (a int, b varchar(32)) PARTITION BY
RANGE (a);
CREATE TABLE sublimits1 (a int, b varchar(32));
CREATE TABLE sublimits2 (a int, b varchar(32));
CREATE TABLE sublimits3 (a int, b varchar(32));
ALTER TABLE moveaccrosspartitions ADD TABLE sublimits1 AS PARTITION BETWEEN 1
AND 100;
ALTER TABLE moveaccrosspartitions ADD TABLE sublimits2 AS PARTITION BETWEEN
101 AND 200;
-ALTER TABLE moveaccrosspartitions ADD TABLE sublimits2 AS PARTITION BETWEEN
201 AND 300;
+ALTER TABLE moveaccrosspartitions ADD TABLE sublimits3 AS PARTITION BETWEEN
201 AND 300;
INSERT INTO moveaccrosspartitions VALUES (50, 'first'), (150, 'second'), (250,
'third'), (60, 'fourth'), (120, 'fifth'), (240, 'sixth');
diff --git a/sql/test/merge-partitions/Tests/mergepart11.stable.err
b/sql/test/merge-partitions/Tests/mergepart11.stable.err
new file mode 100644
--- /dev/null
+++ b/sql/test/merge-partitions/Tests/mergepart11.stable.err
@@ -0,0 +1,51 @@
+stderr of test 'mergepart11` in directory 'sql/test/merge-partitions` itself:
+
+
+# 09:47:13 >
+# 09:47:13 > "mserver5" "--debug=10" "--set" "gdk_nr_threads=0" "--set"
"mapi_open=true" "--set" "mapi_port=39288" "--set"
"mapi_usock=/var/tmp/mtest-21121/.s.monetdb.39288" "--set" "monet_prompt="
"--forcemito"
"--dbpath=/home/ferreira/repositories/MonetDB-merge-partitions/BUILD/var/MonetDB/mTests_sql_test_merge-partitions"
"--set" "embedded_c=true"
+# 09:47:13 >
+
+# builtin opt gdk_dbpath =
/home/ferreira/repositories/MonetDB-merge-partitions/BUILD/var/monetdb5/dbfarm/demo
+# builtin opt gdk_debug = 0
+# builtin opt gdk_vmtrim = no
+# builtin opt monet_prompt = >
+# builtin opt monet_daemon = no
+# builtin opt mapi_port = 50000
+# builtin opt mapi_open = false
+# builtin opt mapi_autosense = false
_______________________________________________
checkin-list mailing list
checkin-list@monetdb.org
https://www.monetdb.org/mailman/listinfo/checkin-list
