From 5aa2388eb11a0091a647f37e9bdc7719152d6a6d Mon Sep 17 00:00:00 2001
From: jimingquan <mingquan.ji@vesoft.com>
Date: Fri, 30 Dec 2022 10:31:11 +0800
Subject: [PATCH] filter out previous path which do not met the conditions
 (#5159)

Co-authored-by: Sophie <84560950+Sophie-Xie@users.noreply.github.com>
---
 src/graph/executor/query/TraverseExecutor.cpp | 35 +++++++++++-
 .../bugfix/MatchNodeLabelFilter.feature       | 53 +++++++++++++++++++
 2 files changed, 86 insertions(+), 2 deletions(-)
 create mode 100644 tests/tck/features/bugfix/MatchNodeLabelFilter.feature

diff --git a/src/graph/executor/query/TraverseExecutor.cpp b/src/graph/executor/query/TraverseExecutor.cpp
index 72540c025b0..d0753b348de 100644
--- a/src/graph/executor/query/TraverseExecutor.cpp
+++ b/src/graph/executor/query/TraverseExecutor.cpp
@@ -162,8 +162,39 @@ folly::Future<Status> TraverseExecutor::handleResponse(RpcResponse&& resps) {
   }
   auto listVal = std::make_shared<Value>(std::move(list));
   auto iter = std::make_unique<GetNeighborsIter>(listVal);
-  if (currentStep_ == 1 && range_ && range_->min() == 0) {
-    result_.rows = buildZeroStepPath(iter.get());
+  if (currentStep_ == 1) {
+    if (range_ && range_->min() == 0) {
+      result_.rows = buildZeroStepPath(iter.get());
+    }
+    // match (v)-[e:Rel]-(v1:Label1)-[e1*2]->() where id(v0) in [6, 23] return v1
+    // the attributes of v1 will be obtained in the second traverse operator
+    // If the conditions are not met, the path in the previous step needs to be filtered out
+    if (!dst2PathsMap_.empty()) {
+      std::unordered_set<Value, VertexHash, VertexEqual> existVids;
+      existVids.reserve(iter->numRows());
+      auto vertices = iter->getVertices();
+      for (auto& vertex : vertices.values) {
+        if (vertex.isVertex()) {
+          existVids.emplace(vertex);
+        }
+      }
+      auto dst2PathIter = dst2PathsMap_.begin();
+      while (dst2PathIter != dst2PathsMap_.end()) {
+        if (existVids.find(dst2PathIter->first) == existVids.end()) {
+          dst2PathIter = dst2PathsMap_.erase(dst2PathIter);
+        } else {
+          dst2PathIter++;
+        }
+      }
+      auto initVidIter = initVids_.begin();
+      while (initVidIter != initVids_.end()) {
+        if (existVids.find(*initVidIter) == existVids.end()) {
+          initVidIter = initVids_.erase(initVidIter);
+        } else {
+          initVidIter++;
+        }
+      }
+    }
   }
   expand(iter.get());
   if (!isFinalStep()) {
diff --git a/tests/tck/features/bugfix/MatchNodeLabelFilter.feature b/tests/tck/features/bugfix/MatchNodeLabelFilter.feature
new file mode 100644
index 00000000000..96b8925c4f5
--- /dev/null
+++ b/tests/tck/features/bugfix/MatchNodeLabelFilter.feature
@@ -0,0 +1,53 @@
+# Copyright (c) 2022 vesoft inc. All rights reserved.
+#
+# This source code is licensed under Apache 2.0 License.
+Feature: match node label filter
+
+  Background:
+    Given a graph with space named "nba"
+
+  Scenario: node label filter
+    When executing query:
+      """
+      MATCH (v0)<-[e0:like]-(v1:bachelor)-[e1*2]->()
+      WHERE id(v0) in ['Tony Parker', 'Spurs', 'Yao Ming']
+      return type(e0), labels(v1)
+      """
+    Then the result should be, in any order:
+      | type(e0) | labels(v1)             |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+      | "like"   | ["bachelor", "player"] |
+    When executing query:
+      """
+      MATCH (v0)<-[e0:like]-(v1:bachelor)-[e1*0..2]->()
+      WHERE id(v0) in ['Tony Parker', 'Spurs', 'Yao Ming']
+      return labels(v1), count(*)
+      """
+    Then the result should be, in any order:
+      | labels(v1)             | count(*) |
+      | ["bachelor", "player"] | 34       |