memgraph/tests/unit/query_variable_start_planner.cpp

330 lines
12 KiB
C++
Raw Normal View History

#include <algorithm>
#include "gtest/gtest.h"
#include "query/frontend/semantic/symbol_generator.hpp"
#include "query/frontend/semantic/symbol_table.hpp"
#include "query/plan/planner.hpp"
#include "utils/algorithm.hpp"
#include "query_plan_common.hpp"
using namespace query::plan;
using query::AstStorage;
using Direction = query::EdgeAtom::Direction;
namespace std {
// Overloads for printing resulting rows from a query.
std::ostream &operator<<(std::ostream &stream,
const std::vector<TypedValue> &row) {
utils::PrintIterable(stream, row);
return stream;
}
std::ostream &operator<<(std::ostream &stream,
const std::vector<std::vector<TypedValue>> &rows) {
utils::PrintIterable(stream, rows, "\n");
return stream;
}
} // namespace std
namespace {
auto MakeSymbolTable(query::Query &query) {
query::SymbolTable symbol_table;
query::SymbolGenerator symbol_generator(symbol_table);
query.Accept(symbol_generator);
return symbol_table;
}
void AssertRows(const std::vector<std::vector<TypedValue>> &datum,
std::vector<std::vector<TypedValue>> expected) {
auto row_equal = [](const auto &row1, const auto &row2) {
if (row1.size() != row2.size()) {
return false;
}
TypedValue::BoolEqual value_eq;
auto row1_it = row1.begin();
for (auto row2_it = row2.begin(); row2_it != row2.end();
++row1_it, ++row2_it) {
if (!value_eq(*row1_it, *row2_it)) {
return false;
}
}
return true;
};
ASSERT_TRUE(std::is_permutation(datum.begin(), datum.end(), expected.begin(),
expected.end(), row_equal))
<< "Actual rows:" << std::endl
<< datum << std::endl
<< "Expected rows:" << std::endl
<< expected;
};
void CheckPlansProduce(
size_t expected_plan_count, query::Query *query, AstStorage &storage,
database::GraphDbAccessor &dba,
std::function<void(const std::vector<std::vector<TypedValue>> &)> check) {
auto symbol_table = MakeSymbolTable(*query);
auto planning_context =
MakePlanningContext(storage, symbol_table, query, dba);
auto query_parts = CollectQueryParts(symbol_table, storage, query);
EXPECT_TRUE(query_parts.query_parts.size() > 0);
auto single_query_parts = query_parts.query_parts.at(0).single_query_parts;
auto plans = MakeLogicalPlanForSingleQuery<VariableStartPlanner>(
single_query_parts, planning_context);
EXPECT_EQ(std::distance(plans.begin(), plans.end()), expected_plan_count);
for (const auto &plan : plans) {
auto *produce = dynamic_cast<Produce *>(plan.get());
ASSERT_TRUE(produce);
auto results = CollectProduce(produce, symbol_table, dba);
check(results);
}
}
TEST(TestVariableStartPlanner, MatchReturn) {
database::GraphDb db;
auto dba = db.Access();
// Make a graph (v1) -[:r]-> (v2)
auto v1 = dba->InsertVertex();
auto v2 = dba->InsertVertex();
dba->InsertEdge(v1, v2, dba->EdgeType("r"));
dba->AdvanceCommand();
// Test MATCH (n) -[r]-> (m) RETURN n
AstStorage storage;
auto *query = QUERY(SINGLE_QUERY(
MATCH(PATTERN(NODE("n"), EDGE("r", Direction::OUT), NODE("m"))),
RETURN("n")));
// We have 2 nodes `n` and `m` from which we could start, so expect 2 plans.
CheckPlansProduce(2, query, storage, *dba, [&](const auto &results) {
// We expect to produce only a single (v1) node.
AssertRows(results, {{v1}});
});
}
TEST(TestVariableStartPlanner, MatchTripletPatternReturn) {
database::GraphDb db;
auto dba = db.Access();
// Make a graph (v1) -[:r]-> (v2) -[:r]-> (v3)
auto v1 = dba->InsertVertex();
auto v2 = dba->InsertVertex();
auto v3 = dba->InsertVertex();
dba->InsertEdge(v1, v2, dba->EdgeType("r"));
dba->InsertEdge(v2, v3, dba->EdgeType("r"));
dba->AdvanceCommand();
{
// Test `MATCH (n) -[r]-> (m) -[e]-> (l) RETURN n`
AstStorage storage;
auto *query = QUERY(SINGLE_QUERY(
MATCH(PATTERN(NODE("n"), EDGE("r", Direction::OUT), NODE("m"),
EDGE("e", Direction::OUT), NODE("l"))),
RETURN("n")));
// We have 3 nodes: `n`, `m` and `l` from which we could start.
CheckPlansProduce(3, query, storage, *dba, [&](const auto &results) {
// We expect to produce only a single (v1) node.
AssertRows(results, {{v1}});
});
}
{
// Equivalent to `MATCH (n) -[r]-> (m), (m) -[e]-> (l) RETURN n`.
AstStorage storage;
auto *query = QUERY(SINGLE_QUERY(
MATCH(PATTERN(NODE("n"), EDGE("r", Direction::OUT), NODE("m")),
PATTERN(NODE("m"), EDGE("e", Direction::OUT), NODE("l"))),
RETURN("n")));
CheckPlansProduce(3, query, storage, *dba, [&](const auto &results) {
AssertRows(results, {{v1}});
});
}
}
TEST(TestVariableStartPlanner, MatchOptionalMatchReturn) {
database::GraphDb db;
auto dba = db.Access();
// Make a graph (v1) -[:r]-> (v2) -[:r]-> (v3)
auto v1 = dba->InsertVertex();
auto v2 = dba->InsertVertex();
auto v3 = dba->InsertVertex();
dba->InsertEdge(v1, v2, dba->EdgeType("r"));
dba->InsertEdge(v2, v3, dba->EdgeType("r"));
dba->AdvanceCommand();
// Test MATCH (n) -[r]-> (m) OPTIONAL MATCH (m) -[e]-> (l) RETURN n, l
AstStorage storage;
auto *query = QUERY(SINGLE_QUERY(
Variable expansion consolidaton Summary: - Removed BreadthFirstAtom, using EdgeAtom only with a Type enum. - Both variable expansions (breadth and depth first) now have mandatory inner node and edge Identifiers. - Both variable expansions use inline property filtering and support inline lambdas. - BFS and variable expansion now have the same planning process. - Planner modified in the following ways: - Variable expansions support inline property filtering (two filters added to all_filters, one for inline, one for post-expand). - Asserting against existing_edge since we don't support that anymore. - Edge and node symbols bound after variable expansion to disallow post-expand filters to get inlined. - Some things simplified due to different handling. - BreadthFirstExpand logical operator merged into ExpandVariable. Two Cursor classes remain and are dynamically chosen from. As part of planned planner refactor we should ensure that a filter is applied only once. The current implementation is very suboptimal for property filtering in variable expansions. @buda: we will start refactoring this these days. This current planner logic is too dense and complex. It is becoming technical debt. Most of the time I spent working on this has been spent figuring the planning out, and I still needed Teon's help at times. Implementing the correct and optimal version of query execution (avoiding multiple potentially expensive filterings) was out of reach also due to tech debt. Reviewers: buda, teon.banek Reviewed By: teon.banek Subscribers: pullbot, buda Differential Revision: https://phabricator.memgraph.io/D852
2017-10-05 17:25:52 +08:00
MATCH(PATTERN(NODE("n"), EDGE("r", Direction::OUT), NODE("m"))),
OPTIONAL_MATCH(PATTERN(NODE("m"), EDGE("e", Direction::OUT), NODE("l"))),
RETURN("n", "l")));
// We have 2 nodes `n` and `m` from which we could start the MATCH, and 2
// nodes for OPTIONAL MATCH. This should produce 2 * 2 plans.
CheckPlansProduce(4, query, storage, *dba, [&](const auto &results) {
// We expect to produce 2 rows:
// * (v1), (v3)
// * (v2), null
AssertRows(results, {{v1, v3}, {v2, TypedValue::Null}});
});
}
TEST(TestVariableStartPlanner, MatchOptionalMatchMergeReturn) {
database::GraphDb db;
auto dba = db.Access();
// Graph (v1) -[:r]-> (v2)
auto v1 = dba->InsertVertex();
auto v2 = dba->InsertVertex();
auto r_type = dba->EdgeType("r");
dba->InsertEdge(v1, v2, r_type);
dba->AdvanceCommand();
// Test MATCH (n) -[r]-> (m) OPTIONAL MATCH (m) -[e]-> (l)
// MERGE (u) -[q:r]-> (v) RETURN n, m, l, u, v
AstStorage storage;
auto *query = QUERY(SINGLE_QUERY(
Variable expansion consolidaton Summary: - Removed BreadthFirstAtom, using EdgeAtom only with a Type enum. - Both variable expansions (breadth and depth first) now have mandatory inner node and edge Identifiers. - Both variable expansions use inline property filtering and support inline lambdas. - BFS and variable expansion now have the same planning process. - Planner modified in the following ways: - Variable expansions support inline property filtering (two filters added to all_filters, one for inline, one for post-expand). - Asserting against existing_edge since we don't support that anymore. - Edge and node symbols bound after variable expansion to disallow post-expand filters to get inlined. - Some things simplified due to different handling. - BreadthFirstExpand logical operator merged into ExpandVariable. Two Cursor classes remain and are dynamically chosen from. As part of planned planner refactor we should ensure that a filter is applied only once. The current implementation is very suboptimal for property filtering in variable expansions. @buda: we will start refactoring this these days. This current planner logic is too dense and complex. It is becoming technical debt. Most of the time I spent working on this has been spent figuring the planning out, and I still needed Teon's help at times. Implementing the correct and optimal version of query execution (avoiding multiple potentially expensive filterings) was out of reach also due to tech debt. Reviewers: buda, teon.banek Reviewed By: teon.banek Subscribers: pullbot, buda Differential Revision: https://phabricator.memgraph.io/D852
2017-10-05 17:25:52 +08:00
MATCH(PATTERN(NODE("n"), EDGE("r", Direction::OUT), NODE("m"))),
OPTIONAL_MATCH(PATTERN(NODE("m"), EDGE("e", Direction::OUT), NODE("l"))),
MERGE(PATTERN(NODE("u"), EDGE("q", Direction::OUT, {r_type}), NODE("v"))),
RETURN("n", "m", "l", "u", "v")));
// Since MATCH, OPTIONAL MATCH and MERGE each have 2 nodes from which we can
// start, we generate 2 * 2 * 2 plans.
CheckPlansProduce(8, query, storage, *dba, [&](const auto &results) {
// We expect to produce a single row: (v1), (v2), null, (v1), (v2)
AssertRows(results, {{v1, v2, TypedValue::Null, v1, v2}});
});
}
TEST(TestVariableStartPlanner, MatchWithMatchReturn) {
database::GraphDb db;
auto dba = db.Access();
// Graph (v1) -[:r]-> (v2)
auto v1 = dba->InsertVertex();
auto v2 = dba->InsertVertex();
dba->InsertEdge(v1, v2, dba->EdgeType("r"));
dba->AdvanceCommand();
// Test MATCH (n) -[r]-> (m) WITH n MATCH (m) -[r]-> (l) RETURN n, m, l
AstStorage storage;
auto *query = QUERY(SINGLE_QUERY(
MATCH(PATTERN(NODE("n"), EDGE("r", Direction::OUT), NODE("m"))),
WITH("n"),
MATCH(PATTERN(NODE("m"), EDGE("r", Direction::OUT), NODE("l"))),
RETURN("n", "m", "l")));
// We can start from 2 nodes in each match. Since WITH separates query parts,
// we expect to get 2 plans for each, which totals 2 * 2.
CheckPlansProduce(4, query, storage, *dba, [&](const auto &results) {
// We expect to produce a single row: (v1), (v1), (v2)
AssertRows(results, {{v1, v1, v2}});
});
}
TEST(TestVariableStartPlanner, MatchVariableExpand) {
database::GraphDb db;
auto dba = db.Access();
// Graph (v1) -[:r1]-> (v2) -[:r2]-> (v3)
auto v1 = dba->InsertVertex();
auto v2 = dba->InsertVertex();
auto v3 = dba->InsertVertex();
auto r1 = dba->InsertEdge(v1, v2, dba->EdgeType("r1"));
auto r2 = dba->InsertEdge(v2, v3, dba->EdgeType("r2"));
dba->AdvanceCommand();
// Test MATCH (n) -[r*]-> (m) RETURN r
AstStorage storage;
Variable expansion consolidaton Summary: - Removed BreadthFirstAtom, using EdgeAtom only with a Type enum. - Both variable expansions (breadth and depth first) now have mandatory inner node and edge Identifiers. - Both variable expansions use inline property filtering and support inline lambdas. - BFS and variable expansion now have the same planning process. - Planner modified in the following ways: - Variable expansions support inline property filtering (two filters added to all_filters, one for inline, one for post-expand). - Asserting against existing_edge since we don't support that anymore. - Edge and node symbols bound after variable expansion to disallow post-expand filters to get inlined. - Some things simplified due to different handling. - BreadthFirstExpand logical operator merged into ExpandVariable. Two Cursor classes remain and are dynamically chosen from. As part of planned planner refactor we should ensure that a filter is applied only once. The current implementation is very suboptimal for property filtering in variable expansions. @buda: we will start refactoring this these days. This current planner logic is too dense and complex. It is becoming technical debt. Most of the time I spent working on this has been spent figuring the planning out, and I still needed Teon's help at times. Implementing the correct and optimal version of query execution (avoiding multiple potentially expensive filterings) was out of reach also due to tech debt. Reviewers: buda, teon.banek Reviewed By: teon.banek Subscribers: pullbot, buda Differential Revision: https://phabricator.memgraph.io/D852
2017-10-05 17:25:52 +08:00
auto edge = EDGE_VARIABLE("r", Direction::OUT);
auto *query = QUERY(
SINGLE_QUERY(MATCH(PATTERN(NODE("n"), edge, NODE("m"))), RETURN("r")));
// We expect to get a single column with the following rows:
TypedValue r1_list(std::vector<TypedValue>{r1}); // [r1]
TypedValue r2_list(std::vector<TypedValue>{r2}); // [r2]
TypedValue r1_r2_list(std::vector<TypedValue>{r1, r2}); // [r1, r2]
CheckPlansProduce(2, query, storage, *dba, [&](const auto &results) {
AssertRows(results, {{r1_list}, {r2_list}, {r1_r2_list}});
});
}
TEST(TestVariableStartPlanner, MatchVariableExpandReferenceNode) {
database::GraphDb db;
auto dba_ptr = db.Access();
auto &dba = *dba_ptr;
auto id = dba.Property("id");
// Graph (v1 {id:1}) -[:r1]-> (v2 {id: 2}) -[:r2]-> (v3 {id: 3})
auto v1 = dba.InsertVertex();
v1.PropsSet(id, 1);
auto v2 = dba.InsertVertex();
v2.PropsSet(id, 2);
auto v3 = dba.InsertVertex();
v3.PropsSet(id, 3);
auto r1 = dba.InsertEdge(v1, v2, dba.EdgeType("r1"));
auto r2 = dba.InsertEdge(v2, v3, dba.EdgeType("r2"));
dba.AdvanceCommand();
// Test MATCH (n) -[r*..n.id]-> (m) RETURN r
AstStorage storage;
Variable expansion consolidaton Summary: - Removed BreadthFirstAtom, using EdgeAtom only with a Type enum. - Both variable expansions (breadth and depth first) now have mandatory inner node and edge Identifiers. - Both variable expansions use inline property filtering and support inline lambdas. - BFS and variable expansion now have the same planning process. - Planner modified in the following ways: - Variable expansions support inline property filtering (two filters added to all_filters, one for inline, one for post-expand). - Asserting against existing_edge since we don't support that anymore. - Edge and node symbols bound after variable expansion to disallow post-expand filters to get inlined. - Some things simplified due to different handling. - BreadthFirstExpand logical operator merged into ExpandVariable. Two Cursor classes remain and are dynamically chosen from. As part of planned planner refactor we should ensure that a filter is applied only once. The current implementation is very suboptimal for property filtering in variable expansions. @buda: we will start refactoring this these days. This current planner logic is too dense and complex. It is becoming technical debt. Most of the time I spent working on this has been spent figuring the planning out, and I still needed Teon's help at times. Implementing the correct and optimal version of query execution (avoiding multiple potentially expensive filterings) was out of reach also due to tech debt. Reviewers: buda, teon.banek Reviewed By: teon.banek Subscribers: pullbot, buda Differential Revision: https://phabricator.memgraph.io/D852
2017-10-05 17:25:52 +08:00
auto edge = EDGE_VARIABLE("r", Direction::OUT);
edge->upper_bound_ = PROPERTY_LOOKUP("n", id);
auto *query = QUERY(
SINGLE_QUERY(MATCH(PATTERN(NODE("n"), edge, NODE("m"))), RETURN("r")));
// We expect to get a single column with the following rows:
TypedValue r1_list(std::vector<TypedValue>{r1}); // [r1] (v1 -[*..1]-> v2)
TypedValue r2_list(std::vector<TypedValue>{r2}); // [r2] (v2 -[*..2]-> v3)
CheckPlansProduce(2, query, storage, dba, [&](const auto &results) {
AssertRows(results, {{r1_list}, {r2_list}});
});
}
TEST(TestVariableStartPlanner, MatchVariableExpandBoth) {
database::GraphDb db;
auto dba = db.Access();
auto id = dba->Property("id");
// Graph (v1 {id:1}) -[:r1]-> (v2) -[:r2]-> (v3)
auto v1 = dba->InsertVertex();
v1.PropsSet(id, 1);
auto v2 = dba->InsertVertex();
auto v3 = dba->InsertVertex();
auto r1 = dba->InsertEdge(v1, v2, dba->EdgeType("r1"));
auto r2 = dba->InsertEdge(v2, v3, dba->EdgeType("r2"));
dba->AdvanceCommand();
// Test MATCH (n {id:1}) -[r*]- (m) RETURN r
AstStorage storage;
Variable expansion consolidaton Summary: - Removed BreadthFirstAtom, using EdgeAtom only with a Type enum. - Both variable expansions (breadth and depth first) now have mandatory inner node and edge Identifiers. - Both variable expansions use inline property filtering and support inline lambdas. - BFS and variable expansion now have the same planning process. - Planner modified in the following ways: - Variable expansions support inline property filtering (two filters added to all_filters, one for inline, one for post-expand). - Asserting against existing_edge since we don't support that anymore. - Edge and node symbols bound after variable expansion to disallow post-expand filters to get inlined. - Some things simplified due to different handling. - BreadthFirstExpand logical operator merged into ExpandVariable. Two Cursor classes remain and are dynamically chosen from. As part of planned planner refactor we should ensure that a filter is applied only once. The current implementation is very suboptimal for property filtering in variable expansions. @buda: we will start refactoring this these days. This current planner logic is too dense and complex. It is becoming technical debt. Most of the time I spent working on this has been spent figuring the planning out, and I still needed Teon's help at times. Implementing the correct and optimal version of query execution (avoiding multiple potentially expensive filterings) was out of reach also due to tech debt. Reviewers: buda, teon.banek Reviewed By: teon.banek Subscribers: pullbot, buda Differential Revision: https://phabricator.memgraph.io/D852
2017-10-05 17:25:52 +08:00
auto edge = EDGE_VARIABLE("r", Direction::BOTH);
auto node_n = NODE("n");
node_n->properties_[std::make_pair("id", id)] = LITERAL(1);
auto *query =
QUERY(SINGLE_QUERY(MATCH(PATTERN(node_n, edge, NODE("m"))), RETURN("r")));
// We expect to get a single column with the following rows:
TypedValue r1_list(std::vector<TypedValue>{r1}); // [r1]
TypedValue r1_r2_list(std::vector<TypedValue>{r1, r2}); // [r1, r2]
CheckPlansProduce(2, query, storage, *dba, [&](const auto &results) {
AssertRows(results, {{r1_list}, {r1_r2_list}});
});
}
TEST(TestVariableStartPlanner, MatchBfs) {
database::GraphDb db;
auto dba_ptr = db.Access();
auto &dba = *dba_ptr;
auto id = dba.Property("id");
// Graph (v1 {id:1}) -[:r1]-> (v2 {id: 2}) -[:r2]-> (v3 {id: 3})
auto v1 = dba.InsertVertex();
v1.PropsSet(id, 1);
auto v2 = dba.InsertVertex();
v2.PropsSet(id, 2);
auto v3 = dba.InsertVertex();
v3.PropsSet(id, 3);
auto r1 = dba.InsertEdge(v1, v2, dba.EdgeType("r1"));
dba.InsertEdge(v2, v3, dba.EdgeType("r2"));
dba.AdvanceCommand();
Variable expansion consolidaton Summary: - Removed BreadthFirstAtom, using EdgeAtom only with a Type enum. - Both variable expansions (breadth and depth first) now have mandatory inner node and edge Identifiers. - Both variable expansions use inline property filtering and support inline lambdas. - BFS and variable expansion now have the same planning process. - Planner modified in the following ways: - Variable expansions support inline property filtering (two filters added to all_filters, one for inline, one for post-expand). - Asserting against existing_edge since we don't support that anymore. - Edge and node symbols bound after variable expansion to disallow post-expand filters to get inlined. - Some things simplified due to different handling. - BreadthFirstExpand logical operator merged into ExpandVariable. Two Cursor classes remain and are dynamically chosen from. As part of planned planner refactor we should ensure that a filter is applied only once. The current implementation is very suboptimal for property filtering in variable expansions. @buda: we will start refactoring this these days. This current planner logic is too dense and complex. It is becoming technical debt. Most of the time I spent working on this has been spent figuring the planning out, and I still needed Teon's help at times. Implementing the correct and optimal version of query execution (avoiding multiple potentially expensive filterings) was out of reach also due to tech debt. Reviewers: buda, teon.banek Reviewed By: teon.banek Subscribers: pullbot, buda Differential Revision: https://phabricator.memgraph.io/D852
2017-10-05 17:25:52 +08:00
// Test MATCH (n) -[r *bfs..10](r, n | n.id <> 3)]-> (m) RETURN r
AstStorage storage;
Variable expansion consolidaton Summary: - Removed BreadthFirstAtom, using EdgeAtom only with a Type enum. - Both variable expansions (breadth and depth first) now have mandatory inner node and edge Identifiers. - Both variable expansions use inline property filtering and support inline lambdas. - BFS and variable expansion now have the same planning process. - Planner modified in the following ways: - Variable expansions support inline property filtering (two filters added to all_filters, one for inline, one for post-expand). - Asserting against existing_edge since we don't support that anymore. - Edge and node symbols bound after variable expansion to disallow post-expand filters to get inlined. - Some things simplified due to different handling. - BreadthFirstExpand logical operator merged into ExpandVariable. Two Cursor classes remain and are dynamically chosen from. As part of planned planner refactor we should ensure that a filter is applied only once. The current implementation is very suboptimal for property filtering in variable expansions. @buda: we will start refactoring this these days. This current planner logic is too dense and complex. It is becoming technical debt. Most of the time I spent working on this has been spent figuring the planning out, and I still needed Teon's help at times. Implementing the correct and optimal version of query execution (avoiding multiple potentially expensive filterings) was out of reach also due to tech debt. Reviewers: buda, teon.banek Reviewed By: teon.banek Subscribers: pullbot, buda Differential Revision: https://phabricator.memgraph.io/D852
2017-10-05 17:25:52 +08:00
auto *bfs = storage.Create<query::EdgeAtom>(
IDENT("r"), EdgeAtom::Type::BREADTH_FIRST, Direction::OUT,
std::vector<storage::EdgeType>{});
bfs->filter_lambda_.inner_edge = IDENT("r");
bfs->filter_lambda_.inner_node = IDENT("n");
bfs->filter_lambda_.expression = NEQ(PROPERTY_LOOKUP("n", id), LITERAL(3));
bfs->upper_bound_ = LITERAL(10);
auto *query = QUERY(
SINGLE_QUERY(MATCH(PATTERN(NODE("n"), bfs, NODE("m"))), RETURN("r")));
// We expect to get a single column with the following rows:
TypedValue r1_list(std::vector<TypedValue>{r1}); // [r1]
CheckPlansProduce(2, query, storage, dba, [&](const auto &results) {
AssertRows(results, {{r1_list}});
});
}
} // namespace