Fix Issue 945 - incorrect count(*) return values (#1288) (#1315)

Fixed issue 945 where count(*) would have incorrect return values.

NOTE -

We need to re-evaluate if we want to use output_node or not.
If output_node is set to false, then it basically short circuits
match for instances where a variable isn't specified -

    MATCH () RETURN 0;
    MATCH () MATCH () RETURN 0;

While, on the surface, this appears to be a good way to improve
execution time of commands that won't do anything, it also
causes chained commands to not work correctly. This is because
a match without a variable will still feed its tuples to the next
stage(s) even though they won't necessarily be sent to the output.
For example, with count(*) -

    MATCH () RETURN count(*);
    MATCH () MATCH RETURN count(*);

With output_node set to false, it won't send the tuples. We will
likely need to remove all of the output_node logic. However, this
needs to be reviewed. For now, we just set it to true and update
the output of the regression tests.

Updated regression tests to accomodate the change.
Added new regression tests to cover overlooked cases.

regress/expected/cypher_delete.out

@@ -707,8 +707,8 @@ SELECT * FROM cypher('detach_delete', $$ MATCH ()-[e]->() RETURN e.name $$) as (
 -------
  "ab"
  "cd"
- "nm"
  "am"
+ "nm"
  "pq"
 (5 rows)
 

regress/expected/cypher_set.out

@@ -167,8 +167,8 @@ SELECT * FROM cypher('cypher_set', $$MATCH ()-[n]->(:other_v) RETURN n$$) AS (a
 ----------------------------------------------------------------------------------------------------------------------------------------
  {"id": 1125899906842625, "label": "e", "end_id": 1407374883553281, "start_id": 281474976710657, "properties": {"i": 3, "j": 20}}::edge
  {"id": 1125899906842626, "label": "e", "end_id": 1407374883553282, "start_id": 844424930131969, "properties": {"i": 3, "j": 20}}::edge
- {"id": 1125899906842627, "label": "e", "end_id": 1407374883553283, "start_id": 844424930131971, "properties": {"i": 3, "j": 20}}::edge
  {"id": 1125899906842628, "label": "e", "end_id": 1407374883553284, "start_id": 844424930131970, "properties": {"i": 3, "j": 20}}::edge
+ {"id": 1125899906842627, "label": "e", "end_id": 1407374883553283, "start_id": 844424930131971, "properties": {"i": 3, "j": 20}}::edge
 (4 rows)
 
 SELECT * FROM cypher('cypher_set', $$

regress/expected/expr.out

@@ -2222,10 +2222,10 @@ SELECT * FROM cypher('expr', $$ MATCH (v) RETURN v $$) AS (expression agtype);
 SELECT * FROM cypher('expr', $$ MATCH ()-[e]-() RETURN e $$) AS (expression agtype);
                                                         expression                                                         
 ---------------------------------------------------------------------------------------------------------------------------
- {"id": 1407374883553282, "label": "e1", "end_id": 1125899906842626, "start_id": 1125899906842625, "properties": {}}::edge
  {"id": 1407374883553281, "label": "e1", "end_id": 1125899906842627, "start_id": 1125899906842626, "properties": {}}::edge
- {"id": 1407374883553282, "label": "e1", "end_id": 1125899906842626, "start_id": 1125899906842625, "properties": {}}::edge
  {"id": 1407374883553281, "label": "e1", "end_id": 1125899906842627, "start_id": 1125899906842626, "properties": {}}::edge
+ {"id": 1407374883553282, "label": "e1", "end_id": 1125899906842626, "start_id": 1125899906842625, "properties": {}}::edge
+ {"id": 1407374883553282, "label": "e1", "end_id": 1125899906842626, "start_id": 1125899906842625, "properties": {}}::edge
 (4 rows)
 
 -- id()
@@ -2234,10 +2234,10 @@ SELECT * FROM cypher('expr', $$
 $$) AS (id agtype);
         id        
 ------------------
- 1407374883553282
  1407374883553281
- 1407374883553282
  1407374883553281
+ 1407374883553282
+ 1407374883553282
 (4 rows)
 
 SELECT * FROM cypher('expr', $$
@@ -2276,10 +2276,10 @@ SELECT * FROM cypher('expr', $$
 $$) AS (start_id agtype);
      start_id     
 ------------------
- 1125899906842625
  1125899906842626
- 1125899906842625
  1125899906842626
+ 1125899906842625
+ 1125899906842625
 (4 rows)
 
 -- should return null
@@ -2309,10 +2309,10 @@ SELECT * FROM cypher('expr', $$
 $$) AS (end_id agtype);
       end_id      
 ------------------
- 1125899906842626
  1125899906842627
- 1125899906842626
  1125899906842627
+ 1125899906842626
+ 1125899906842626
 (4 rows)
 
 -- should return null

regress/sql/cypher_match.sql

@@ -1136,12 +1136,50 @@ SELECT count(*) FROM cypher('test_enable_containment', $$ MATCH (x:Customer {pho
 -- information in the plan for each run.
 -- SELECT * FROM cypher('test_enable_containment', $$ EXPLAIN (COSTS OFF) MATCH (x:Customer {school:{name:'XYZ',program:{degree:'BSc'}},phone:[987654321],parents:{}}) RETURN x $$) as (a agtype);
 
+--
+-- Issue 945
+--
+SELECT create_graph('issue_945');
+SELECT * FROM cypher('issue_945', $$
+    CREATE (a:Part {part_num: '123'}),
+           (b:Part {part_num: '345'}),
+           (c:Part {part_num: '456'}),
+           (d:Part {part_num: '789'})
+    $$) as (result agtype);
+
+-- should match 4
+SELECT * FROM cypher('issue_945', $$
+    MATCH (a:Part) RETURN a
+    $$) as (result agtype);
+
+-- each should return 4
+SELECT * FROM cypher('issue_945', $$ MATCH (:Part) RETURN count(*) $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (a:Part) RETURN count(*) $$) as (result agtype);
+
+-- each should return 4 rows of 0
+SELECT * FROM cypher('issue_945', $$ MATCH (:Part) RETURN 0 $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (a:Part) RETURN 0 $$) as (result agtype);
+
+-- each should return 16 rows of 0
+SELECT * FROM cypher('issue_945', $$ MATCH (:Part) MATCH (:Part) RETURN 0 $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (a:Part) MATCH (:Part) RETURN 0 $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (:Part) MATCH (b:Part) RETURN 0 $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (a:Part) MATCH (b:Part) RETURN 0 $$) as (result agtype);
+
+-- each should return a count of 16
+SELECT * FROM cypher('issue_945', $$ MATCH (:Part) MATCH (:Part) RETURN count(*) $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (a:Part) MATCH (:Part) RETURN count(*) $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (:Part) MATCH (b:Part) RETURN count(*) $$) as (result agtype);
+SELECT * FROM cypher('issue_945', $$ MATCH (a:Part) MATCH (b:Part) RETURN count(*) $$) as (result agtype);
+
+
 --
 -- Clean up
 --
 SELECT drop_graph('cypher_match', true);
 SELECT drop_graph('test_retrieve_var', true);
 SELECT drop_graph('test_enable_containment', true);
+SELECT drop_graph('issue_945', true);
 
 --
 -- End

src/backend/parser/cypher_clause.c

@@ -4036,7 +4036,7 @@ static List *transform_match_entities(cypher_parsestate *cpstate, Query *query,
                                     node->name),
                             parser_errposition(pstate, node->location)));
                 }
-                
+
                 /*
                  * Checks the previous clauses to see if the variable already
                  * exists.
@@ -4053,6 +4053,22 @@ static List *transform_match_entities(cypher_parsestate *cpstate, Query *query,
             output_node = (special_VLE_case && !node->name && !node->props) ?
                           false :
                           INCLUDE_NODE_IN_JOIN_TREE(path, node);
+            /*
+             * TODO
+             *
+             * We need to re-evaluate if we want to use output_node or not.
+             * If output_node is set to false, then it basically short circuits
+             * the match for instances where a variable isn't specified. While,
+             * on the surface, this appears to be a good way to improve
+             * execution time of commands that won't do anything, it also
+             * causes chained commands to not work correctly. This is because
+             * a match without a variable will still feed its tuples to the next
+             * stage(s). With this set to false, it won't. So we likely need to
+             * remove all of the output_node logic. This needs to be reviewed,
+             * though. For now, we just set it to true and update the output of
+             * the regression tests.
+             */
+            output_node = true;
 
             /* transform vertex */
             expr = transform_cypher_node(cpstate, node, &query->targetList,