FOSS4G-NA notes from PeterK

29 May 2013

##FOSS4G-NA, May 22- 24, 2013## Topics and Highlights from Minneapolis, MN by Peter Keum

Stats: 375 registered, 11% women, 34 states ###Day 1### Morning Keynote Erek Dyskant - Team lead for geospatial analytic for DNC 2012.

  • Self service tools and data are KEY to analyst
  • GRASS to script to run travel time model using OSM data
  • User expericne: only two layers allowed, no layer reordering, simple styling, and focus user experience
  • End users to choose: raw database level, QGIS/GRASS, web service, web browser
  • Server side platform - PostGIS solved many problems
  • Tools to drive work flow “How can we use that info to drive action?”“
  • Tagged data and show more ways to display them - tabular, time series and graphs

Presentations

Shaping OSM into global basemap - AJ Ashton Methodology on using OSM to create various basemaps used in MapBox product.

  • MapBox uses customized basempa using OSM data
  • Using OSM to create basemap still feels like art project
  • OSM data Features
    • Nodes - points and vertices
    • Ways - lines or polygons
    • Relation data - no GIS equivalent
    • Tagging system - no set schema or pre-defined feature classes, conventions by consensus & popularity
    • Have to look at surround tags to identify some objects in OSM data
    • Difficult to use OSM without pre-processing
  • Creating basemap using OSM is like creating a sculpture of art project
  • OSM data + natural earth data = good basemap
  • http://taginfo.openstreetmap.org/ site to inspect commonly used tags
  • Site to read only API, alternative to OSM - explore particular facets of data http://overpass-turbo.eu/

  • http://ideditor.com/ Editor tool to make additions & corrections to OSM data

  • Tools to make basemap
    • osm2pqsql & postgresSQL (postGIS, hstore, PL/pgsql, PY/pgsql)
    • Create sensible data structure by reducing the large number of table columns - one table for water features, one for road etc.
  • Do data transformation
    • Just bring basic data - id, geometries and tags in a single store column
    • Simplified geometries for low zoom levels

OSM iD Editor - Tom MacWright, MapBox

  • Slide Presentation - http://macwright.org/presentations/id-foss4g/#0
  • It is built from pure-javascript to edit and to edit data into OSM using web browser.
  • Very easy to use
  • iD editor displays various data layers according to geogrphaic areas (euro imagery, US etc)- scale dependency
  • It uses tiny libraries - GeoJSON, convert GPS and KML to GeoJSON
  • Only 10,258 javascript lines
  • Motto used “You are not going to need it”
  • Libraries - Rtree, d3, also built libraries - osm-auth, ohauth
  • edit-GeoJSON tool from Tom - Draw feature and return GeoJSON http://macwright.org/edit-geojson/

High Performance Data Visualization in Javascript - Vladimir Agafonkin “Mr Leaflet” Method to Increase Javascript Performance

  • Data visualization - lots of types
    • Charts, maps, became more interactive and dynamic
  • Navigate through data
  • Demand for real-time interactivity is increasing
  • Processing data is shifting from the server to the browser more recently
  • Pure javascript - very fast but rendering in DOM is very slow
  • To Increase speed
    • Data processing first then show to render
    • Data reduction (in order to render at browser)
    • load the data once then search the data later to increase the speed
    • Using Grid method - load the data then only pick the data that falls under viewing box but still not fast enough
    • Use tree data structures - binary help, range tree, k-d tree, quadtree, R-tree
    • Crossfilter (many dimensions)
    • Cohon-sutherland clipping - for line clipping
    • Sutherland-hodgemann algorithm for polygon clipping
    • Douglas-Peucker simplification & distance based simplification
    • Simplification line/polyline library http://mourner.github.io/simplify-js/
    • Clustering - grouping of the objects and increase performance used hierarchical clustering used in Leaflet
    • Data loading and processing - UI-js-UI(browser), freezes on large data so solution is to use web worker js.
    • Loading and sending to Worker using multi-thread
    • Used in transferable objects in Chrome and Firefox - data is sent as an array buffer in the browser to prevent freeze
    • asm.js soon to be implemented in Chrome and now Firefox - allows the computational intense to relieve bottleneck (Didn’t quite understood this web worker js methodology)
    • Rendering SVG uses low number of objects- ltos of stuff to draw use Canvas (if you want to render lots of objects)

CartoCSS for Styling Maps - Tom MacWright, MapBox

  • Slide Presentation :http://macwright.org/presentations/carto-foss4g/#0
  • Developed for styling language for TileMill
  • Labeling and anti-aliasing etc for easy and beautiful maps
  • CartoDB uses cartoCSS styling language
  • Copy and paste is very important
  • Stylesheet Language Ph.D thesis http://people.opera.com/howcome/2006/phd/
  • Not designed for inter-operable but for mapnik only
  • Has variables, nesting selectors
  • Written in javascript and run in node.js
  • Make assumption to use in spherical projection (WGS 84)
  • What’s important about CartoCSS
    • very fast to render
    • very friendly, gives nice error reporting with flag
    • design to show part of the properties and push tab, it’ll show up different options.
    • it has very good in-line help

ESRI and Large Data: Andrew Turner, ESRI

###Day 2###

Keynote speaker - Eric Gundersen, MapBox (CEO of MapBox).

It was very inspirational and positive keynote speech.

  • MapBox goal - make easy maps for people, be a canvas for everything location
  • Moving to open source that can be made into money business (using open source tools and open data) - “dog food” company of 30 folks.
  • Building in the open makes better product
  • The value is in the packing
  • 100% Development Seed is going into MapBox
  • Wants to go against Google, Bing, ESRI, MapQuest, TomTom, Nokia
  • Using open source and open data gives very large leverage that go against the giants
  • The global geo service industry is valued at up to $270 Billion, 30% increase every year
  • Keep building as fast as possible
  • Amazon & Node.js, backbone for MapBox (infrastructure availability)
  • Using Amazon web services as distributor
  • Open Street Map as data to be used
  • Find a tool to fix the errors very fast and constantly, that’s the key to success
  • Released MapBox satellite recently
  • TileMill design studio- to create basemap and other maps using CartoCSS style sheet language
  • Feb 2011, lunched TileMill
  • Find the way to iterative and keep building
  • Create more opportunity instead of competitive
  • Blog constantly but do repetitively and recursively
  • Started the business w/ consulting and use that as seed to develop and open products
  • Future of making better software is to open method and share with others
  • Created vector tiles
  • Map is going to be very central focus in lot of things
  • Value is not in the code but the value around it (hence importance of open source) packaging
  • Invest heavily in design, developing the best design not necessary.
  • “How are you ever going to create anything interesting if you are terrified of cannibalizing your own work?”
  • “the value is not in the code(data), it’s in the service around the code”
  • “be consistently open and communicate repetitively”
  • “foss4g crowd” resurgent of spatial”

Mobile, Offline Maps -Justin Miller, MapBox

  • justin@mapbox.com / @incanus77
  • MapBox 115 repo at github
  • TileMill desktop app for making maps
  • MapBox iOS SDK - rebuild of apple’s mapkit
  • Works w/ MapBox, OSM and various other tiles
  • Add map tile then add point/polyline overlays
  • 15k line of codes
  • Problems w/ off-line mapping
    • very large number of tiles
    • transferility of tiles over network or USB painful
    • maintaining data integrity
  • Traditionally file storage system (x,y coordinate)
  • Better ways to manage the tiles, hence created MbTile (SQLite, self contained and server-less database), simple and easy to grok, fast and light weight –> MbTiles
  • MbTile is open standard tile storage, one SQLite contains all the necessary tiles in the directory
  • Place tiles into database associate with x and y and binary png file then insert into database in SQLite - MbTile
  • Building the stack - TileMill exports, iOS off line, WhirlyGlobe/Maply
  • Preview content for MbTile
  • MapProxy/Arc2Earth export out to web
  • Futher benefits -made to file size smaller, 3GB to 1GB to 300 MB -de-duplication, blue ocean tiles, light sand tiles ect, can get in vector -Future plan - vector tiles (efficient storing data client-side) and rendering at client side, style on-the fly, overzooming

GeoJSON is Spectacularly Wrong - Sean Gillies

  • Slide Presentation: http://sgillies.github.io/foss4gna-2013-geojson-is-spectacularly-wrong/#/
  • GeoJSON is very flexible and evolable
  • Likes json, minimal, portable and textual, self-describing, subset of javascript
  • Order of coordinates is wrong, lat/long not long/lat
  • It doesn’t confirm to ISO 191***
  • Good for programmers
  • It’s flexible (tillable)
  • Some things not correct but very useful

Moving up to an Enterprise open source geospatial plaform - Nancy Read, Metro Mosquito Control District, MN.

  • Efficiency, accuracy and transparency of data to control mosquito
  • Issue (want to create common view to take various spatially reference data) and make into the table
  • How to build business case for making enterprise open source -identify immediate ROI based efficiency (document the open source) -money for license ersi product, they can use that fund to make customization of OS tools
  • field data entry: all collected data go to same location
  • due to turn ovver staff, customization software vs training for new people. Decided to go w/ customization with minimal training
  • Business case
    • powerful tools
    • easy access for manager, pilot, public
    • custom software and can share development w/ other agencies
    • avoided per-seat licenses fees
    • treat it as capital expenses (w/ cycle of maintenance and operation)
    • more money for development in first year, no additional cost to scale the application other than cloud server, allowed for great flexibility for custom development

Automatically geotagging unstructured text with open source tools- Charlie Greenbacker, Berico Technology

  • Ingest unstrctured text
  • Extract place names - geo entity extraction
  • Diambiguate names - geo entity resolution
  • Named entity recognition
  • Using open source tools, named entity recognition - standford natural language process to pull out names from the content.
  • Using GeoName.org to show names of the places
  • Intelligent context-based heuristic - rank by population
  • Use CLAVIN - location resolver http://clavin.bericotechnologies.com/

Leaflet - Past, Present and Future - Vladimir Agafonkin He was very positive and excited about the project.

  • Leaflet History
    • started CloudMade April 2008
    • first slippy map w/ 200 line of codes then built full-feature API
    • came w/ very minimal geographic knowledge
    • first product: “Web Maps Lite” turned into Leaflet
    • very light but for two years, he was only person to maintain and became too much work
    • Open sourced it on Sept 2010
    • Rewrote the libraries from scratch and made it better
    • 2.5 years of development, 6531 lines of code, with over 129 contributors
    • plugins 63 plugins (examples: vector editing, clustering)
  • Future of Leaflet
    • Simplicity very important
    • number of features will be reduced
    • just basic features
    • focus on quality of features instead of quantity of features
    • keep API as simple as possible
    • convention over configuration/ simple inside and outside
    • complexity limits contribution and github has revolutionized the code contribution
    • need to be not just open but TRANSPARENT
  • Suggestions for foss4g software developers
    • make it simple, refractoring for flexibility
    • improving performance -improving usability -improving plugin infrastructure -better web site, -more tutorial sites -“get excited, build cool stuff, believe in yourself”
Next CUGOS meeting: November 20th 2024

Code of Conduct

CUGOS strives to be an inclusive community for anyone interested in open source geography. Before attending our meetings, please familiarize yourself with our Code of Conduct.

About CUGOS

Learn more about CUGOS, our mission, and our annual reports.

Want to stay up to date with CUGOS news? Join our Google Group.

You'll have to request access, we'll approve you ASAP.