Harness MATCH tests refactored

Summary:
1. Test setup rewritten to take cca 8 seconds. Note that edges are created by using:
`MATCH (a) WITH a MATCH (b) WITH b WHERE rand() < X CREATE (a)-[:ET]->(b)`
Where `X` is a threshold calculated so the desired edge count is the expectation. This seems the only feasable way of generating a large number of edges since query execution does not depend on edge count, but on vertex count.

2. Using the new `assert` function to verify graph state. I recommend doing that in all the harness tests (I don't think we currently have something better).

3. All tests rewritten to take around 200ms per iteration.

4. Test are using SKIP to avoid sending data to the client, but ensure that appropriate operations get executed. This currently seems like the best way of removing unwanted side-effects.

Harness will cost us our sanity. And it doesn't even provide good quality regression testing we really need :(

Reviewers: buda, mislav.bradac, mferencevic

Reviewed By: mferencevic

Subscribers: pullbot

Differential Revision: https://phabricator.memgraph.io/D752

tests/macro_benchmark/harness/groups/match/config.json

@@ -1,3 +1,3 @@
 {
-    "iterations": 3
+    "iterations": 10
 }

tests/macro_benchmark/harness/groups/match/pattern_cycle.run.cypher

@@ -0,0 +1 @@
+MATCH (n)-[r1]->(m)-[r2]->(n) RETURN r1 SKIP 100000000

tests/macro_benchmark/harness/groups/match/pattern_cycle.run.py

@@ -1 +0,0 @@
-print("MATCH (n)-[r1]->(m)-[r2]->(n) RETURN *")

tests/macro_benchmark/harness/groups/match/pattern_long.run.cypher

@@ -0,0 +1 @@
+MATCH (n1)-[r1]->(n2)-[r2]->(n3)-[r3]->(n4)<-[r4]-(n5) RETURN n5 SKIP 100000 LIMIT 1

tests/macro_benchmark/harness/groups/match/pattern_long.run.py

@@ -1 +0,0 @@
-print("MATCH (n1)-[r1]->(n2)<-[r2]-(n3)-[r3]->(n4) RETURN *")

tests/macro_benchmark/harness/groups/match/pattern_short.run.cypher

@@ -0,0 +1 @@
+UNWIND range(0, 5) AS i MATCH (n)-[r]->(m) RETURN r SKIP 10000000

tests/macro_benchmark/harness/groups/match/pattern_short.run.py

@@ -1 +0,0 @@
-print("MATCH (n)-[r]->(m) RETURN *")

tests/macro_benchmark/harness/groups/match/pattern_where.run.py

@@ -1,3 +0,0 @@
-from setup import ID, rint, VERTEX_COUNT
-print("MATCH (n)-[r]->(m) WHERE n.%s = %d AND m.%s = %d RETURN *" % (
-    ID, rint(VERTEX_COUNT), ID, rint(VERTEX_COUNT)))

tests/macro_benchmark/harness/groups/match/setup.py

@@ -8,25 +8,29 @@ from random import randint
 def rint(upper_bound_exclusive):
     return randint(0, upper_bound_exclusive - 1)
 
-VERTEX_COUNT = 10000
+VERTEX_COUNT = 1500
-EDGE_COUNT = VERTEX_COUNT * 3
+EDGE_COUNT = VERTEX_COUNT * 15
 
 # numbers of *different* labels, edge types and properties
 LABEL_COUNT = 10
-EDGE_TYPE_COUNT = 10
 
-MAX_LABELS = 3  # maximum number of labels in a vertex
+MAX_LABELS = 5  # maximum number of labels in a vertex
 MAX_PROPS = 4   # maximum number of properties in a vertex/edge
+MAX_PROP_VALUE = 1000
 
 # some consts used in mutiple files
+LABEL_INDEX = "LabelIndex"
 LABEL_PREFIX = "Label"
-PROP_PREFIX = "Property"
+PROP_PREFIX = "Prop"
 ID = "id"
 
 
+
 def labels():
-    return "".join(":%s%d" % (LABEL_PREFIX, rint(LABEL_COUNT))
+    labels = ":" + LABEL_INDEX
-                   for _ in range(randint(1, MAX_LABELS - 1)))
+    for _ in range(rint(MAX_LABELS)):
+        labels += ":" + LABEL_PREFIX + str(rint(LABEL_COUNT))
+    return labels
 
 
 def properties(id):
@@ -34,22 +38,22 @@ def properties(id):
     Note that if PropX is generated, then all the PropY where Y < X
     are generated. Thus most labels have Prop0, and least have PropMAX_PROPS.
     """
-    return "{%s: %d, %s}" % (ID, id, ",".join(
+    props = {"%s%d" % (PROP_PREFIX, i): rint(MAX_PROP_VALUE)
-        ["%s%d: %d" % (PROP_PREFIX, prop_ind, rint(100))
+             for i in range(rint(MAX_PROPS))}
-         for prop_ind in range(randint(1, MAX_PROPS - 1))]))
+    props[ID] = id
+    return "{" + ", ".join("%s: %s" % kv for kv in props.items()) + "}"
 
 
 def vertex(vertex_index):
     return "(%s %s)" % (labels(), properties(vertex_index))
 
 
-def edge(edge_index):
-    return "[:EdgeType%d %s]" % (rint(EDGE_TYPE_COUNT), properties(edge_index))
-
-
 def main():
+    # create an index to speed setup up
+    print("CREATE INDEX ON :%s(%s);" % (LABEL_INDEX, ID))
+
     # we batch CREATEs because to speed creation up
-    BATCH_SIZE = 50
+    BATCH_SIZE = 30
 
     # create vertices
     for vertex_index in range(VERTEX_COUNT):
@@ -57,14 +61,14 @@ def main():
         if (vertex_index != 0 and vertex_index % BATCH_SIZE == 0) or \
                 vertex_index + 1 == VERTEX_COUNT:
             print(";")
+    print("MATCH (n) RETURN assert(count(n) = %d);" % VERTEX_COUNT)
 
     # create edges
-    for edge_index in range(EDGE_COUNT):
+    print("MATCH (a) WITH a MATCH (b) WITH a, b WHERE rand() < %f "
-        print("MATCH (a {%s: %d}), (b {%s: %d}) MERGE (a)-%s->(b)" % (
+          " CREATE (a)-[:EdgeType]->(b);" % (EDGE_COUNT / VERTEX_COUNT ** 2))
-            ID, randint(0, VERTEX_COUNT - 1),
+    print("MATCH (n)-[r]->() WITH count(r) AS c "
-            ID, randint(0, VERTEX_COUNT - 1),
+          "RETURN assert(c >= %d AND c <= %d);" % (
-            edge(edge_index)))
+            EDGE_COUNT * 0.98, EDGE_COUNT * 1.02))
-        print(";")
 
 
 if __name__ == "__main__":

tests/macro_benchmark/harness/groups/match/vertex_on_index.run.py

@@ -0,0 +1,5 @@
+from setup import LABEL_INDEX, ID, VERTEX_COUNT, rint
+
+print("UNWIND range(0, 10000) AS i "
+      "MATCH (n:%s {%s: %d}) RETURN n SKIP 1000000" % (
+        LABEL_INDEX, ID, rint(VERTEX_COUNT)))

tests/macro_benchmark/harness/groups/match/vertex_on_label.run.py

@@ -1,2 +1,5 @@
-from setup import LABEL_COUNT, rint
+from setup import LABEL_COUNT, LABEL_PREFIX
-print("MATCH (n:Label%d) RETURN n" % rint(LABEL_COUNT))
+
+for i in range(LABEL_COUNT):
+    print("UNWIND range(0, 30) AS i MATCH (n:%s%d) "
+          "RETURN n SKIP 1000000;" % (LABEL_PREFIX, i))

tests/macro_benchmark/harness/groups/match/vertex_on_label_property.run.py

@@ -1,3 +1,5 @@
-from setup import LABEL_PREFIX, PROP_PREFIX, MAX_PROPS, LABEL_COUNT, rint
+from setup import LABEL_PREFIX, PROP_PREFIX, MAX_PROPS, MAX_PROP_VALUE, LABEL_COUNT, rint
-print("MATCH (n:%s%d {%s%d: %d}) RETURN n" % (
+
-    LABEL_PREFIX, rint(LABEL_COUNT), PROP_PREFIX, rint(MAX_PROPS), rint(10)))
+for i in range(LABEL_COUNT):
+    print("UNWIND range(0, 50) AS i MATCH (n:%s%d {%s%d: %d}) RETURN n SKIP 10000;" % (
+        LABEL_PREFIX, i, PROP_PREFIX, rint(MAX_PROPS), rint(MAX_PROP_VALUE)))

tests/macro_benchmark/harness/groups/match/vertex_on_property.run.py

@@ -1,3 +1,4 @@
-from setup import PROP_PREFIX, MAX_PROPS, rint
+from setup import PROP_PREFIX, MAX_PROPS, rint, MAX_PROP_VALUE
-print("MATCH (n {%s%d: %d}) RETURN n" % (
+
-    PROP_PREFIX, rint(MAX_PROPS), rint(10)))
+print("UNWIND range(0, 50) AS i MATCH (n {%s%d: %d}) RETURN n SKIP 10000" % (
+    PROP_PREFIX, rint(MAX_PROPS), rint(MAX_PROP_VALUE)))

tests/macro_benchmark/harness/harness.py

@@ -217,8 +217,14 @@ class _QuerySuite:
 
         def execute(config_name, num_client_workers=1):
             queries = scenario.get(config_name)
-            return runner.execute(queries(), num_client_workers) if queries \
+            start_time = time.time()
-                else None
+            if queries:
+                r_val = runner.execute(queries(), num_client_workers)
+            else:
+                r_val = None
+            log.info("\t%s done in %.2f seconds" % (config_name,
+                                                    time.time() - start_time))
+            return r_val
 
         measurements = []
 

tools/apollo/generate

@@ -193,7 +193,7 @@ binary_release_link_path = os.path.join(BUILD_RELEASE_DIR, "memgraph")
 # macro benchmark tests
 MACRO_BENCHMARK_ARGS = (
     "QuerySuite MemgraphRunner "
-    "--groups aggregation 1000_create unwind_create dense_expand "
+    "--groups aggregation 1000_create unwind_create dense_expand match "
     "--no-strict")
 macro_bench_path = os.path.join(BASE_DIR, "tests", "macro_benchmark")
 harness_client_binaries = os.path.join(BUILD_RELEASE_DIR, "tests",