understanding the converter

[bonfirefan]

Hello - thank you for creating this gtfs converter - I'm surprised this does not get more use. I've used gtfs2graphs to convert the gtfs feed for a specific agency into network graphs for the most part, but wanted to understand the project a bit better. There also appear to be some possibly deprecated code that refer to conf files that do not exist (in gml_extract_types.py).

To clarify -

• get_feeds.py : connects to transitfeeds and downloads the gtfs files for all available agencies, once we have inputted our transitapi keys in get_feeds_conf_.yaml
• gtfs2graphs.py : converts a gtfs zip file to our selected graph output
• gml_extract_types.py : this does not work due to the conf file mentioned
• gmlgtf2slp.py : converts gml files to lp files
• run4allfeeds.sh : appears to be bash script to run gtfs2graphs.py for all feeds available

While the gtf2graphs.py appears to have worked for the feed I am interested in working with, it appears that the graphs output with non-identifiable ID's. It also appears that there are some config files that are set up specifically for Germany - is this something we will have to set regionally?

Thanks again in advance.

[daajoe]

Hi,
thanks!
Yes, you are absolutely right the repo misses a documentation.
I wrote one for a solver competition PACE 2016 transit feeds. It also contains a brief description about the converter parts.

• get_feeds.py : yes
• gtfs2graphs.py : yes. atm only dimacs/lp/gr/gml. note that gr and dimacs are integer based formats.
• gml_extract_types.py ... is deprecated (it was only used to split already available gml graphs with available agency/route type info)
• gmlgtf2slp.py : yes.
• run4allfeeds.sh : yes, but does it only for the gr format

Issue (non-identifiable IDs):

• this was a decision, because of a legal issue/unclarity. many transit feed providers don't use CC. making the data anonymous should avoid this for solver competitions.
• the option --print_symtab or --print_labels should allow you to output also the labels/symbol table for the integer based formats
• if you want to modify the output in the gml file graph_output.py:write_gml should have all the necessary data also for identifiable IDs

Configs:
The config file gtfs_into_conf.yml contains only info for the Berlin Metro region. It was only used to have a trivial mapping to cities.

The proper way to do a geo-location would probably be to take the lat/long data stored in the stops and run a geo-localization. The networkx graphs should already have the lat/long data (gtfs2graphs.py:L97). Google's geo-location is just a RESTful call, but the free plan is limited afaik to 1k calls each day (150k after registration).

[bonfirefan]

Thank you, the documentation is extremely helpful! I'll be continuing to work on this and will ask if I have further questions.