Drawing a dataset from CartoDB's API without geo tools

My first approach to resolve this problem was focusing in GIS theory. But I was stuck with some cases and I was sure there was another way plus simple. Finally, I applied image drawing techniques and all was good. The only GIS thing is the conversion latitude / longitude to x / y pixels. In this case, I use an equirectangular projection, but we can use another projection.

In this moment, the application fetch the whole dataset from CartoDB API (70 MB) in GeoJSON format, extracts the geometries, calculates the bounds geometries, calculates the scale with these bounds, and draws the polygons. I implemented the panning and an zoom system that needs improve. All this in 4 seconds.

There is few points for improve performance. In server calls side:

• simplify geometries with PostGIS or/and with TopoJSON, or codifying the data for reduce the space needed
• parallel downloads with webworkers or similar
• client caching, an simple memory hash or maybe a client database
• call data within the view’s bounds

In client rendering (needs monitoring):

• now, every polygon starts a context.beginPath() ans ends with context.closePath(). Maybe I can call these functions with groups of polygons, or only one time globally.
• maybe, ordering the polygons
• maybe, filter only the polygons that will be within the bounds before drawing
• maybe, draw groups of polygons in many canvas layers
• maybe, draw the dataset in an in-memory canvas and batch the result as image in the main canvas

Of course, investigate the techniques in HTML5 gaming, HTML5 video rendering, streaming, database streaming (like pipelinedb), BI, Big Data (MapReduce), etc. And monitoring, monitoring, monitoring and deciding.