gtfsclean

Tidy (and validate) GTFS feeds.

Fixes small inconsistencies, minimizes the overall feed size, and prepares the feed for secure, standard-compliant further use.

Output feeds are semantically equivalent to the input feed. In this context, semantical equivalency means that the output feed provides exactly the same trips with exactly the same attributes (routes, stop-times, shapes, agency, fares etc.). In other words, the output feed is equivalent to the input feed from a passenger's perspective.

0. Features

• Clean CSV output. Only quote string values where needed, use dynamic float precision and remove whitespace. Only output files that are necessary.
• Default-value error handling. If non-required fields in the input-feed have errors, fall back to the default value according to the GTFS standard
• Drop-entities error handling. If non-fixable errors occur, drop the affected entity (trip, route, stop, etc.).
• Orphan deletion. Delete stops, routes, stop times and shapes that aren't referenced anywhere
• ID minimization. Replace IDs with dense integer or character IDs
• Shape minimization. Minimize shape geometries using the Douglas-Peucker algorithm
• Service minimization. Minimize services in calender.txt and calender_dates.txt by searching optimal progression and/or exception covers.
• Trip/Stop-Time minimization. Minimize trips and stop-times by analyzing stop_times.txt and frequencies.txt and searching for optimal frequency covers.
• Shape remeasurement. If shape measures (shape_dist_traveled) have gaps, try to fill them by interpolating surrounding measurements
• Duplicate removal. Safely remove routes, shapes and services that are semantically equivalent to others and combine them into one.
• Extensive feed validation. Validation includes for example checking stop time progressions, ID references, ID collisions, missing required fields, shape measurements, timezone strings, general field value range validity, URLs, mail addresses, ISO language codes, timepoint validity ...
• Entropy-minimizing heuristics. Sort lines prior to writing them to keep the entropy of the resulting CSV files small. This often leads to better compression results (see below).

For more features, see the help page (--help).

1. Installation

$ go install github.com/public-transport/gtfsclean@latest

Then you need you need to check where it has been installed by running

$ go env GOPATH

and ensure that your terminal's path contains the GOPATH. If everything is set up correctly then running

$ gtfsclean --help

should yield a help message with all of the various cmdline flags.

2. Usage

Each GTFS processor has to be enabled explicitly. See

$ gtfsclean --help

for possible options.

You can use

$ gtfsclean -v <feed>

to do a simple feed validation.

3. Example

Process the SFMTA-Feed with all processors enabled:

$ gtfsclean -SCRmTcdsOeD sanfrancisco.zip

Validate the SFMTA-Feed:

$ gtfsclean -v sanfrancisco.zip

4. Evaluation

All zipped sizes using zip -9.

SFMTA feed

Processed with -SCRmTcdsOeD.

File	# lines before	size before (B)		# lines after	size after (B)
agency.txt	2	159		2	153
calendar_dates.txt	11	172		11	161
calendar.txt	4	194		4	190
fare_attributes.txt	3	109		3	107
fare_rules.txt	83	1k		83	392
frequencies.txt	---	---		3,904	111k
routes.txt	83	3,2K		83	1,9k
shapes.txt	201,831	7.4M		19,863	620K
stops.txt	3,556	212K		3,540	172K
stop_times.txt	1,115,598	46M		893,344	25M
trips.txt	28,989	1,4M		21,741	739K

Compression gain:

File	processors	size before (B)		size after (B)
feed.zip	-SCRmTcdsOeD	8.8M		5.9M
feed.zip	none	8.8M		8.5M

Switzerland feed

Processed with -SCRmTcdsOeD.

File	# lines before	size before (B)		# lines after	size after (B)
agency.txt	424	40K		424	33K
calendar_dates.txt	4,387,195	113M		455,843	6.6M
calendar.txt	32,822	2M		28,873	1014K
fare_attributes.txt	---	---		---	---
fare_rules.txt	---	---		---	---
frequencies.txt	---	---		2,023	57K
routes.txt	5,564	254K		4,613	105K
shapes.txt	---	---		---	---
stops.txt	31,753	2.4M		31,753	1.4M
stop_times.txt	11,569,991	790M		10,771,471	323M
transfers.txt	21,691	846K		21,691	295K
trips.txt	1,055,231	86M		662,771	26M

Compression gain:

File	processors	size before (B)		size after (B)
feed.zip	-SCRmTcdsOeD	80M		63M
feed.zip	none	80M		77M

Prague feed

Processed with -SCRmTcdsOeD.

File	# lines before	size before (B)		# lines after	size after (B)
agency.txt	21	1.7K		21	1.6K
calendar_dates.txt	1	32		8	122
calendar.txt	49	1.8K		27	972
fare_attributes.txt	---	---		---	---
fare_rules.txt	---	---		---	---
frequencies.txt	---	---		21,425	604K
routes.txt	434	25K		434	22K
shapes.txt	784,062	22M		578,211	15M
stops.txt	7,401	385K		7,401	323K
stop_times.txt	1,879,757	69M		751,415	23M
trips.txt	87,348	3.3M		34,812	1.1M

Compression gain:

File	processors	size before (B)		size after (B)
feed.zip	-SCRmTcdsOeD	16M		9.4M
feed.zip	none	16M		16M

5. Available processors

There are two classes of processors. Processors with a lowercase flag modify existing entries. Processors with an uppercase flag delete existing entries, either because they are duplicates or because they can be combined with other entries.

ID minimizer

---

IDs are packed into dense integer arrays, either as base 10 or base 36 integers. You should not use this processor if you are referencing entities from outside the static feed (for example, if the IDs are references from a GTFS-realtime feed).

Flags

• -i: pack IDs into dense base 10 integers
• -d: pack IDs into dense base 36 integers

Modifies

Every file.

Example

Before

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_desc,route_type,route_url,route_color,route_text_color
AB,DTA,10,Airport - Bullfrog,,3,,,
BFC,DTA,20,Bullfrog - Furnace Creek Resort,,3,,,
STBA,DTA,30,Stagecoach - Airport Shuttle,,3,,,
CITY,DTA,40,City,,3,,,
AAMV,DTA,50,Airport - Amargosa Valley,,3,,,
AAMV2,DTA,50,Airport - Amargosa Valley,,3,,,

After

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_type
2,1,20,Bullfrog - Furnace Creek Resort,3
3,1,30,Stagecoach - Airport Shuttle,3
4,1,40,City,3
5,1,50,Airport - Amargosa Valley,3
6,1,50,Airport - Amargosa Valley,3
1,1,10,Airport - Bullfrog,3

Orphan remover

---

Feed is checked for entries that are not referenced anywhere. These entries are removed from the output.

Flags

• -O: remove entities that are not referenced anywhere

Modifies

trips.txt, stops.txt, routes.txt, calendar_dates.txt, calendar.txt

Example

Before

stops.txt

stop_id,stop_name,stop_desc,stop_lat,stop_lon,zone_id,stop_url,parent_station
META1,Furnace Creek,,36.425288,-117.1333162,,,
META2,Furnace Creek,,36.425288,-117.1333162,,,
FUR_CREEK_RES,Furnace Creek Resort (Demo),,36.425288,-117.133162,,,META1

After

stops.txt

stop_id,stop_name,stop_lat,stop_lon,parent_station
META1,Furnace Creek,36.42529,-117.133316,
FUR_CREEK_RES,Furnace Creek Resort (Demo),36.42529,-117.13316,META1

Shape minimizer

---

Minimizes shape geometries using Douglas-Peucker. This processor implicitly calls the shape remeasurer! The shape coordinates are projected to web mercator (EPSG:3857) prior to minimization. The ε value for Douglas-Peucker is set to 1.0

Flags

• -s: minimize shapes (using Douglas-Peucker)

Modifies

shapes.txt

Example

Before

shapes.txt

shape_id,shape_pt_lat,shape_pt_lon,shape_pt_sequence,shape_dist_traveled
A_shp,0,0,1,
A_shp,0.6,0.5,2,6.8310
A_shp,1,1,3,15.8765
A_shp,2,1,4
A_shp,3,1,5,36.76
A_shp,3.5,1,6,
B_shp,0,0,1,
B_shp,0.6,0.5,2,6.8310
B_shp,1,1,3,15.8765
B_shp,2,1,4
B_shp,3,1,5,36.76
B_shp,3.5,1,6,

After

shapes.txt

shape_id,shape_pt_lat,shape_pt_lon,shape_pt_sequence,shape_dist_traveled
A_shp,0,0,1,0
A_shp,0.6,0.5,2,6.831
A_shp,1,1,3,15.8765
A_shp,3.5,1,6,42.910156
B_shp,0,0,1,0
B_shp,0.6,0.5,2,6.831
B_shp,1,1,3,15.8765
B_shp,3.5,1,6,42.910156

Service minimizer

---

Minimizes service ranges in calendar.txt and calendar_dates.txt by searching for optimal coverages of range entries in calendar.txt and exception entries in calendar_dates.txt.

Flags

• -c: minimize services by searching for the optimal exception/range coverage

Modifies

calendar.txt, calendar_dates.txt

Example

Before

calendar.txt

(empty)

calendar_dates.txt

service_id,date,exception_type
FULLW,20160814,1
FULLW,20160815,2
FULLW,20160816,1
FULLW,20160817,1
FULLW,20160818,1
FULLW,20160819,1
FULLW,20160820,1
FULLW,20160821,1

After

calendar.txt

service_id,monday,tuesday,wednesday,thursday,friday,saturday,sunday,start_date,end_date
FULLW,0,1,1,1,1,1,1,20160814,20160821

calendar_dates.txt

(empty)

Trip/Stop-times minimizer

Minimizes stop times in stop_times.txt and trips in trips.txt by searching for progression (frequency) covers on the stop times. If multiple trips with equivalent attributes (route, shapes etc) and the same relative stop times are found, they are checked for frequency patterns. If a pattern can be found, the trips are combined into a single frequency-based trip (via frequency.txt). Existing frequencies in frequencies.txt are also optimized and/or combined with entries in stop_times.txt.

The algorithm is based on a CAP (Cover by Arithmetic Progression) algorithm proposed by Hannah Bast and Sabine Storandt.

Flags

• -T: search for frequency patterns in explicit trips and combine them

Modifies

trips.txt, stop_times.txt, frequencies.txt

Example

Before

stop_times.txt

trip_id,arrival_time,departure_time,stop_id,stop_sequence
AB1a,8:00:00,8:00:00,BEATTY_AIRPORT,1
AB1a,8:10:00,8:15:00,BULLFROG,2
AB1b,8:10:00,8:10:00,BEATTY_AIRPORT,1
AB1b,8:20:00,8:25:00,BULLFROG,2
AB1c,8:20:00,8:20:00,BEATTY_AIRPORT,1
AB1c,8:30:00,8:35:00,BULLFROG,2
AB1d,8:30:00,8:30:00,BEATTY_AIRPORT,1
AB1d,8:40:00,8:45:00,BULLFROG,2

trips.txt

route_id,service_id,trip_id,trip_headsign,direction_id,block_id,shape_id
AB,FULLW,AB1a,to Bullfrog,0,1,A_shp
AB,FULLW,AB1b,to Bullfrog,0,1,A_shp
AB,FULLW,AB1c,to Bullfrog,0,1,A_shp
AB,FULLW,AB1d,to Bullfrog,0,1,A_shp

frequencies.txt

(empty)

After

stop_times.txt

trip_id,arrival_time,departure_time,stop_id,stop_sequence
AB1a,8:00:00,8:00:00,BEATTY_AIRPORT,1
AB1a,8:10:00,8:15:00,BULLFROG,2

trips.txt

route_id,service_id,trip_id,trip_headsign,direction_id,block_id,shape_id
AB,FULLW,AB1a,to Bullfrog,0,1,A_shp

frequencies.txt

trip_id,start_time,end_time,headway_secs,exact_times
AB1a,8:00:00,8:40:00,600,1

Route duplicate remover

---

Removes duplicate routes (routes that have the same attributes and the same fare rules), updates references in trips.txt and deletes redundant rules in fare_rules.txt as well.

Flags

• -R: remove route duplicates

Modifies

routes.txt, trips.txt, fare_rules.txt

Example

Before

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_desc,route_type,route_url,route_color,route_text_color
AB,DTA,10,Airport - Bullfrog,,3,,,
BFC,DTA,20,Bullfrog - Furnace Creek Resort,,3,,,
CFC,DTA,20,Bullfrog - Furnace Creek Resort,,3,,,

fare_rules.txt

fare_id,route_id,origin_id,destination_id,contains_id
p,AB,,,
p,BFC,,,
p,CFC,,,

After

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_type
AB,DTA,10,Airport - Bullfrog,3
CFC,DTA,20,Bullfrog - Furnace Creek Resort,3

fare_rules.txt

fare_id,route_id
p,AB
p,CFC

Service duplicate remover

---

Removes duplicate services (services that cover the same set of dates) and updates references.

Flags

• -C: remove duplicate services in calendar.txt and calendar_dates.txt

Modifies

calendar_dates.txt, calendar.txt, trips.txt

Example

Before

calendar_dates.txt

service_id,date,exception_type
B,20160815,1
B,20160816,1
B,20160817,1
B,20160818,1
B,20160819,1
B,20160820,1

calendar.txt

service_id,monday,tuesday,wednesday,thursday,friday,saturday,sunday,start_date,end_date
A,1,1,1,1,1,1,0,20160814,20160821

After

calendar_dates.txt

(empty)

calendar.txt

service_id,monday,tuesday,wednesday,thursday,friday,saturday,sunday,start_date,end_date
A,1,1,1,1,1,1,0,20160814,20160821

Shape duplicate remover

---

Removes duplicate shapes and updates references in trips.txt. Shape equality testing is done with a simple heuristic which resembles the Fréchet-Distance but is faster. The check never underestimates the distance between two shapes, but overestimates it for shapes with total distances that are >> the max distance. This processor implicitely calls the shape remeasurer.

Flags

• -S: remove shape duplicates

Modifies

shapes.txt, trips.txt

Example

Before

shapes.txt

shape_id,shape_pt_lat,shape_pt_lon,shape_pt_sequence,shape_dist_traveled
A_shp,0,0,1,0
A_shp,0.6,0.5,2,6.831
A_shp,1,1,3,15.8765
A_shp,2,1,4,26.315065
A_shp,3,1,5,36.76
A_shp,3.5,1,6,42.910156
B_shp,0,0,1,0
B_shp,0.6,0.5,2,6.831
B_shp,1,1,3,15.8765
B_shp,2,1,4,26.315065
B_shp,3,1,5,36.75
B_shp,3.500005,1,6,42.91

After

shapes.txt

shape_id,shape_pt_lat,shape_pt_lon,shape_pt_sequence,shape_dist_traveled
B_shp,0,0,1,0
B_shp,0.6,0.5,2,6.831
B_shp,1,1,3,15.8765
B_shp,2,1,4,26.315065
B_shp,3,1,5,36.75
B_shp,3.500005,1,6,42.91

Shape remeasurer

---

Remeasures shapes and fills measurement gaps.

Flags

• -m: remeasure shapes (filling measurement-holes)

Modifies

shapes.txt

Example:

Before

shapes.txt

shape_id,shape_pt_lat,shape_pt_lon,shape_pt_sequence,shape_dist_traveled
A_shp,0,0,1,
A_shp,0.6,0.5,2,6.8310
A_shp,1,1,3,15.8765
A_shp,2,1,4
A_shp,3,1,5,36.76
A_shp,3.5,1,6,

After

shapes.txt

shape_id,shape_pt_lat,shape_pt_lon,shape_pt_sequence,shape_dist_traveled
A_shp,0,0,1,0
A_shp,0.6,0.5,2,6.831
A_shp,1,1,3,15.8765
A_shp,2,1,4,26.315065
A_shp,3,1,5,36.76
A_shp,3.5,1,6,42.910156

Set erroneous values to standard defaults

---

If optional field values of feed entries have errors, this processors sets them to the default values specified in the GTFS standard.

Flags

• -e: if non-required fields have errors, fall back to the default values

Modifies

Every file, if errors are present.

Example:

Before

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_desc,route_type,route_url,route_color,route_text_color
AB,DTAoopserror,10,Airport - Bullfrog,,3,,,
BFC,DTA,20,Bullfrog - Furnace Creek Resort,,3,,,
STBA,DTA,30,Stagecoach - Airport Shuttle,,3,,,
CITY,DTA,40,City,,3,,,
AAMV,DTA,50,Airport - Amargosa Valley,,3,,,
AAMV2,DTA,50,Airport - Amargosa Valley,,3,,,

After

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_type
CITY,DTA,40,City,3
AAMV,DTA,50,Airport - Amargosa Valley,3
AAMV2,DTA,50,Airport - Amargosa Valley,3
AB,,10,Airport - Bullfrog,3
BFC,DTA,20,Bullfrog - Furnace Creek Resort,3
STBA,DTA,30,Stagecoach - Airport Shuttle,3

Drop erroneous entries

---

If feed entries have errors that can't be fixed in any other way (e.g. by -e), this processor completely removes them. This cascades through the entire feed: if an erroneous route is deleted, all trips that refer to this route are also deleted, all stop times for trips referencing this route are deleted as well and so on. You effectively get an error-free subset of the input feed.

Flags:

• -D: drop erroneous entries from feed

Modifies:

Every file, if errors are present.

Example:

Before

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_desc,route_type,route_url,route_color,route_text_color
AB,DTAoopserror,10,Airport - Bullfrog,,3,,,
BFC,DTA,20,Bullfrog - Furnace Creek Resort,,3,,,
STBA,DTA,30,Stagecoach - Airport Shuttle,,3,,,
CITY,DTA,40,City,,3,,,
AAMV,DTA,50,Airport - Amargosa Valley,,3,,,
AAMV2,DTA,50,Airport - Amargosa Valley,,3,,,

After

routes.txt

route_id,agency_id,route_short_name,route_long_name,route_type
CITY,DTA,40,City,3
AAMV,DTA,50,Airport - Amargosa Valley,3
AAMV2,DTA,50,Airport - Amargosa Valley,3
BFC,DTA,20,Bullfrog - Furnace Creek Resort,3
STBA,DTA,30,Stagecoach - Airport Shuttle,3

6. Processing order

The order in which feed processors are called is always the same, regardless of the flag ordering. The processors are run in this order:

1. Default erroneous values (-e)
2. Drop erroneous entries (-D)
3. Remove orphans (-O)
4. Remeasure shapes (-m)
5. Minimize shapes (-s)
6. Remove shape duplicates (-S)
7. Remove route duplicates (-R)
8. Remove service duplicates (-C)
9. Minimize services (-c)
10. Minimize stop-times/trips (-T)
11. Minimize IDs (-i or -d)

7. License

GPL v2, see LICENSE