Changeset: a2a3d51e1f72 for MonetDB
URL: https://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=a2a3d51e1f72
Modified Files:
sql/backends/monet5/rel_bin.c
sql/server/rel_optimizer.c
sql/test/mergetables/Tests/sqlsmith-exists.stable.out
Branch: default
Log Message:
in rel_bin always use full name (ie iff no rname, don't match just on cnmae)
diffs (truncated from 332 to 300 lines):
diff --git a/sql/backends/monet5/rel_bin.c b/sql/backends/monet5/rel_bin.c
--- a/sql/backends/monet5/rel_bin.c
+++ b/sql/backends/monet5/rel_bin.c
@@ -150,9 +150,10 @@ list_find_column(backend *be, list *l, c
} else {
for (; e; e = e->chain) {
stmt *s = e->value;
+ const char *rnme = table_name(be->mvc->sa, s);
const char *nme = column_name(be->mvc->sa, s);
- if (nme && strcmp(nme, name) == 0) {
+ if (!rnme && nme && strcmp(nme, name) == 0) {
res = s;
break;
}
@@ -177,9 +178,10 @@ list_find_column(backend *be, list *l, c
}
} else {
for (n = l->h; n; n = n->next) {
+ const char *rnme = table_name(be->mvc->sa, n->data);
const char *nme = column_name(be->mvc->sa, n->data);
- if (nme && strcmp(nme, name) == 0) {
+ if (!rnme && nme && strcmp(nme, name) == 0) {
res = n->data;
break;
}
@@ -279,9 +281,8 @@ distinct_value_list(backend *be, list *v
s = stmt_append(be, s, i);
}
- // Probably faster to filter out the values directly in the underlying
list of atoms.
- // But for now use groupby to filter out duplicate values.
-
+ /* Probably faster to filter out the values directly in the underlying
list of atoms.
+ But for now use groupby to filter out duplicate values. */
stmt* groupby = stmt_group(be, s, NULL, NULL, NULL, 1);
stmt* ext = stmt_result(be, groupby, 1);
@@ -354,17 +355,17 @@ handle_in_exps(backend *be, sql_exp *ce,
}
}
} else {
- // TODO: handle_in_exps should contain all necessary logic for
in-expressions to be SQL compliant.
- // For non-SQL-standard compliant behavior, e.g. PostgreSQL
backwards compatibility, we should
- // make sure that this behavior is replicated by the sql
optimizer and not handle_in_exps.
-
- stmt* last_null_value = NULL; // CORNER CASE ALERT: See
description below.
-
- // The actual in-value-list should not contain duplicates to
ensure that final join results are unique.
+ /* TODO: handle_in_exps should contain all necessary logic for
in-expressions to be SQL compliant.
+ For non-SQL-standard compliant behavior, e.g. PostgreSQL
backwards compatibility, we should
+ make sure that this behavior is replicated by the sql
optimizer and not handle_in_exps. */
+
+ stmt* last_null_value = NULL; /* CORNER CASE ALERT: See
description below. */
+
+ /* The actual in-value-list should not contain duplicates to
ensure that final join results are unique. */
s = distinct_value_list(be, nl, &last_null_value);
if (last_null_value) {
- // The actual in-value-list should not contain null
values.
+ /* The actual in-value-list should not contain null
values. */
s = stmt_project(be, stmt_selectnonil(be, s, NULL), s);
}
@@ -373,21 +374,21 @@ handle_in_exps(backend *be, sql_exp *ce,
if (!in) {
if (last_null_value) {
- // CORNER CASE ALERT:
- // In case of a not-in-expression with the
associated in-value-list containing a null value,
- // the entire in-predicate is forced to always
return false, i.e. an empty candidate list.
- // This is similar to postgres behavior.
- // TODO: However I do not think this behavior
is in accordance with SQL standard 2003.
-
- // Ugly trick to return empty candidate list,
because for all x it holds that: (x == null) == false.
- //list* singleton_bat = sa_list(sql->sa);
- // list_append(singleton_bat, null_value);
+ /* CORNER CASE ALERT:
+ In case of a not-in-expression with the
associated in-value-list containing a null value,
+ the entire in-predicate is forced to always
return false, i.e. an empty candidate list.
+ This is similar to postgres behavior.
+ TODO: However I do not think this behavior
is in accordance with SQL standard 2003.
+
+ Ugly trick to return empty candidate list,
because for all x it holds that: (x == null) == false.
+ list* singleton_bat = sa_list(sql->sa);
+ list_append(singleton_bat, null_value); */
s = stmt_uselect(be, c, last_null_value,
cmp_equal, NULL, 0);
return s;
}
else {
- // BACK TO HAPPY FLOW:
- // Make sure that null values are never
returned.
+ /* BACK TO HAPPY FLOW:
+ Make sure that null values are never
returned. */
stmt* non_nulls;
non_nulls = stmt_selectnonil(be, c, NULL);
s = stmt_tdiff(be, non_nulls, s);
@@ -505,7 +506,6 @@ exp_bin(backend *be, sql_exp *e, stmt *l
/* handle table returning functions */
if (l->type == e_psm && l->flag & PSM_REL) {
stmt *lst = r->op1;
- //if (r->type == st_table && lst->nrcols == 0
&& lst->key) {
if (r->type == st_table && lst->nrcols == 0 &&
lst->key && e->card > CARD_ATOM) {
node *n;
list *l = sa_list(sql->sa);
@@ -916,7 +916,6 @@ exp_bin(backend *be, sql_exp *e, stmt *l
r2 = exp_bin(be, re2, left, right, grp, ext, cnt, sel);
if (!l || !r || (re2 && !r2)) {
- //assert(0);
fprintf(stderr, "query: '%s'\n", sql->query);
return NULL;
}
@@ -1649,35 +1648,38 @@ rel2bin_table(backend *be, sql_rel *rel,
}
l = sa_list(sql->sa);
if (f->func->res) {
- if (f->func->varres) {
- for(i=0, en = rel->exps->h, n =
f->res->h; en; en = en->next, n = n->next, i++ ) {
- sql_exp *exp = en->data;
- sql_subtype *st = n->data;
- const char *rnme =
exp_relname(exp)?exp_relname(exp):exp->l;
- stmt *s = stmt_rs_column(be,
psub, i, st);
-
- s = stmt_alias(be, s, rnme,
exp_name(exp));
- list_append(l, s);
- }
- } else {
- for(i = 0, n = f->func->res->h; n; n =
n->next, i++ ) {
- sql_arg *a = n->data;
- stmt *s = stmt_rs_column(be,
psub, i, &a->type);
- const char *rnme =
exp_find_rel_name(op);
+ if (f->func->varres) {
+ for(i=0, en = rel->exps->h, n = f->res->h; en;
en = en->next, n = n->next, i++ ) {
+ sql_exp *exp = en->data;
+ sql_subtype *st = n->data;
+ const char *rnme =
exp_relname(exp)?exp_relname(exp):exp->l;
+ stmt *s = stmt_rs_column(be, psub, i,
st);
- s = stmt_alias(be, s, rnme,
a->name);
- list_append(l, s);
- }
- if (list_length(f->res) ==
list_length(f->func->res) + 1) {
- /* add missing %TID% column */
- sql_subtype *t =
f->res->t->data;
- stmt *s = stmt_rs_column(be,
psub, i, t);
- const char *rnme =
exp_find_rel_name(op);
-
- s = stmt_alias(be, s, rnme,
TID);
- list_append(l, s);
- }
+ s = stmt_alias(be, s, rnme,
exp_name(exp));
+ list_append(l, s);
}
+ } else {
+ assert(list_length(f->func->res) ==
list_length(rel->exps));
+ node *m;
+ for(i = 0, n = f->func->res->h, m =
rel->exps->h; n && m; n = n->next, m = m->next, i++ ) {
+ sql_arg *a = n->data;
+ sql_exp *exp = m->data;
+ stmt *s = stmt_rs_column(be, psub, i,
&a->type);
+ const char *rnme =
exp_relname(exp)?exp_relname(exp):exp_find_rel_name(op);
+
+ s = stmt_alias(be, s, rnme, a->name);
+ list_append(l, s);
+ }
+ if (list_length(f->res) ==
list_length(f->func->res) + 1) {
+ /* add missing %TID% column */
+ sql_subtype *t = f->res->t->data;
+ stmt *s = stmt_rs_column(be, psub, i,
t);
+ const char *rnme =
exp_find_rel_name(op);
+
+ s = stmt_alias(be, s, rnme, TID);
+ list_append(l, s);
+ }
+ }
}
if (rel->flag == TABLE_PROD_FUNC && sub && sub->nrcols) {
assert(0);
@@ -1703,7 +1705,7 @@ rel2bin_table(backend *be, sql_rel *rel,
sql_exp *c = n->data;
stmt *s = stmt_rs_column(be, sub, i, exp_subtype(c));
const char *nme = exp_name(c);
- const char *rnme = NULL;
+ const char *rnme = exp_relname(c);
s = stmt_alias(be, s, rnme, nme);
if (fr->card <= CARD_ATOM) /* single value, get result
from bat */
@@ -1720,12 +1722,7 @@ rel2bin_table(backend *be, sql_rel *rel,
for( en = rel->exps->h; en; en = en->next ) {
sql_exp *exp = en->data;
const char *rnme = exp_relname(exp)?exp_relname(exp):exp->l;
- stmt *s;
-
- /* no relation names */
- if (exp->l)
- exp->l = NULL;
- s = exp_bin(be, exp, sub, NULL, NULL, NULL, NULL, NULL);
+ stmt *s = bin_find_column(be, sub, exp->l, exp->r);
if (!s) {
assert(0);
@@ -2188,7 +2185,6 @@ rel2bin_antijoin(backend *be, sql_rel *r
}
/* handle join-ing conditions first */
if (!list_empty(jexps)) {
- // assert(0);
if (list_empty(mexps))
mexps = jexps;
}
@@ -3741,7 +3737,7 @@ rel2bin_insert(backend *be, sql_rel *rel
ret = s;
}
- if (be->cur_append) //building the total number of rows affected across
all tables
+ if (be->cur_append) /* building the total number of rows affected
across all tables */
ret->nr = add_to_merge_partitions_accumulator(be, ret->nr);
if (ddl)
@@ -3868,7 +3864,6 @@ update_check_ukey(backend *be, stmt **up
upd = updates[c->c->colnr];
/*
} else if (updates) {
- //assert(0);
//upd = updates[updcol]->op1;
//upd = stmt_project(be, upd,
stmt_col(be, c->c, dels));
upd = stmt_project(be, tids,
stmt_col(be, c->c, dels));
@@ -4747,7 +4742,7 @@ rel2bin_update(backend *be, sql_rel *rel
cnt = s;
}
- if (be->cur_append) //building the total number of rows affected across
all tables
+ if (be->cur_append) /* building the total number of rows affected
across all tables */
cnt->nr = add_to_merge_partitions_accumulator(be, cnt->nr);
if (sql->cascade_action)
@@ -4848,7 +4843,7 @@ sql_delete_ukey(backend *be, stmt *utids
s = stmt_join(be, s, utids, 0, cmp_equal); /* join over
the join index */
s = stmt_result(be, s, 0);
tids = stmt_project(be, s, tids);
- if(cascade) { //for truncate statements with the
cascade option
+ if(cascade) { /* for truncate statements with the
cascade option */
s = sql_delete_cascade_Fkeys(be, fk, tids);
list_prepend(l, s);
} else {
@@ -4952,7 +4947,7 @@ sql_delete(backend *be, sql_table *t, st
return sql_error(sql, 02, SQLSTATE(27000) "DELETE: triggers
failed for table '%s'", t->base.name);
if (rows)
s = stmt_aggr(be, rows, NULL, NULL, sql_bind_aggr(sql->sa,
sql->session->schema, "count", NULL), 1, 0, 1);
- if (be->cur_append) //building the total number of rows affected across
all tables
+ if (be->cur_append) /* building the total number of rows affected
across all tables */
s->nr = add_to_merge_partitions_accumulator(be, s->nr);
return s;
}
@@ -4990,7 +4985,7 @@ struct tablelist {
struct tablelist* next;
};
-static void //inspect the other tables recursively for foreign key dependencies
+static void /* inspect the other tables recursively for foreign key
dependencies */
check_for_foreign_key_references(mvc *sql, struct tablelist* list, struct
tablelist* next_append, sql_table *t, int cascade, int *error) {
node *n;
int found;
@@ -5153,7 +5148,7 @@ sql_truncate(backend *be, sql_table *t,
goto finalize;
}
- if (be->cur_append) //building the total number of rows
affected across all tables
+ if (be->cur_append) /* building the total number of rows
affected across all tables */
other->nr = add_to_merge_partitions_accumulator(be,
other->nr);
}
@@ -5326,7 +5321,7 @@ rel2bin_exception(backend *be, sql_rel *
stmt *s = exp_bin(be, e, l, r, NULL, NULL, NULL, NULL);
append(slist, s);
}
- } else { //if there is no exception condition, just generate a
statement list
+ } else { /* if there is no exception condition, just generate a
statement list */
list_append(slist, l);
list_append(slist, r);
}
diff --git a/sql/server/rel_optimizer.c b/sql/server/rel_optimizer.c
--- a/sql/server/rel_optimizer.c
+++ b/sql/server/rel_optimizer.c
@@ -6187,10 +6187,10 @@ rel_push_project_up(int *changes, mvc *s
/* Don't rewrite refs, non projections or constant or
order by projections */
- if (!l || rel_is_ref(l) ||
+ if (!l || rel_is_ref(l) || is_topn(l->op) ||
(is_join(rel->op) && (!r || rel_is_ref(r))) ||
(is_select(rel->op) && l->op != op_project) ||
- (is_join(rel->op) && l->op != op_project && r->op !=
op_project) ||
+ (is_join(rel->op) && ((l->op != op_project && r->op !=
op_project) || is_topn(r->op))) ||
((l->op == op_project && (!l->l || l->r ||
project_unsafe(l,is_select(rel->op)))) ||
(is_join(rel->op) && (is_subquery(r) ||
(r->op == op_project && (!r->l || r->r ||
project_unsafe(r,0)))))))
diff --git a/sql/test/mergetables/Tests/sqlsmith-exists.stable.out
b/sql/test/mergetables/Tests/sqlsmith-exists.stable.out
--- a/sql/test/mergetables/Tests/sqlsmith-exists.stable.out
_______________________________________________
checkin-list mailing list
[email protected]
https://www.monetdb.org/mailman/listinfo/checkin-list
