jenks

[morganherlocker]

Jenks natural breaks classifier

[drewda]

Perhaps of relevance: https://gist.github.com/drewda/1299198

[morganherlocker]

Hey drew, that code looks really great! For now I am going to use @tmcw's simple-statistics algo, since it is already implemented in js. I will, however, swing back and use this if I have any issues with ss. Thanks for the suggestion!

[morganherlocker]

This feature has been implemented and tested. A quantile classifier has also been implemented, which just leaves equal interval, which should be trivial.

[drewda]

Sounds good. Also, great to see your velocity on this project!

[morganherlocker]

Thanks, I am running as fast as I can from the existing systems out of business necessity, which tends to be a pretty strong motivator! If you have any other suggestions or requests from the project, please reach out. Any help or input is much appreciated. :)

[drewda]

Great. I'll look through some of my past projects and see if any of the one-off implementations I've written to work with vector features are worth trying to generalize and contribute to this project.

In general, how do you see Turf in relation to the Java Topology Suite (and its JS cousins)? Similar scope of functionality but implemented with GeoJSON in mind?

(If there's a more appropriate place to continue this conversation, I'm glad to continue over there.)

[morganherlocker]

Let's take it to email. morgan.herlocker@gmail.com

[morganherlocker]

Hey Drew, I didn't see an email, but it might have hit the spam filter. Regardless, I am planning to write up a post about the relation between turf and jsts soon. Here is the basic rough draft of the post that I wrote up in response to someone else who asked the same question:

JSTS is an awesome project. I may even depend on some of its code in turf in the future. At the same time, it is far more complicated than turf, and the learning curve is much steeper than turf. Here are the main things I would point to that distinguish turf from JSTS:

• dead simple api. JSTS tends to follow Java idioms rather than node.js (which is understandable, as that is where it came from).
• geojson from the ground up. This means you can load any geojson file into memory, pass it directly into any turf function, and get geojson back that can be directly displayed into a map. No complex data structures makes it way easier to use.
• functional programming style. This will make it possible to distribute work out over node.js computing clusters and do a lot of cool "mapreduce-y" type stuff.
• focus on data analysis, instead of topological manipulations (turf will do many of these things, but the real goal is more complex data analysis like spatial clustering, contours, tins, data classification, etc.).

[tmcw]

:+1: There is lots of room for an alternative to jsts that doesn't over-abstract, is in idiomatic js, and is modular.

[morganherlocker]

I made a post fleshing out my response here: http://morganherlocker.com/post/turf-vs-JSTS