web.config for React and Node.js on Azure App Service

I’ve been creating a web application using React and Node.js, and decided to host it on Azure App Service. I had a lot of difficulty finding an appropriate web.config rewrite rules that would host static assets, redirect deep React Router links to index.html and send API calls to Node.js via iisnode. I finally came across this post which got me most of the way there, and I just had to adapt it to use Node.js:

WebdriverIO with ChromeDriver headless without Selenium

I was excited to discover that WebdriverIO supports the use of ChromeDriver running in headless mode without the use of Selenium (because ChromeDriver supports the WebDriver protocol). Here’s a starter guide (assuming Node.js is installed):

  1. Install Chrome Web Browser
  2. Download ChromeDriver
  3. Run ChromeDriver
  4. Install WebdriverIO by running npm i webdriverio
  5. Run the following file (based on the WebdriverIO example)

const webdriverio = require('webdriverio')
const client = webdriverio.remote({
  host: "localhost",
  port: 9515,
  path: "/",
  desiredCapabilities: {
    browserName: "chrome"
  .setValue('#search_form_input_homepage', 'WebdriverIO')
  .getTitle().then(function(title) {
    console.log('Title is: ' + title);

To run Chrome in headless, use the following config:

const client = webdriverio.remote({
  host: "localhost",
  port: 9515,
  path: "/",
  desiredCapabilities: {
    browserName: "chrome",
    chromeOptions: {
      args: ["headless", "disable-gpu"]

Creating a ArcGIS map service raster layer in PyQGIS

It’s relatively straightforward creating layers programatically in QGIS using PyQGIS. However, the documentation isn’t clear on how to use create a raster layer from a ArcGIS map service.

The list of raster providers can be found by searching for addRasterProviderDialog in qgsdatasourcemanagerdialog.cpp.

There are two mandatory parts to the data source string when specifying the ArcGIS map service, the url and layer, which need to be space separated. For example:

iface.addRasterLayer("url='http://gis.infrastructure.gov.au/infrastructure/rest/services/KeyFreightRoute/KFR/MapServer' layer='5'", "Key road freight routes", "arcgismapserver")

or using the constructor:

layer = QgsRasterLayer("url='http://gis.infrastructure.gov.au/infrastructure/rest/services/KeyFreightRoute/KFR/MapServer' layer='5'", "Key road freight routes", "arcgismapserver")

To understand what data source parameters are supported by the ArcGIS map service provider (AMSProvider), search for dataSource.Param in qgsamsprovider.cpp.

Websockets are easy

I was interested in getting a minimal example working for Websockets and it was surprisingly easy to get a demo working between Node.js and a browser.

First install the ws library:

npm install ws

Create an index.js file with the contents:

const WebSocket = require('ws');
const wss = new WebSocket.Server({ port: 8080 });

wss.on('connection', function(ws) {
  console.log('Starting connection');
  ws.on('message', function(message) {
    console.log('Received message: ' + message);
    ws.send('You sent ' + message);

Start the Node.js process:

node index.js

In a browser console, send a message to the Node.js process:

var ws = new WebSocket("ws://localhost:8080");
ws.onmessage = function(e) {
ws.send('Hello World');

Re-rendering map layers

I recently was optimising the performance of a Leaflet-based map that rendered TopoJSON layers via Omnivore. The layers were a visualisation using the ABS’s Postal Areas, and while there was only a single TopoJSON file, this resulted in a number of feature layers being displayed, with each bound to their own data. The data for each feature layer could change depending on what filters were set (these filters were displayed in a left panel).

In the original implementation, whenever a change was made to a filter, the entire TopoJSON layer was removed, and then re-joined to the data set and rendered:

var dataLayer;
function renderDataLayer() {
    var prevLayer = dataLayer;
    dataLayer = L.geoJson(null, {
        filter: filter,
        style: style,
        onEachFeature: onEachFeature
        .topojson('postal_areas.topojson', null, dataLayer)
        .on('ready', function() {
            // Remove previous layer when current layer is ready to avoid flickering
            if (prevLayer) {

Using Chrome’s “Record JavaScript CPU Profile”, it clearly showed the code invoking Omnivore was the problem:

Changing the render function to (1) iterate over the existing layers and (2) change the style of the layer, made the map feel a lot more responsive.

var dataLayer = omnivore.topojson('postal_areas.topojson', null, dataLayer);
function renderDataLayer() {
    dataLayer.eachLayer(function(featureLayer) {

One of the disadvantages was that we can’t use the filter function on the geoJson object anymore, since in order for a style change to occur, the feature layer must be present (although it can be hidden). This potentially can lead to slowness when dragging or zooming a map.

Another related disadvantage is hiding layers both visually and from mouse events. Setting the opacity and fillOpacity of the layer to 0 will take care of the first, while this CSS will prevent the mouse cursor from changing when hovering over a hidden layer:

/* Don't show pointer on hidden layers */
.leaflet-pane > svg path.leaflet-interactive[stroke-opacity="0"][fill-opacity="0"] {
    pointer-events: none;

Templated SQL

I’ve recently been working on a data migration SQL script that performs the same operations against numerous tables. While I could use dynamic SQL – where SQL statements are built as a string and then executed – I don’t particularly like the downsides:

  • It cannot be checked at compile time
  • It is difficult to read, debug and reason about
  • I’d have to use a cursor to iterate through each of the table names

Since this was a one-off migration task, I decided to render the SQL dynamically ahead of time and wrote a little utility in Node.js. So the template contained the SQL statements I wanted to execute, while the data contained a single column with the table names. For example:




ALTER TABLE {{table}} ADD col VARCHAR(50);



Wrap the output in a BEGIN/COMMIT TRANSACTION and we’re done.

Leaflet and Google Maps

I’ve recently been developing an application that uses Leaflet to interactivity with a geographic map. One of the business requirements was to use Google Maps as a basemap, since it is pervasively used by our customers. A naive implementation used Leaflet’s tileLayer to render the tiles directly:

var map = L.map('map').setView([-29, 133], 4);

This approach, while simple, does not conform to the Google Maps API Terms of Service. Section 10.1.a indicates that:

No access to APIs or Content except through the Service. You will not access the Maps API(s) or the Content except through the Service. For example, you must not access map tiles or imagery through interfaces or channels (including undocumented Google interfaces) other than the Maps API(s).

To conform with the Terms of Service, I initially tried using leaflet-plugins, which creates a layer using the Google Maps Javascript API. However, when panning the map, the Google Map tiles lag compared to vector layers (e.g. polylines, polygons). This is a known, but unresolved issue and occurs because the Google Map setCenter method is asynchronous and thus will get out of sync with the Leaflet controlled elements.

I next used a wrapper around leaflet-plugins. The wrapper (1) hides the leaflet-plugins Google layer and (2) finds the right tile image in the DOM (rendered by leaflet-plugins) and renders it when it becomes available. Here is a simplified example:

var GoogleGridLayer = L.GridLayer.extend({
  // googleLayer is a leaflet-plugins Google object
  initialize: function(googleLayer) {
    this.googleLayer = googleLayer;
  onAdd: function(map) {
    L.GridLayer.prototype.onAdd.call(this, map);
    // Hide the leaflet-plugins layer
    $(this.googleLayer._container).css("visibility", "hidden");
  onRemove: function(map) {
    L.GridLayer.prototype.onRemove.call(this, map);
    $(map._container).find("#" + this.googleLayer._container.id).remove();
  createTile: function(coords, done) {
    var img = L.DomUtil.create("img");
    var googleLayer = this.googleLayer;
    var interval = setInterval(function() {
      var src;
      var id = "#" + googleLayer._container.id;
      var googleImg = $(id + " .gm-style img").filter(function(i, el) {
          var src = $(el).attr("src");
          return src.indexOf("!1i" + coords.z + "!") > 0
                 && src.indexOf("!2i" + coords.x + "!") > 0
                 && src.indexOf("!3i" + coords.y + "!") > 0;
        if (googleImg.length) {
          googleImg = googleImg.first();
          src = googleImg.attr("src");
        if (src) {
          img.src = src;
          done(null, img);
    return img;

This approach seems to work reasonably well and is as responsive as the original tile layer implementation.

Other enhancements I added later were:

  • Caching the image src rather than having to poll the DOM on every create tile request
  • Adding support for Satellite and Hybrid maps (the latter requires a div with two nested images – one for the Satellite tile and one for the labels/roads tile)
  • Adding a timeout to polling – there are certain tile requests by Leaflet that aren’t fulfilled by the Google tile layer. It may be that Leaflet tries to load tiles past the viewport
  • Adding attribution (including logo and copyright) by moving some of the .gm-style elements in front of the GoogleGridLayer. The Google Map stores a reference to these elements and appropriately changes their content.

One of the disadvantages at this stage is that there is no easy method to access Street View coverage tiles without simulating a drag of the Pegman. It would be nice if the Google Maps Javascript API supports a function to turn on/off the Street View coverage tiles.

Organise photos by date

I organise my photos by grouping them in a directory, where the directory takes on the name “created_date title” (e.g. 2016-05-16 Birthday Party). As it became tedious to create a directory and manually group photo sets, I wrote a Ruby script to assist with the heavy lifting. After that, it’s just a matter of providing a title to each created directory.

require 'exifr'
require 'fileutils'

Dir['*.jpg'].each do |file|
  date = EXIFR::JPEG.new(file).date_time
  if date
    folder = date.strftime('%Y-%m-%d')
    FileUtils.mkdir_p(folder) unless File.directory?(folder)
    FileUtils.mv(file, folder)

Note: you will first need to run gem install exifr to install dependencies.