Capping Distance/Time or just returing K Nearest Points in Table Service

[jamaps]

Greetings OSRM,

I'm setting up OSRM to compute several large OD matrices (each ~ 15,000 x 15,000) for a large portion of southern Ontario.

I was wondering if there's a way to limit the size of the search e.g. only return points and durations that are less than a certain distance or time threshold - or only return the closest k number of points. I only need to analyze travel times for nearby points, not those on opposite sides of my map. Plus I would like to save some computation time if possible.

I did some searching but couldn't find anything. Apologies if this has been answered elsewhere.

Cheers,

• JA

[daniel-j-h]

Your description sounds suspiciously like isochrones? There is an experimental pull request: #3652

There is no option for stopping a Table request "early" on in the API.

For matrices this large you can also try

• Use the node bindings or even libosrm directly (avoid http if you can)
• Save results more efficiently than in json format
• Decompose a single big request into smaller ones (run in parallel) and re-assemble them again; might come with some coordinate to internal data structure overhead, so keep the smaller matrices still as big as possible

[danpat]

To add to this - you can use the sources and destinations option on the table plugin to perform a 1:N query - generate coordinates in a radius around your centerpoint, and gradually repeat queries with a growing radius until you have no more coordinates within range.

Other than that, the table service isn't really intended for this. There are a couple of work-in-progress PR's for true isochrone searches #3652 #2477 and sort of #3903, but none are in the main codebase yet and all need some work.

[jamaps]

Thanks guys! I've decided to divide and conquer in parallel which seems to do the trick faster than anticipated.