Home - Vanuan/research GitHub Wiki

OSM Issues

Research

Database size

Are RDBMS size effective?

How to test:

  1. Load Planet.OSM into PostGIS
  2. Do the same for GeoCouch

Performance

Does NoSQL provide a better performance?

Conclusion:

Based on the research on comparing NoSQL to RDBMS in this field of application: : http://igitur-archive.library.uu.nl/student-theses/2012-0822-200532/MSc_report_final_v101.pdf

NoSQL is no competitor to PostGIS (probably just yet). The main advantage of NoSQL is that is scales well. Until PostGIS won't reach the critical point, it'll suit best for general purpose. NoSQL wins in certain applicatons, but the price is greater memory and storage usage.

Design

GeoJSON is a must. TopoJSON provides 80% smaller size.

Is it possible to use D3? SVG is much faster than canvas. http://bost.ocks.org/mike/map/ http://blog.cartodb.com/post/39680106243/cartodb-makes-d3-maps-a-breeze

Kothic looks promising. Although it's extremely slow.

Leaflet supports geoJSON: http://stackoverflow.com/questions/11570669/leaflet-geojson-display?rq=1

Interesting article on line simplification: http://bost.ocks.org/mike/simplify/

As a first step REST API implementation for GeoJSON is needed. The two possibilities are here:

  1. api/{z}/{x}/{y}.json
  2. api?sql=select * from ...

The first option has an advantage in terms of performance.

The second option has an advantage in terms of flexibility (and downloading size).

To create a convertor, we need to map OSM features to GeoJSON: https://gist.github.com/1396990

Implementation

  1. translate kothic json_getter.py to javascript:
  • twms analog for js?