Hacker News Reader

Comment by Doctor_Fegg | Hacker News Reader

Building your own elevation API is absolutely trivial; I don't know why anyone would even want to pirate Google's API for large-scale usage.

Firstly, the data is freely available. It's NASA's SRTM [1], downloadable from a zillion mirror sites.

Then just take some code to calculate a lat/long offset and find the right position in the right tile. Bob Osola has some easily portable PHP for this if you need it [2].

Free hint: Putting both your tiles and your (Lua) lookup code in Redis, then calling the latter with EVALSHA, is a really neat and fast way to do this.

If you really need ocean depths, higher resolutions than SRTM offers, or stuff above 60 degrees latitude, then it gets a little more complex. But for 90% of cases doing it yourself is eminently practicable.

[1] http://www2.jpl.nasa.gov/srtm/ [2] http://www.osola.org.uk/elevations/

View on HN

Building a robust elevation api is definitely not trivial for the average dev using the google api. If someone thought the only way to get elevations was through google then they definitely will not know how to find and compile the data, then write the TIN (triangulated irregular network) algos necessary to appropriately estimate the elevation of an arbitrary point. No, it is not that hard, since tools like Grass can do a lot of the TINing stuff for you, but just running Grass, much less scripting it, is not trivial for someone unfamiliar with GIS. Making this stuff stable and performant on a server (as opposed to a one off calculation on a dataset) is also not trivial in the least.

Worrying about TIN is the 10% case. Simple interpolation within SRTM raster tiles is all that's needed for 90% of cases (IMHO, obviously).

I would not think interpolation would be that much easier than a creating a TIN. I have not looked at the dataset though, so it might be "pre-interpolated" (probably using a TIN if I had to guess), in which case this would be pretty easy with something like pyqgis on a flask server or whatever. The entire api could basically just be:

    @app.route("/elevation")
    def elevation():
      fileName = "/path/to/raster/srtmRaster.tif"
      fileInfo = QFileInfo(fileName)
      baseName = fileInfo.baseName()
      srtmRaster = QgsRasterLayer(fileName, baseName)
      res, ident = srtmRaster.identify(QgsPoint(request.form['lat'], request.form['lon']))
      return ident[0]

I can test this out and throw it on github with some polish if anyone is interested.

interesting. I think of these data are freely available, then there ought to be more people using it and making competing versions so that google will not have a monopoly on provision of such data apis.

Don't wait for "other people;" if you think it should be done, I'd encourage you to do it yourself!

How do you know it is just SRTM?

Sorry, should have made myself clearer. I'm not saying that Google just uses SRTM - I'm sure they throw lots of other sources and general Google Magick in there. But for most people in most places, SRTM is good enough.

np and cheers!

Yeah, I think so too. And querying SRTM tiles is prerry simple.

Reply on news.ycombinator.com