GeoJSON-triggered Chrome crash test case

[tmcw]

• http://bl.ocks.org/tmcw/raw/e256d5b639d3b4705cdc/ffeac6fbdbf27202ed21bec92f85b648f9f6af8c/

Triggers a sad-face tab crash in Chrome Version 38.0.2125.104

[jfirebaugh]

Upstream: https://code.google.com/p/chromium/issues/detail?id=426116

[jfirebaugh]

Upstream says it's an OOM crash.

[tmcw]

Interesting. I wonder if there's something we can do to make this consume less memory? 1024 polygons are included, and are within the realm of Leaflet etc.

[mourner]

@tmcw this happens probably because you're fetching this GeoJSON into the main thread, and then passing it to the worker. Loading big GeoJSON data in the worker directly (by passing the URL in the GeoJSONSourcce options instead) should work fine.

[tmcw]

I'll try that. Can you say more about what the difference is here? The size of the geojson isn't super huge - 600kb. Is the memory thing in the JSON transfer, or something else?

[mourner]

Yeah, JSON load + transfer. Of course I'm not sure it's the culprit, but this is the first thing to try.

[jfirebaugh]

It's not the JSON transfer, it's tiling the GeoJSON. GeoJSON is tiled at zooms 1, 5, 9, and 13. This GeoJSON generates 29700 tiles at z9 and 536580 tiles at z13, and the worker either takes forever to process that (master), or crashes (0.4.1). So basically this reduces to #464.

[mourner]

Going to take a crack at fixing this, see #464 (comment)