Mapping Hacks

During my recent month on the circuit I had the chance to run a couple of open mapping tools talk/workshops; I need to write up a narrative or schematic soon. One tool I showed a lot of and have become increasingly impressed with its solidity and functional simplicity, is JOSM, the Java-based offline OpenStreetmap editor written by Imi Scholz.

For me one of the best things about being back in Europe is the proximity to active mapmaking again. In new places I start actively going out of my way to collect traces of different streets for OSM. Back in London, I find the model to be very much more complete, and I’ve been going back over a lot of the features that I sketched in roughly a year and a half ago (and that I’m sure others have borne with my initial messiness and made corrections too) and trying to do a spot of data gardening.

To get started with JOSM one has to feed it tracks. I have a tiny test track of hundred metres or so around Limehouse Town Hall I use to load it up with. Then to haul the data down from OSM, and see how much has changed in the coverage. After this stage it’s also possible to save the raw trackpoint data back locally in GPX format.

Hooray! Lots of new and refined data showing where I’d drawn streets in error; where smaller features really start to look reliably filled in; where a goodish data density of trackpoints provides a good indicator of “primariness” or otherwise of streets on the basis of how frequently they are traversed. The very heavily marked streets provide a good reference point, too. I find myself moving nodes a lot and adding new nodes into segments that used to provide a broader picture of primary roads.

Having the e-courier data is interesting; one of these modern, dynamic and intelligent cab goods transfer companies that does optimised scheduling based on GPS. It accumulates a vast quantity of quite literal “drive-by data” which it was throwing away before coming to a data sharing agreement with OpenStreetmap. In places the e-courier data is a bit overwhelming. Canary Wharf is a massive business hub, a relatively recent development complex providing a counterweight to the City of London. An awful lot of couriers have reason to take things there. So the model of Canary Wharf looks like this - a more or less undifferentiated blob some hundred metres across.

I don’t really know where to begin, here; I suppose some things still need to be surveyed on foot, by hand. But on the whole the data density is tremendous - and the completion level of the map really seems to be coming together. The completion of the street-level vectors, that is; annotation-wise there’s a lot more to be done, and I want to think about ways of making that easier.

JOSM provides basic editing capacity for key-value pairs which correspond to properties of features. The latest version (which is predictably stable and always recommended, rather than the “latest release” lacking new features) provides a list of common pre-set keys - specifically those supported by Osmarender, the XLST-based SVG visualisation tool for OpenStreetmap maps.

However, JOSM doesn’t provide or suggest recommended values for those keys. The supported values are quite fully set out in OSM’s Map Features document. This is stable enough to be given a namespace and talked about authoritatively, maybe backing it up with an RDF schema one day. When I talked with Sven Anders in Hamburg he said the German approach was to use the English language key-value pairs and provide a mapping to the colloquial German terms on an equivalent wiki page.

One could describe JOSM as a “power-user tool” for OpenStreetmap. I wish a tool at one level of sophistication could unfold or fold into another. This is a special case though as the tasks at hand really are at different levels of detail. It gets to the point where the drawing is really done, at least of street segments, and the task is to collect them into “ways” representing contiguous street sections which share properties, and to add and augment those properties. The second pass calls for a metadata editor for features, not a feature editor.

As such I think it’s really too much to ask the “casual user” to deal with the overhead of a Java client that does much more than this. OpenLayers is a good candidate for a UI driver, the question then is hooking into the OSM API. I recall Chris Schmidt got about that far during the spate of OSM-reuse prototype demos we made and never finished last year in Boston. Near-instant gratification makes a lot of difference.

Posted in geodata, collaborative mapping, london, metadata, openstreetmap |

You can follow any responses to this entry through the RSS 2.0 feed. Trackback from your own site.