1.2.7

Original release on 011-05-06:

• New features
  • Adding possibility to warn about "deprecated column names" (r1423, other related commits: r1463, r1521)
  • Adding the first FeedValidator extension: this “googletransit” extension adds and alters validations of feeds to be used by Google Transit (r1452, see also FeedValidatorExtensions)
    • extension of transitfeed.FareAttribute
    • validation of new file feedinfo.txt
    • extension of transitfeed.Schedule
    • extension of transitfeed.Frequency (r1455)
    • extension of transitfeed.Route and transitfeed.Stop (r1509)
    • extension of transitfeed.Agency (includes the BCP-47 language code validation) (r1521, other related commits: r1524)
    • adding feedvalidator_googletransit.py for using the googletransit extension without having to add the -extension flag (r1498)
• General validation changes
  • Removed date dependency from two tests (r1393)
  • Allowing trip_id to contain a hash '#' (r1399)
  • Avoiding duplicate errors from schedule.Validate due to validate_children=true (r1408)
  • Fix handling of empty string stop_lat and stop_lon values in stops.txt ( issue #258 , r1419)
  • Harmonizing and extending date validation in ServicePeriod. Dates must be between 1900 and 2100 (r1424)
• transitfeed API changes
  • Changes to the extensibility mechanism to make creating and testing of extensions easier (r1409 and r1438)
  • Code formatting changes: cleaned code and removed trailing whitespaces (r1420)
  • Code refactoring: harmonized the Expect... and Expect...InClosure functions in the ValidationTestCase (r1421)
  • Making the error level configurable on all problem types (r1439)
  • Making feedvalidator.py extensible (r1498)
• Other changes
  • Making transitfeed nosetests run on Windows 7 (r1458)
  • Adding a line to the output (both Console and HTML) saying which extension, if any, was used to validate the feed (r1497)

1.2.6

Original release on 2010-08-05:

• New features
  • Introduction of an extensibility mechanism so validation can be altered without directly changing the base transitfeed code (r1296. Other related commits: r1319)
• Validation changes
  • Transfers are now ID'd by the pair (from_stop_id, to_stop_id). While not prohibited by GTFS duplicate IDs are still allowed, but a warning is issued (r1268)
  • Warning when two trips in the same block have overlapping stop time intervals, and when both trips run on the same date. (Change by Brian Ferris - issue #230, r1297)
  • Removed error when transfers = 0 and transfer_duration > 0, as it is allowed by GTFS (issue #232, r1298)
  • Added a warning when there are no exceptions defined in calendar_dates.txt (r1306)
  • Added detection of large transfers. Issues a warning if the stops are >2km appart, and an error if they are >10km appart. Also issues a warning if min_transfer_time + 120s is not enough to cover the distance when walking at 2m/s (very fast walking). (issue #234, r1307)
  • util.FloatStringToFloat and util.NonNegIntStringToInt are now more flexible, issuing a warning when the number formats are not recognized but can be parsed by Python. As this is implementation dependent, it should be moved to an error as soon as possible. (r1350)
• transitfeed API changes
  • Introduction of an object factory (r1281)
  • Introduction of the HeadwayPeriod (later renamed to Frequency) and GtfsFactory classes (r1281)
  • Trip._FIELD_NAMES_HEADWAY was removed. Frequency._FIELD_NAMES should be used instead. (r1296)
  • Invalid values in routes.txt's route_color and route_text_color are now replaced by the respective default values (r1281)
  • Schedule.AddAgencyObject() no longer calls Validate() by default on the Agency object being added (r1281)
  • Schedule.AddTripObject() no longer calls Validate() by default on the Trip object being added (r1281)
  • Schedule.AddRouteObject() no longer calls Validate() on the Route object being added (r1281)
  • Instanciation of GTFS classes should not be done directly, but through GtfsFactory instead (r1281)
  • Introduction of a ShapePoint class (r1289)
  • Separation between problem reporting (what warnings and errors exist) and accumulation (how to handle them). (r1319)
  • Renamed a couple of classes to more closely mirror their respective GTFS names. Fare was renamed FareAttribute and HeadwayPeriod was renamed Frequency. GenericGTFSObject was also renamed to GtfsObjectBase. (r1324)
• Other changes
  • FeedValidator on Windows now deletes database file successfully (issue #206, r1257)
  • Introduction of a GenericGTFSObject (later renamed GtfsObjectBase) from which all GTFS classes should inherit (r1258)
  • Moved each GTFS class to its own file and removed the big transitfeed/_transitfeed.py. The API was not changed. (issue #196, r1275)
  • Fixed google_random_queries.py usage info, added URL to usage of several tools (r1280)
  • Added single date selection to Schedule Viewer (issue #17, r1300)
  • Unit tests now pass in Python 2.6 (issue #237, r1323)
  • All classes, except StopTime for performance reasons, go through the GtfsFactory instead of directly instantiating classes. (r1321)

1.2.5

Original release on 2009-12-15:

• Fixing the Windows binary distribution, which by mistake did not include the pytz package for timezone validation.

1.2.4

Original release on 2009-12-11:

• Moved transitfeed.py to transitfeed/_transitfeed.py to package it properly with transitfeed/util.py and transitfeed/shapelib.py; provided transitfeed/init.py to make the contents of transitfeed available under the same names as before.
• Validation changes
  • Warn when there are several consecutive days without service in the feed validation (issue #188)
  • Warn when the successive stop-times have the same value of shape_dist_traveled and Error when they are decreasing (issue #125, issue #203)
  • The W3C formula for luminance comparison of line colors is used (partial fix for issue #108)
  • Warn when an input file has more than one column with the same name (issue #87)
  • Warn when start date or end date are set, but are not valid dates (issue #41)
  • For problem types that can be sorted by severity keep the N most severe instead of the first N found. These include !StopTooFarFromParentStation, !StopsTooClose, !StationsTooClose, !DifferentStationTooClose, !StopTooFarFromShapeWithDistTraveled, and !TooFastTravel. (issue #184)
• transitfeed API changes
  • Schedule.Validate() is now not run by Loader.Load() by default. To call it implicitly set extra_validation to True, or call it explicitly indicating the service_gap_interval.
  • For performance reasons !ServicePeriod.!IsActiveOn now takes both a date string and, optionally, a date object representing the same date as the string.
• Other changes
  • All tools now have a crash handler. The crash dump file is now called transitfeedcrash.txt
  • --help available in all tools
  • Make bottom panel in schedule viewer fixed height to partially address issue #45
  • Changes to make all tests pass in python 2.4

1.2.2

• Validation changes
  • Warn if a stop is served by a route_type 1 and 3 (Subway and Bus). (Guangfan)
  • Warn if vehicles travel too fast between stops (Clancy)
  • Output warning or error if a stop is more than 100m or 1km from its parent (Fabien)
  • Warn if a schedule starts in the future, issue #159
  • Warn when a new version of transitfeed has been created (Guangfan)
  • Include summary of trips per day in validation-results.html
  • parent_station of stop X is not station Y but they are near changed from error to warning (issue #167)
  • Support HHH:MM:SS for really long distance trips
• New features
  • feedvalidator.py --duplicate_trip_check Check for duplicate trips which go through the same stops with same service and start times. (Guangfan)
  • feedvalidator.py --limit_per_type=N default is 5. only show the first N warnings or errors of each type. This saves memory when there are many problems. (issue #37)
  • unusual_trip_filter.py --route_type apply filter only to selected route type. (Jiri)
  • feedvalidator.py -o CONSOLE print all errors and warnings to the command console. This lets the validator output useful information before a crash and saves memory when there are many problems.
  • location_editor.py adds support for editing station locations to schedule_viewer (Jiri)
• Fix for
  • issue #57, sort of fixed. Added Schedule(load_stop_times=False) used by google_random_queries.py
  • issue #160, feedvalidator crash when given bad path
  • issue #153, merger crash when given wrong number of arguments
  • issue #156, merge raises error and crashes on input with warning or error
  • issue #177, Inconsistent line ends error message should say what file had the bad lines
  • examples/google_random_queries.py prints errors and warnings to console instead of crashing
• transitfeed.Schedule gets new method !GetServicePeriodsActiveEachDate
• The check for stops served by subway and bus and duplicate_trip_check use a second scan of stop_times, which increases validation time by ~10%. They may be disabled by default in a future release.

1.2.1

Original release on 2009-05-26:

• Validation changes
  • Change unknown 'route_type' value in 'route.txt' from error into warning. (Jiri)
  • Change unknown 'location_type' value in 'stops.txt'from error into warning. (Leio)
  • Warn if shape_dist_traveled is larger than max shape_dist_traveled of its shape (Leio)
  • issue #114, issue an Error if a trip has frequencies defined, but no stop_times. (Clancy)
  • check that for a stop entrance with location_type=2, its parent_station must have location_type=1. (Leio)
  • Stricter checking for route_long_name containing route_short_name. (Leio)
  • Some checking that shape and stop location are within 1000m when using shape_dist_traveled (Leio)
• [Merge Merger] change
  • Use a temporary file instead of memory for stops_times. This allows feeds to be merged with less RAM but is slower. If you have small files or lots of RAM the option --memory_db reverts to the previous in-memory storage.
  • Fixed issue #165, merge when schedules have the same trip_id
• Fix for
  • issue #143, issue #144 Unicode related crashes
  • issue #148, crash on badly formatted shape_dist_traveled (Clancy)
  • issue #115, assert stop_id and !IndexError crash hit when column isn't found
  • issue 3120, "%s" isn't a recognized time zone; improved error message
• ScheduleViewer tweaks
  • Marks patterns with unusual trips (those having non-zero trip_type) with gray background. (Jiri)
  • Add support for mouse wheel zooming (Jiri)
  • Displays station in red as opposed to stops in blue. (Jiri)
  • Add a new option --host to support ditu.google.com maps (Leio)
• New tool unusual_trip_filter.py finds infrequent patterns of routes, setting a GTFS attribute 'trip_type' of the corresponding trips to value '1'. No Windows binary yet.

1.2.0

Original release on 2009-02-27:

• GetTimeInterpolatedStops uses polyline connecting stops instead of straight-line between timepoints. (Thanks to William Lachance)
• slightly stricter time format validation. minutes and second must start with 0-5 instead of any digit (Thanks to Clancy Childs)
• add crash handler to merge.py
• protect crash handler from objects that crash when you try to convert them to a string
• Modify shape_importer to parse shapefiles with polylines between stops instead of for entire trips (Thanks to Dennis Lee)
• Support for arbitrary columns in routes.txt, trips.txt
• ScheduleViewer tweaks:
  • lat/lng in the station bubble (Thanks to Dennis Lee)
  • trip_id search box, issue #96 (Thanks to Clancy Childs)
• Fix for
  • Issue #67 (and duplicates): bad date format causes feedvalidator crash
  • Issue #136 (half of it): zip files written are now compressed
  • Issue #110: Check that times are in order of sequences in stop_times (Thanks to Clancy Childs)
  • Issue #112: Warn about unexpected file names in feed (Thanks to Clancy Childs)
  • Issue #119: Warn about unexpected subdirectories in feed (Thanks to Clancy Childs)
  • Issue #113: Don't raise a RuntimeError for new location_type (Thanks to Clancy Childs)

1.1.9

Original release on 2008-12-01:

• Fix kmlwriter, which was broken in 1.1.8 (Thanks to PK for the fix)

1.1.8

Original release on 2008-11-25:

• Support for transfers.txt (Thanks to Ivan)
• Make Problem reporter more robust when given a mix of unicode and basic strings
• new command line option for kmlwriter: --date_filter "Restrict to trips active on date YYYYMMDD"
• transitfeed.Trip gets new method ReplaceStopTimeObject
• Fix for
  • Issue #75: validate stop_latitude and lng are in format -?[1-9]\d+.\d+
  • Issue #106: TypeError: unsupported operand type(s) for -: 'NoneType' and 'NoneType' in Schedule.Validate
  • Issue #118: error in transfers.txt file does not appear in validation-results.html document

1.1.6

• Store stop_times table in sqlite. This greatly reduces the memory requirements.
• schedule_viewer no longer requires --key or --feed_filename when connecting to localhost
• added option to kmlwriter to show each shape point
• added new shape parsing library, shapelib.py, but not used by any tools yet
• new methods ServicePeriod.IsActiveOn and ServicePeriod.ActiveDates
• merge stops with station hierarchy (location_type and parent_station)
• fix some bugs

Measured in linux 2.6.22 on an Intel Core 2 Duo CPU T7300 at 2GHz laptop using feedvalidator.py -p.

Version	Dataset	stop_time rows	Memory (MB)	Time (s)
1.1.5	bart	33k	31	12
	portland	1134k	984.2	551
1.1.6	bart	33k	29.5	20
	portland	1134k	149.0	746
1.1.6 --memory_db	portland	1134k	235.8	738