Emit warning when osrm-extract encounters ways referencing missing nodes

CHANGELOG.md

@@ -2,6 +2,8 @@
   - Changes from 6.0.0
     - Routing:
       - FIXED: Crash when route starts or ends at `type=manoeuvre` relation via node [#7287](https://github.com/Project-OSRM/osrm-backend/issues/7287)
+    - Extraction:
+      - ADDED: Emit warning when ways reference nodes not present in input data [#1596](https://github.com/Project-OSRM/osrm-backend/issues/1596)
     - Profiles:
       - ADDED: Make `max_collapse_distance` configurable via Lua profiles to preserve short road crossings in pedestrian routing [#6171](https://github.com/Project-OSRM/osrm-backend/issues/6171)
       - CHANGED: Apply configurable penalty (default 60s) to gates in car profile [#6757](https://github.com/Project-OSRM/osrm-backend/issues/6757)

features/options/extract/missing_nodes.feature

@@ -0,0 +1,19 @@
+@extract @options @missing_nodes
+Feature: osrm-extract - Missing node references
+
+    Background:
+        Given the profile "testbot"
+
+    Scenario: osrm-extract - Warn about missing node references
+        Given the node map
+            """
+            a b c
+            """
+        And the ways
+            | nodes |
+            | abc   |
+        And node "b" is removed from OSM data
+        And the data has been saved to disk
+        When I try to run "osrm-extract {osm_file} --profile {profile_file}"
+        Then it should exit with an error
+        And stderr should contain "referenced by ways were not found"

features/step_definitions/data.js

@@ -354,6 +354,13 @@ Given(
   },
 );
 
+Given(/^node "([a-z])" is removed from OSM data$/, function (name, callback) {
+  const node = this.findNodeByName(name);
+  if (!node) throw new Error(`*** unknown node ${name}`);
+  this.OSMDB.nodes = this.OSMDB.nodes.filter((n) => n.id !== node.id);
+  callback();
+});
+
 Given(/^the data has been saved to disk$/, function (callback) {
   this.writeOSM();
   callback();

src/extractor/extraction_containers.cpp

@@ -472,6 +472,7 @@ void ExtractionContainers::PrepareNodes()
         util::UnbufferedLog log;
         log << "Building node id map      ... " << std::flush;
         TIMER_START(id_map);
+        const auto original_used_count = used_node_id_list.size();
         auto node_iter = all_nodes_list.begin();
         auto ref_iter = used_node_id_list.begin();
         auto used_nodes_iter = used_node_id_list.begin();
@@ -500,6 +501,13 @@ void ExtractionContainers::PrepareNodes()
 
         // Remove unused nodes and check maximal internal node id
         used_node_id_list.resize(std::distance(used_node_id_list.begin(), used_nodes_iter));
+        const auto dropped_count = original_used_count - used_node_id_list.size();
+        if (dropped_count > 0)
+        {
+            util::Log(logWARNING) << dropped_count
+                                  << " node(s) referenced by ways were not found in the input data"
+                                     " and will be ignored";
+        }
         if (used_node_id_list.size() > std::numeric_limits<NodeID>::max())
         {
             throw util::exception("There are too many nodes remaining after filtering, OSRM only "