[Talk-GB] TfL cycle data published - schema mapping

Martin - CycleStreets list-osm-talk-gb at cyclestreets.net
Sat Sep 14 13:56:28 UTC 2019



> I will shortly be circulating a proposed mapping of CID -> OSM 
> attributes, reflecting the published schema of the finalised version of 
> the dataset, for comment.


A proposed mapping of the data is here:
https://bikedata.cyclestreets.net/tflcid/conversion/

I would very much welcome comments and thoughts!

This is also available as an Excel (and JSON) file at:
https://github.com/cyclestreets/tflcid-conversion/tree/master/schema

This reflects the final released data rather than earlier versions, with 
the data viewable and filterable at at:
https://bikedata.cyclestreets.net/tflcid/


For each field, I've looked at the data and proposed what is shown as "OSM 
tag(s)", and there is an associated "OSM comment" where relevant.

The TfL CID database is large (234,333 assets), but quite a number of the 
aspects are not really relevant to OSM. Across the 9 asset types, there are 
95 fields, but 23 of them I propose would be discarded (search for 
"Discard" on the webpage).

Cycle parking data is in my view the most useful part of the database - it 
seems to be more comprehensive than OSM has. The cycle lane data is also 
excellent, in that it contains the actual start-stop locations, compared to 
OSM's tendency sometimes just to have a single lane marked for the whole 
length of a street even though it can come and go.

The data is high quality - I've very rarely come across errors, having now 
spent quite a lot of time looking at it. There are two images of each 
location so it is verifiable easily.

As an example of data that is likely to be unwanted, 118,893 (~50%) of 
these assets are signs (the actual signage head or paint on the ground). 
While OSM does have support for traffic_sign=*, in practice this are rarely 
used, as the tagging on the Ways that the signage actually represents is 
more important, and that data is in any case also represented in the CID as 
line geometries.

For instance, a contraflow cycleway has both the line geometry with 
clt_contra=TRUE but there are also signs for that contraflow. Only the line 
geometry is really needed for OSM. There are a few cases where this doesn't 
quite hold true (e.g. where a sign says "No cycling" but the path beyond it 
isn't actually present, because that is not classed as cycle 
infrastructure).

I'll shortly e-mail again with more detailed commentary on various aspects 
of what is shown, in particular cases where new tags are suggested. 
Searching on the webpage for "Community" will find cases where there are 
particular issues which as I say I'll e-mail about separately.


PS The webpage display, which reads from the JSON file, is a bit of a work 
in progress - I'm aware the ID links aren't yet activated, and the Browse 
map link isn't yet picking up the field (only the type). I'll fix these 
shortly.


Martin,                     **  CycleStreets - For Cyclists, By Cyclists
Developer, CycleStreets     **  https://www.cyclestreets.net/






More information about the Talk-GB mailing list