fix(MergeLineContext): Partially remove unused service ids

src/main/java/com/conveyal/datatools/manager/jobs/MergeFeedsJob.java

@@ -23,6 +23,7 @@
 import com.fasterxml.jackson.annotation.JsonIgnore;
 import com.fasterxml.jackson.databind.ObjectMapper;
 import com.google.common.collect.Lists;
+import com.google.common.collect.Sets;
 import org.bson.codecs.pojo.annotations.BsonIgnore;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -87,6 +88,10 @@ public class MergeFeedsJob extends FeedSourceJob {
     public Set<String> serviceIdsToCloneRenameAndExtend = new HashSet<>();
     @JsonIgnore @BsonIgnore
     public Set<String> serviceIdsFromActiveFeedToTerminateEarly = new HashSet<>();
+    @JsonIgnore @BsonIgnore
+    public Set<String> serviceIdsFromActiveFeedToRemove = new HashSet<>();
+    @JsonIgnore @BsonIgnore
+    public Set<String> serviceIdsFromFutureFeedToRemove = new HashSet<>();
 
     private List<TripAndCalendars> sharedConsistentTripAndCalendarIds = new ArrayList<>();
 
@@ -441,17 +446,40 @@ private void determineMergeStrategy() {
                 getActiveServiceIds(feedMergeContext.getActiveTripIdsNotInFutureFeed())
             );
 
+            // Build the set of calendars ids from the future feed to be removed
+            // because they become no longer used after shared trips are remapped to another service id.
+            serviceIdsFromFutureFeedToRemove = Sets.difference(
+                feedMergeContext.future.feedToMerge.serviceIds,
+                getFutureServiceIds(feedMergeContext.getFutureTripIdsNotInActiveFeed())
+            );
+
+            // Build the set of calendars ids from the active feed to be removed
+            // because they become no longer used after shared trips are remapped to another service id.
+            serviceIdsFromActiveFeedToRemove = Sets.difference(
+                feedMergeContext.active.feedToMerge.serviceIds,
+                getActiveServiceIds(feedMergeContext.getActiveTripIdsNotInFutureFeed())
+            );
+
             mergeFeedsResult.mergeStrategy = CHECK_STOP_TIMES;
         }
     }
 
     /**
      * Obtains the service ids corresponding to the provided trip ids.
      */
-    private List<String> getActiveServiceIds(Set<String> tripIds) {
+    private Set<String> getActiveServiceIds(Set<String> tripIds) {
         return tripIds.stream()
             .map(tripId -> feedMergeContext.active.feed.trips.get(tripId).service_id)
-            .collect(Collectors.toList());
+            .collect(Collectors.toSet());
+    }
+
+    /**
+     * Obtains the service ids corresponding to the provided trip ids.
+     */
+    private Set<String> getFutureServiceIds(Set<String> tripIds) {
+        return tripIds.stream()
+            .map(tripId -> feedMergeContext.future.feed.trips.get(tripId).service_id)
+            .collect(Collectors.toSet());
     }
 
     /**

src/main/java/com/conveyal/datatools/manager/jobs/feedmerge/CalendarMergeLineContext.java

@@ -86,19 +86,27 @@ private boolean checkCalendarIds(Set<NewGTFSError> idErrors, FieldContext fieldC
                         .format(GTFS_DATE_FORMATTER));
                 }
             }
-/*        } else {
+
+            // Remove calendar entries that are no longer used.
+            if (job.serviceIdsFromActiveFeedToRemove.contains(keyValue)) {
+                LOG.warn(
+                    "Skipping active calendar entry {} because it will become unused in the merged feed.",
+                    keyValue);
+                mergeFeedsResult.skippedIds.add(key);
+                shouldSkipRecord = true;
+            }
+        } else {
             // If handling the future feed, the MTC revised feed merge logic is as follows:
             // - Calendar entries from the future feed will be inserted as is in the merged feed.
-            // so no additional processing needed here, unless the calendar entry is no longer used.
+            // so no additional processing needed here, unless the calendar entry is no longer used,
+            // in that case we drop the calendar entry.
             if (job.serviceIdsFromFutureFeedToRemove.contains(keyValue)) {
                 LOG.warn(
                     "Skipping future calendar entry {} because it will become unused in the merged feed.",
                     keyValue);
                 mergeFeedsResult.skippedIds.add(key);
                 shouldSkipRecord = true;
             }
-
- */
         }
 
 
@@ -137,7 +145,7 @@ public void addClonedServiceId() throws IOException {
             String originalServiceId = keyValue;
             if (job.serviceIdsToCloneRenameAndExtend.contains(originalServiceId)) {
                 // FIXME: Do we need to worry about calendar_dates?
-                String[] clonedValues = getRowValues().clone();
+                String[] clonedValues = getOriginalRowValues().clone();
                 String newServiceId = clonedValues[keyFieldIndex] = String.join(":", getIdScope(), originalServiceId);
                 // Modify start date only (preserve the end date from the future calendar entry).
                 int startDateIndex = Table.CALENDAR.getFieldIndex("start_date");

src/main/java/com/conveyal/datatools/manager/jobs/feedmerge/MergeLineContext.java

@@ -52,6 +52,7 @@ public class MergeLineContext {
     private CsvReader csvReader;
     private boolean skipRecord;
     protected boolean keyFieldMissing;
+    private String[] originalRowValues;
     private String[] rowValues;
     private int lineNumber = 0;
     protected final Table table;
@@ -198,6 +199,7 @@ public boolean iterateOverRows() throws IOException {
             if (!constructRowValues()) {
                 return false;
             }
+
             finishRowAndWriteToZip();
         }
         return true;
@@ -491,6 +493,7 @@ public void initializeRowValues() {
         skipRecord = false;
         // Reset the row values (this must happen after the first line is checked).
         rowValues = new String[sharedSpecFields.size()];
+        originalRowValues = new String[sharedSpecFields.size()];
     }
 
     public void writeValuesToTable(String[] values, boolean incrementLineNumbers) throws IOException {
@@ -522,6 +525,7 @@ public void writeHeaders() throws IOException {
      * @return false, if a failing condition was encountered. true, if everything was ok.
      */
     public boolean constructRowValues() throws IOException {
+        boolean result = true;
         // Piece together the row to write, which should look practically identical to the original
         // row except for the identifiers receiving a prefix to avoid ID conflicts.
         for (int specFieldIndex = 0; specFieldIndex < sharedSpecFields.size(); specFieldIndex++) {
@@ -533,65 +537,73 @@ public boolean constructRowValues() throws IOException {
                 field,
                 csvReader.get(fieldsFoundList.indexOf(field))
             );
-            // Handle filling in agency_id if missing when merging regional feeds. If false is returned,
-            // the job has encountered a failing condition (the method handles failing the job itself).
-            if (!updateAgencyIdIfNeeded(fieldContext)) {
-                return false;
-            }
-            // Determine if field is a GTFS identifier (and scope if needed).
-            scopeValueIfNeeded(fieldContext);
-            // Only need to check for merge conflicts if using MTC merge type because
-            // the regional merge type scopes all identifiers by default. Also, the
-            // reference tracker will get far too large if we attempt to use it to
-            // track references for a large number of feeds (e.g., every feed in New
-            // York State).
-            if (job.mergeType.equals(SERVICE_PERIOD)) {
-                // Remap service id from active feed to distinguish them
-                // from entries with the same id in the future feed.
-                // See https://github.com/ibi-group/datatools-server/issues/244
-                if (handlingActiveFeed && fieldContext.nameEquals(SERVICE_ID)) {
-                    updateAndRemapOutput(fieldContext);
+            originalRowValues[specFieldIndex] = fieldContext.getValueToWrite();
+            if (!skipRecord) {
+                // Handle filling in agency_id if missing when merging regional feeds. If false is returned,
+                // the job has encountered a failing condition (the method handles failing the job itself).
+                if (!updateAgencyIdIfNeeded(fieldContext)) {
+                    result = false;
                 }
+                // Determine if field is a GTFS identifier (and scope if needed).
+                scopeValueIfNeeded(fieldContext);
+                // Only need to check for merge conflicts if using MTC merge type because
+                // the regional merge type scopes all identifiers by default. Also, the
+                // reference tracker will get far too large if we attempt to use it to
+                // track references for a large number of feeds (e.g., every feed in New
+                // York State).
+                if (job.mergeType.equals(SERVICE_PERIOD)) {
+                    // Remap service id from active feed to distinguish them
+                    // from entries with the same id in the future feed.
+                    // See https://github.com/ibi-group/datatools-server/issues/244
+                    if (handlingActiveFeed && fieldContext.nameEquals(SERVICE_ID)) {
+                        updateAndRemapOutput(fieldContext);
+                    }
 
-                updateServiceIdsIfNeeded(fieldContext);
+                    updateServiceIdsIfNeeded(fieldContext);
 
-                // Store values for key fields that have been encountered and update any key values that need modification due
-                // to conflicts.
-                if (!checkFieldsForMergeConflicts(getIdErrors(fieldContext), fieldContext)) {
+                    // Store values for key fields that have been encountered and update any key values that need modification due
+                    // to conflicts.
+                    if (!checkFieldsForMergeConflicts(getIdErrors(fieldContext), fieldContext)) {
+                        skipRecord = true;
+                        continue;
+                    }
+                }
+                // If the current field is a foreign reference, check if the reference has been removed in the
+                // merged result. If this is the case (or other conditions are met), we will need to skip this
+                // record. Likewise, if the reference has been modified, ensure that the value written to the
+                // merged result is correctly updated.
+                if (!checkForeignReferences(fieldContext)) {
                     skipRecord = true;
-                    break;
+                    continue;
                 }
+                rowValues[specFieldIndex] = fieldContext.getValueToWrite();
             }
-            // If the current field is a foreign reference, check if the reference has been removed in the
-            // merged result. If this is the case (or other conditions are met), we will need to skip this
-            // record. Likewise, if the reference has been modified, ensure that the value written to the
-            // merged result is correctly updated.
-            if (!checkForeignReferences(fieldContext)) {
-                skipRecord = true;
-                break;
-            }
-            rowValues[specFieldIndex] = fieldContext.getValueToWrite();
         }
-        return true;
+        return result;
     }
 
-    public void finishRowAndWriteToZip() throws IOException {
+    private void finishRowAndWriteToZip() throws IOException {
+        boolean shouldWriteCurrentRow = true;
         // Do not write rows that are designated to be skipped.
         if (skipRecord && job.mergeType.equals(SERVICE_PERIOD)) {
             mergeFeedsResult.recordsSkipCount++;
-            return;
+            shouldWriteCurrentRow = false;
         }
         // Store row and stop values. If the return value is true, the record has been skipped and we
         // should skip writing the row to the merged table.
         if (storeRowAndStopValues()) {
-            return;
+            shouldWriteCurrentRow = false;
         }
+
         // Finally, handle writing lines to zip entry.
         if (mergedLineNumber == 0) {
             writeHeaders();
         }
-        // Write line to table.
-        writeValuesToTable(rowValues, true);
+
+        if (shouldWriteCurrentRow) {
+            // Write line to table.
+            writeValuesToTable(rowValues, true);
+        }
 
         // Optional table-specific additional processing.
         afterRowWrite();
@@ -631,7 +643,7 @@ protected int getLineNumber() {
         return lineNumber;
     }
 
-    protected String[] getRowValues() { return rowValues; }
+    protected String[] getOriginalRowValues() { return originalRowValues; }
 
     /**
      * Retrieves the value for the specified CSV field.

src/test/java/com/conveyal/datatools/manager/jobs/MergeFeedsJobTest.java

@@ -69,6 +69,11 @@ public class MergeFeedsJobTest extends UnitTest {
      * and some added trips, and a trip from the base feed removed.
      */
     private static FeedVersion fakeTransitSameSignatureTrips;
+    /**
+     * The base feed (transposed to the future dates), with some trip_ids from the base feed with the same signature,
+     * and a trip from the base feed removed.
+     */
+    private static FeedVersion fakeTransitSameSignatureTrips2;
     private static FeedSource bart;
     private static FeedVersion noAgencyVersion1;
     private static FeedVersion noAgencyVersion2;
@@ -137,6 +142,7 @@ public static void setUp() throws IOException {
         fakeTransitModService = createFeedVersion(fakeTransit, zipFolderFiles("merge-data-mod-services"));
         fakeTransitNewSignatureTrips = createFeedVersion(fakeTransit, zipFolderFiles("merge-data-mod-trips"));
         fakeTransitSameSignatureTrips = createFeedVersion(fakeTransit, zipFolderFiles("merge-data-added-trips"));
+        fakeTransitSameSignatureTrips2 = createFeedVersion(fakeTransit, zipFolderFiles("merge-data-added-trips-2"));
 
         // Feeds with no agency id
         FeedSource noAgencyIds = new FeedSource("no-agency-ids", project.id, MANUALLY_UPLOADED);
@@ -345,17 +351,24 @@ void mergeMTCShouldHandleExtendFutureStrategy() throws SQLException {
         // assert service_ids start_dates have been extended to the start_date of the base feed.
         String mergedNamespace = mergeFeedsJob.mergedVersion.namespace;
 
+        // There should be no unused service ids.
+        assertThatSqlCountQueryYieldsExpectedCount(
+            String.format("SELECT count(*) FROM %s.errors where error_type = 'SERVICE_UNUSED'", mergedNamespace),
+            0
+        );
+
         // - calendar table
-        // expect a total of 5 records in calendar table
+        // expect a total of 1 record in calendar table that
+        // corresponds to the trip ids present in both active and future feed.
         assertThatSqlCountQueryYieldsExpectedCount(
             String.format("SELECT count(*) FROM %s.calendar", mergedNamespace),
-            5
+            1
         );
         // expect that two records in calendar table have the correct start_date
         // (one for the original calendar entry, one for the extended service id for trips with same signature)
         assertThatSqlCountQueryYieldsExpectedCount(
             String.format("SELECT count(*) FROM %s.calendar where start_date = '20170918' and monday = 1", mergedNamespace),
-            2
+            1
         );
     }
 
@@ -432,6 +445,60 @@ void mergeMTCShouldHandleMatchingTripIdsWithSameSignature() throws SQLException
         );
     }
 
+    /**
+     * Ensures that an MTC merge of feeds with exact matches of service_ids and trip_ids,
+     * trip ids having the same signature (same stop times) will utilize the
+     * {@link MergeStrategy#CHECK_STOP_TIMES} strategy correctly and drop unused future service ids.
+     */
+    @Test
+    void mergeMTCShouldHandleMatchingTripIdsAndDropUnusedFutureCalendar() throws SQLException {
+        Set<FeedVersion> versions = new HashSet<>();
+        versions.add(fakeTransitBase);
+        versions.add(fakeTransitSameSignatureTrips2);
+        MergeFeedsJob mergeFeedsJob = new MergeFeedsJob(user, versions, "merged_output", MergeFeedsType.SERVICE_PERIOD);
+        // Run the job in this thread (we're not concerned about concurrency here).
+        mergeFeedsJob.run();
+        // Check that correct strategy was used.
+        assertEquals(
+            MergeStrategy.CHECK_STOP_TIMES,
+            mergeFeedsJob.mergeFeedsResult.mergeStrategy
+        );
+        // Result should succeed.
+        assertFalse(
+            mergeFeedsJob.mergeFeedsResult.failed,
+            "Merge feeds job should succeed with CHECK_STOP_TIMES strategy."
+        );
+
+        String mergedNamespace = mergeFeedsJob.mergedVersion.namespace;
+
+        // - calendar table
+        // expect a total of 4 records in calendar table:
+        // - common_id from the active feed (but start date is changed to one day before first start_date in future feed),
+        // - common_id from the future feed (because of one future trip not in the active feed),
+        // - common_id cloned and extended for the matching trip id present in both active and future feeds
+        //   (from MergeFeedsJob#serviceIdsToCloneAndRename),
+        // - only_calendar_id used in the future feed.
+        assertThatSqlCountQueryYieldsExpectedCount(
+            String.format("SELECT count(*) FROM %s.calendar", mergedNamespace),
+            3
+        );
+
+        // Out of all trips from the input datasets, expect 4 trips in merged output.
+        // 1 trip from active feed that is not in the future feed,
+        // 1 trip in both the active and future feeds, with the same signature (same stop times),
+        // 2 trips from the future feed not in the active feed.
+        assertThatSqlCountQueryYieldsExpectedCount(
+            String.format("SELECT count(*) FROM %s.trips", mergedNamespace),
+            3
+        );
+
+        // There should be no unused service ids.
+        assertThatSqlCountQueryYieldsExpectedCount(
+            String.format("SELECT count(*) FROM %s.errors where error_type = 'SERVICE_UNUSED'", mergedNamespace),
+            0
+        );
+    }
+
     /**
      * Ensures that an MTC merge of feeds with trip_ids matching in the active and future feed,
      * but with different signatures (e.g. different stop times) fails.

src/test/resources/com/conveyal/datatools/gtfs/merge-data-added-trips-2/agency.txt

@@ -0,0 +1,2 @@
+agency_id,agency_name,agency_url,agency_lang,agency_phone,agency_email,agency_timezone,agency_fare_url,agency_branding_url
+1,Fake Transit,,,,,America/Los_Angeles,,

src/test/resources/com/conveyal/datatools/gtfs/merge-data-added-trips-2/calendar.txt

@@ -0,0 +1,3 @@
+service_id,monday,tuesday,wednesday,thursday,friday,saturday,sunday,start_date,end_date
+common_id,1,1,1,1,1,1,1,20170923,20170925
+only_calendar_id,1,1,1,1,1,1,1,20170920,20170927

src/test/resources/com/conveyal/datatools/gtfs/merge-data-added-trips-2/feed_info.txt

@@ -0,0 +1,2 @@
+feed_id,feed_publisher_name,feed_publisher_url,feed_lang,feed_version
+fake_transit,Conveyal,http://www.conveyal.com,en,1.0

src/test/resources/com/conveyal/datatools/gtfs/merge-data-added-trips-2/routes.txt

@@ -0,0 +1,3 @@
+agency_id,route_id,route_short_name,route_long_name,route_desc,route_type,route_url,route_color,route_text_color,route_branding_url
+1,1,1,Route 1,,3,,7CE6E7,FFFFFF,
+1,2,2,Route 2,,3,,7CE6E7,FFFFFF,

src/test/resources/com/conveyal/datatools/gtfs/merge-data-added-trips-2/stop_attributes.txt

@@ -0,0 +1,3 @@
+stop_id,accessibility_id,cardinal_direction,relative_position,stop_city
+4u6g,0,SE,FS,Scotts Valley
+johv,0,SE,FS,Scotts Valley

src/test/resources/com/conveyal/datatools/gtfs/merge-data-added-trips-2/stop_times.txt

@@ -0,0 +1,9 @@
+trip_id,arrival_time,departure_time,stop_id,stop_sequence,stop_headsign,pickup_type,drop_off_type,shape_dist_traveled,timepoint
+trip3,07:00:00,07:00:00,4u6g,1,,0,0,0.0000000,
+trip3,07:01:00,07:01:00,johv,2,,0,0,341.4491961,
+only-calendar-trip1,07:00:00,07:00:00,4u6g,1,,0,0,0.0000000,
+only-calendar-trip1,07:01:00,07:01:00,johv,2,,0,0,341.4491961,
+only-calendar-trip2,07:00:00,07:00:00,johv,1,,0,0,0.0000000,
+only-calendar-trip2,07:01:00,07:01:00,4u6g,2,,0,0,341.4491961,
+only-calendar-trip999,07:00:00,07:00:00,johv,1,,0,0,0.0000000,
+only-calendar-trip999,07:01:00,07:01:00,4u6g,2,,0,0,341.4491961,