Luis Sevillano LuisSevillano

Geocoding 10,000 addresses a second with NYC's Geosupport library and Node FFI

Following on Chris Whong's excellent writeup of how to make calls directly to NYC's Geosupport client and this first attempt at generalizing it, here's a way that let me geocode about 10,000 addresses a second on Ubuntu using Node FFI.

Note: this assumes Ubuntu - other Linux is probably fine but may need adjustments.

First, install the basics:

# Update, install Node and unzip (if needed)

Convert SVG to an image on-the-fly.

This article from MDN explains the process of rendering a DOM object on a canvas element: Drawing DOM object into a canvas. Rendering to an <img> element is very similar. This works for any DOM object, not just SVG. Not sure if this is well supported by older browsers. Works on Chrome, Firefox and IE 10 for me. The bar chart is a fork of this block by Mike Bostock: Canvas Bar Chart.

This is a scatterplot using two ordinal scales and the RangePoints() function, with data showing the country of origin of films in competition at the Cannes film festival since 2006.

	license: mit
	height: 1000
	scrolling: no
	border:

	import * as d3 from 'd3';
	import React from 'react';

	const top = 1;
	const right = 2;
	const bottom = 3;
	const left = 4;
	const epsilon = 1e-6;

	function translateX(scale0, scale1, d) {

	var multiCrowbar = (function() {
	/*
	* SVG Export
	* converts html labels to svg text nodes
	* will produce incorrect results when used with multi-line html texts
	*
	* Author: Gregor Aisch
	* based on https://github.com/NYTimes/svg-crowbar/blob/gh-pages/svg-crowbar-2.js
	*/

	<!DOCTYPE qgis_style>
	<qgis_style version="0">
	<symbols/>
	<colorramps>
	<colorramp type="gradient" name="Magma">
	<prop k="color1" v="0,0,3,255"/>
	<prop k="color2" v="251,252,191,255"/>
	<prop k="stops" v="0.04;4,4,21,255:0.08;14,10,42,255:0.12;26,16,65,255:0.16;40,17,89,255:0.20;57,15,110,255:0.23;74,16,121,255:0.27;90,21,126,255:0.31;105,28,128,255:0.35;121,34,129,255:0.39;137,40,129,255:0.43;153,45,127,255:0.47;169,50,124,255:0.51;185,55,120,255:0.55;202,62,114,255:0.59;217,70,106,255:0.62;230,81,98,255:0.66;240,96,93,255:0.70;247,113,91,255:0.74;251,132,96,255:0.78;253,151,104,255:0.82;254,170,116,255:0.86;254,188,130,255:0.90;253,207,146,255:0.94;253,225,163,255:0.98;252,243,181,255"/>
	</colorramp>
	<colorramp type="gradient" name="Inferno">

	// Pipe to ffmpeg with something like:
	// node draw.js \| ffmpeg -y -c:v png -f image2pipe -r 20 -i - -an -c:v libx264 -pix_fmt yuv420p -movflags +faststart myvideo.mp4

	var Canvas = require("canvas"),
	d3 = require("d3"),
	topojson = require("topojson"),
	rw = require("rw"),
	world = require("./world-110m.json");

	var width = 1280,

	/*

	NZZ Euro 2016 Orakel

	http://nzz.ch/-ld.17757

	The algorithm


	this works like this: