Newbie Questions on OmniSci


1 comment

  • Avatar
    Michael Flaxman

    Hi RTPSR,

    Glad you're having fun with the platform. Good questions, let me take a stab

    1. no you can't, since that function would require the renderer API which is only in the enterprise edition

    2. in existing Immerse client, you are correct, there is no geo download function. If you are using CE, then presumably using your own front end anyway. In that case, possible but not trivial to add one depending on use case. If you don't mind geo-CSV or geo-JSON in WGS84, you can get that in a single call from javascript (basically EXPORT TO). Just make sure to LIMIT a bit so you don't peg your server with requests to serialize 10B polygons to geoJSON!

    For more control and options, I once wrote a python WFS3 client which downloaded geojson, but with some sanity limits by default (<100k features or so). Its been awhile, but it was a "flask app" on the server which used pymapd to grab the features, and geopandas to serve out into various formats. The issues I delegated to geopandas were reprojection and file formats. In that way you can support download of zipped shapefiles in various coordinate systems for example.

    1. To summarize across various vector feature geographies you generally need to figure out a scheme for subdividing them with some common geo key. Visually, you can do this in Immerse using the heat maps charts for either grids or hexes. To do it analytically, varies based on feature types and size relationships. Generally its a two-step. First you add a column to each input geometry layer and assign it a hex ID or lat/lon bin ID. Then you join on those IDs across layers.

    2. Partially. We use normal Mapbox, to which you can add whatever you please. We also add "cross filtering" (through an OS library), which is often considered a major feature/improvement over the standard slippery map. OmniSci does support WMS within Immerse, and Mapbox lets you do that for a custom front end as well. We don't usually worry about 3rd party vector tile sets because we're usually dealing with dataset sizes you would never want to send across the wire. But in concept you could do it just by adding Mapbox config pointing to various other GIS resources.

    You are correct that there is no raster support per se. However, the server doesn't mind billions of points, and does provide a heatmap grid render. So my recommended workflow is to use GDAL to convert rasters to points, then render them as points or heatmaps. The performance is just fine, and you get the additional benefit of being able to cross-filter points dynamically (which rasters don't usually let you do). For example, if you have a raster of land use, you can create an interactive histogram of land use class by record count or area. Selecting one or more of those bars will instantly update the map display (and everything else cross filtered).

    Good luck, and if you can tell us more about your specific use case and I can perhaps provide more targeted advice.


    Comment actions Permalink

Please sign in to leave a comment.