Meeting Info

Date & time
20 Jan 2016, 6:00pm

Host
University of Washington

Address
UW Anderson Hall, Forest Club room 207

It's the room with the elk in it.

47.651737, -122.30754

Loose agenda:

  • @christyheaton will talk about her experience using Python to extract Open Street Map data and the challenges that come with using free, user-contributed data.
  • @jsheedy will present his work on d3-grid-map, a package for presenting regularly gridded datasets.
  • @cliffordsnow Bid for 2016 OSM State of the Map US
  • @aaronr Quick demo of bboxfinder.com and projfinder.com as possible ideas for Spring Fling 2016. Also talk about migrating both tools under the cugos github account.
  • @foundatron will give us an update on this year’s CUGOS Spring Fling and let you know how you can get involved!
  • @you tell us about what you’re trying to solve.

Notes:

Christy Heaton

Works at SGSI - woop!

Worked on a world wide basemap using Spectrum Spatial, but needed all data in TAB format. They had some data but needed everything else from the world. Therefore, they went with OpenStreetMap for non-US data!

The plan: download OSM data into some kind of database > convert into TAB format > add to serverside repository > style the maps as they needed. They decided to use PostGIS databases and split up per country to keep things small. Used python to extract data from databases and added to repository.

Christy helped collect the data and wrote the python to extract it. The python uses QGIS and requires the scripts to be run within the python window in QGIS. This was tough because QGIS module documentation is super limited.

The first python script was simple - it went into each DB and extracted info from each table and turned it into a shapefile. This didn’t do everything they needed, so she wrote a second version that takes specific information based on the columns they need and puts them into a useful structure. It picks out specific columns to minimize file size.

Initially, there were some problems. and fixes

  • Tranlation to tab format was unsuccessful. Some data were missing some huge chunks of data. There were no errors in the script, but data showed no roads. This was due to memory errors because Germany has HUGE amounts of data and the CPU couldn’t handle it. broke these datasets into smaller tables to reduce the load.
  • Tags were a huge issue. OSM tagging is unstructured, which means you could be searching for one tag but the data you actually need exist as another tag. Requires a lot of tag searching. tag queries just got really long and included all potential tag names to get all needed features
  • Points of interest were way too vague, and didn’t return expected results
  • Large rivers were missing since they didn’t get tags for those
  • needed place names in English, but most places in OSM were in local language. obtained english names from an external database

Lessons

  • Tags can be all over the place
  • Files sizes are HUGE, watch out!
  • Consider your output projection and make sure it is projected properly
  • Be flexible, if things don’t go right the first time - consider some alternative workflows

Joseph Sheedy

Joseph made a tool called d3-grid-map, a package that plots gridded data sets. Allows you to hover over pixels to get values from them. Needed to plot gridded data sets in real time. Needed binary arrays from a server and plot them fastly in ANY projection (non-mercator). Required adaptable color maps. Needed to query values (mouseover) and do analysis on those. Also required a universal data format.

Made for a project called “Sea Around Us”, which maps global fisheries data. This tool allowed them to show a bunch of different grid data sets. It’s not terribly fast, but it is “barely reasonable” (his words).

How does it work?

Here’s how you would initialize it in an app:

d3.geo.GridMap('id', options);
map.addLayer(countries);
map.addLayer(dataArray, {gridSize: [720, 360]});

Data format

  • Can’t have anything other than a rectangular grid
  • Float32Array for space saving
  • Options must contain gridSize with width and height array

Transferring data into the browser

Using standard XHR requests in browsers using the arraybuffer response types and turning them into javascript Float32Array objects. Then you can use the above code to add to a grid along with the width and height.

You can put as many layers of things on top of the map as you’d like. For each layer he has cached it, but it changes when you rotate the map around and when you change the container. As long as the map stays in one location, you can recycle that calculation.

Check out the sweet example of pressure systems across the world: velotronheavyindustries.com/wx - takes a while to load 30mb.

Joseph pushes it to the limits. If you aren’t pushing your browser to the limit, you aren’t doing it right.

Hotrodding - let’s speed this sucker up

How can we make it faster? Golden solution is to get all calculations running on GPU instead of the DOM. Potentially a custom d3 build that doesn’t rely on the DOM? Whoa. Hopefully someone else has some ideas?

Thanks to Vulcan for letting him open source this thing. They are hiring!

Paul McCombs in place of Clifford Snow

Paul is talking about Seattle’s bid for the 2016 State of the Map US conference. Here are the details:

we will be presenting a bid to host the 2016 SOTM-US in Seattle. We’ve contacted Microsoft to see if they can give us a good deal on conference space. (MS already donates their satellite imagery for tracing.) Some number about the conference.

  • 600 people are expected to attend
  • Three tracks of presentations over a two day weekend event
  • Summer event but are flexible if the cost is right
  • One “hack” day for workshops and coding. We are looking for space. UW would be a great location!
  • Drone demo would be great!
  • Deadline is Feb 15th.
  • Anyone interested in volunteering to help with the bid should contact me
  • We’ll reuse last years bid as much as possible.

Aaron Racicot

Snuck himself into the agenda late. Talking about projfinder.com and bboxfinder.com. He wants to turn them into CUGOS projects (take them away from his personal github account). They are great projects to hack on as a group that utilize a lot of concepts that we talk about meeting to meeting. He believes these projects have a good opportunity to be focuses for the spring fling.

He’s been getting a ton of inquiries about Projfinder (specifically to the API backend for , which is a tool that allows you to find potential localized projections in case you don’t know what your dataset’s project is set in. Let’s do a few things:

  1. update database on the backend, make it filter better
  2. front end, make it more generic and usable for a general audience

Ryan Small and the (increasingly) annual Spring Fling

SPRING FLING IS ON APRIL 15th, 2016

Check out his

Next meeting: 02/17/2016   Previous meeting: 12/16/2015 All meetings