From Wikipedia, the free encyclopedia
Jump to: navigation, search


Hello world!

My interest in editing Wikipedia is in illustration, so let me know if you wish some article on science, technology, architecture or mathematics illustrated.

Here is some of my work to date:

Atmospheric composition Langley.svg
About this image
CIE1931xy ColorChecker SMIL.svg
About this image
Drawings and
3D renders
vector graphics
Jane Coston cycle bridge deck.jpg
About this image
Photographs Maths and tables
Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Halley's Comet Sun Eris (dwarf planet) Makemake (dwarf planet) Haumea (dwarf planet) Pluto Ceres (dwarf planet) Neptune Uranus Saturn Jupiter Mars Earth Venus Mercury (planet) Astronomical unit Astronomical unit Dwarf planet Dwarf planet Comet Planet

Distances of selected bodies of the Solar System from the Sun. The left and right edges of each bar correspond to the perihelion and aphelion of the body, respectively, hence long bars denote high orbital eccentricity. The radius of the Sun is 0.7 million km, and the radius of Jupiter (the largest planet) is 0.07 million km, both too small to resolve on this image.

−ve +ve
+ve Science All
−ve Teachers Crazy
Charts using
Templates for
general use selected pieces of interest...

Cmglee logo.svg


One million dots 1080p.png 28 Aug 2017[edit]

Visualisation of powers of ten from 1 to 1 million
One million black dots (pixels) – each tile with white or grey background contains 1000 dots (full image)

I noticed that a full HD screen has just over 2 million pixels (1920 × 1080 = 2 073 600) and had the idea of a two-dimensional visualisation of the size of one million to complement my three-dimensional one to the left.

By using a checkerboard pattern, I could show one dot for every two dots. The naive approach would result in a big blob of a million dots without any grouping, so I wondered if there might be a way to separate them into 1000 tiles of a thousand each. Having 10 rows of 100 columns would "use up" 9 rows and 99 columns of pixels to separate the tiles, resulting in 1821 × 1071 = 1 950 291 pixels, just shy of the 2 million pixels needed. However, 10 rows of 10 columns gives 1911 × 1071 = 2 046 681 pixels, so I'm on the right track...

I could use the background of the dots to separate tiles of 1000 dots. I settled on white and very light grey to make the difference just visible yet not too distracting. Interestingly, 19 × 106 = 2014, i.e. 1007 dots, so each group of 10000 dots can be 190 pixels wide and 106 pixels tall. A 10×10 matrix of these would leave a neat 20 rows and 20 columns of pixels to separate the groups. To make the groups easier to count, I divided them as follows: 2_1_2_1_2_4_2_1_2_1_2, where _ represents a group, in both dimensions. As 7 dots had to be removed from each tile, I removed 3 from one corner and 4 from another, conveniently resulting in ticks appearing in each group.

Hacking up the Python script (which uses the module) on file:one_million_dots_1080p.png, a bit of tweaking to make the graphic symmetrical, and ta-da – 1 million black dots in full HD!

Universe blocks timeline visualisation.svg 31 Jul 2017[edit]

A brief (proportional) history of the time

magine a 138-page book chronicling the history of time. Each page has exactly 100 lines, and each line, exactly 100 letters. The first letters on the first page describe the most recent events; the last letters on the last page, the beginning of the Universe as we know it. If the amount of text is linearly proportional to time, how much of this tome narrates written human history?

Well, with 13.8 billion years denoted by 138 pages of 10000 letters, each letter represents 10000 years. Thus, written history is worth about half a letter!

This graphic shows where some other notable events lie. Note that everything involving humans lies on the first line, representing the most recent 1 million years.

Going down two-thirds of a page, we come to the big die-off of the dinosaurs 66 million years ago. At the end of the first page, 100 million years ago, we're firmly in dinosaur and reptilian territory. Just five pages later (500 million years earlier), complex animals hadn't evolved. After the tenth page (1 billion) years, the most complex lifeforms were green algae.

A third of the book in, the 45th page marks the formation of the Earth, the Sun and much of the rest of the Solar System. Most is unknown about the Universe until perhaps the last 8 or so pages when the first galaxies formed.

The final page, likely the 138th, is once again a hive of activity, where much of the early Universe takes place. Exactly how many lines and letters it has, noone knows...

I hope this analogy gives another perspective on the timeline of the Universe. Though one might bemoan our insignificance in the big picture, I find it amazing that scientists were able to discover these milestones in just the first half-letter.

P.S. At the same scale, how long is the Diary of the Universe to people who think it started in 4004 BC? Six-tenths of a letter! ୯ ͡°  ͜১͡° ੭

horizon distance graphs.svg 24 Jun 2017[edit]

Distances to the horizon along the Earth's surface h, straight-line distance d, and approximate straight-line distance assuming h << Earth's radius ~d, for altitude h
How far away is the horizon.png

The graph on the left employs a clever technique to have gridlines in different units, such as metric and imperial, on the same graph: by having them straddle the line!

I wondered if it might work for any odd number of graphs, and drew this month's graphic. Getting the gridlines to alternate between the curves was tricky, but creating clip-paths by alternating the directions of the paths worked. Each set of gridlines could then be made up of rects textured with their corresponding patterns. I even managed to reasonably smoothly transition from feet to miles for imperial altitude. And of course, multiplicative factors between units appear simply as translations on a logarithmic scale.

It was fun picking reference altitudes, though I couldn't find a good one between ISS and GPS orbits. It's also rather surprising how good the approximation is, and how similar the straight-line and along-Earth's-surface distances are, all the way to the official boundary of outer space.

North Korean missile range.svg 20 May 2017[edit]

Azimuthal equidistant projection of estimated maximum range of some North Korean missiles

User:Ethanmayersweet suggested that I update my unexpectedly popular graphic and I thought it would be the perfect opportunity to update the map to an azimuthal equidistant projection – I'd found it mildly annoying that the radii of the rings weren't to scale, and much of the globe (especially mainland USA, whose readers would most likely be interested in) was hidden.

User:Jason Davies had usefully uploaded File:North-korean-missile-ranges.svg, but I preferred a coloured globe without coastline shading. The other thing I wasn't so keen about was the 5 MB file size. So I wrote this Python script to

  1. convert absolute to relative coordinates in the SVG path,
  2. round to the nearest integer, and
  3. simplify horizontal and vertical movements to "h" and "v" commands, respectively.

The script could have combined multiple "h" and "v", and handle commands other than "M", "L" and "Z", but that will have to wait until another day. That reduced the total file size (even after "beautification" and updating the ranges) to 162 KB, a reduction of 97%!

By the way, I found it curious that UK and mainland USA (and Africa) are just beyond 8 000 km – coincidence?

Menger sponge diagonal section.gif 29 Apr 2017[edit]

GIF animation of cross-sections of a level-4 Menger sponge perpendicular to a space diagonal. The cross-section through its centroid comprises regular hexagrams arranged in six-fold symmetry. The cross-sections are true-view and to scale.
GIF animation showing how an orthographic projection of a tetrix can exactly fill a plane

After hearing about the Menger sponge at a talk about fractals in this year's Cambridge Science Festival, I challenged my companion to guess what a cross-section along a space diagonal and passing through its centroid looks like.

Back home, it occurred to me that I could use the constructive solid geometry feature of POV-Ray to automatically generate cross-sections, and decided to make an animation of the cross-sections along the space diagonal similar to but viewed perpendicular to the section so that each section is a true view.

The biggest challenge was coming up with an efficient way to create the sponge. The method I used to render the level-4 tetrix on the left would result in 204 = 160 000 cubes – far too strenuous for my measly computer!

I eventually decided to subtract three orthogonal sets of long holes from a simple cube. My initial arrangement resulted in artifacts due to holes from different levels overlapping, but I managed to find a set without overlap – see my POV-Ray source code below if you're interested.

Additionally, I wanted to use a perspective projection for a more realistic look, but still wanted to keep every section at the same scale, so I had to translate the cube at exactly the rate which keeps the cross-section plane in place. To confirm that I got it right, I replaced the sponge with a simple cylinder with its axis perpendicular to the plane.

After a day rendering the 55 frames (27 × 2 + 1 – chosen so that every vertex has a plane passing through it), I thought the colour was a bit bland, so modified the palette after converting them into GIF in XnView's batch mode. Good thing I decided to turn off anti-aliasing to increase rendering speed.

I also decided to add pauses at the level-0 vertices, and of course, a longer one at the cross-section through the centroid.

Lastly, I came up with a tidier presentation of my POV-Ray source on the file description page, so that anyone can continue my work:

Now I'm tempted to make more animated cross-sections, and one that comes to mind is a rotating double cone to generate conic sections!

nucleosynthesis periodic table.svg 29 Mar 2017[edit]

Periodic table showing origin of elements in the Solar System

Three years on from its first publication, I've updated my periodic table of the origin of the chemical elements, made infamous by NASA choosing it as Astronomy Picture of the Day. I was surprised by its selection (it wasn't me who submitted it for consideration – I wasn't even particularly proud of it, as I'd quickly hacked it together after attending a talk showing the figure from which it was derived) and the subsequent criticism of its veracity.

Since then, astronomer Jennifer Johnson created a more well-researched version and I decided to combine with her data with some "improvements":

  • The percentages of each origin are represented by squares (out of a hundred) instead of oblique areas to make it easier to estimate proportions – hovering over the SVG gives precise figures as tooltips, too.
  • It uses a light colour scheme to aid printing.

ancient seven wonders timeline.svg 21 Jan 2017[edit]

Timeline and map of the Seven Wonders. Dates in bold green and dark red are of their construction and destruction, respectively.

I became interested in the Seven Wonders of the Ancient World after watching a documentary and read the Wikipedia article. Seeing room for improvement of their timeline, I set about drawing one.

It felt a little clumsy as the massive range of the Great Pyramid of Giza compared to the others caused huge gaps to appear.

I wondered, "Why not fill them with useful information?" and thus added a map and silhouettes to scale, with dimensions obtained as far as possible from their respective Wikipedia articles.

The outlines show how amazing an achievement the building of not one, but two pyramids of comparable size was, 2000 years before the other wonders.

Lastly, from the scale of the timeline, it seems a shame that we just missed living in a time when three wonders existed. If only they hung on for 600 more years!


snow css3 animation example.svg 31 Dec 2016[edit]

Merry Christmas/Happy New Year ambigram

I learnt that SVG filters can be animated using SMIL and added an ambigram that reads Merry/Happy and Christmas/New Year to my CSS snow animation. Happy 2017!

block stacking problem.svg 6 Nov 2016[edit]

The first eight blocks in the solution to the single-wide block-stacking problem
The source texture bitmap

I had just realised that the block-stacking problem article lacked an illustration despite my interest in harmonic series. This presented a good opportunity to experiment with an idea to use affine transformation matrices (which I explored last month) to transform bitmaps to make more realistic pseudo-3D graphics.

In the SVG, the upper-right 3×3 swatches in the image on the left were split into individual textures, each assigned to a pattern. Appropriate affine transformations were applied to the faces of each block, some overlaid with semi-transparent black to simulate shading, and defined as a group. The fill parameter of each group was assigned a texture pattern, and the block moved into position.

It's a shame that SVG doesn't do perspective transforms (even though CSS does), so I'm limited to parallel projection. Still, that suits technical illustrations where conceptual clarity takes precedence over geometric fidelity.

Expect to see more textured SVGs!

UPDATE 18 NOV 2016:
A rotating textured pseudo-3D model

cmglee trimetric template.svg 21 Oct 2016[edit]

My trimetric projection roughly approximates a cube with small integers

Back in 2012, I discovered an aesthetically pleasing 3D trimetric projection that lets one construct semi-realistic-looking scenes using vectors with components comprising small integers, namely

  • (4
    for the x-axis (up and right)
  • (−3
    for the y-axis (up and left)
  • (0
    for the z-azis (directly upwards)

assuming the image y-axis is positive downwards. An added advantage over dimetric or isometric projections is that the rear edges and vertex of a cube does not overlap with a front one, so that they can be clearly shown. Curiously, a 2 × n × 1 block has the same width and height.

I also found some useful affine transformation matrices to automagically apply rectangular textures onto the three faces of a cuboid:

  • Top face: matrix( 4,-2, -3,-3, 0,0)
  • Left face: matrix(-3,-3, 0, 4, 0,0)
  • Right face: matrix( 4,-2, 0, 4, 0,0)

As can be seen from the "F" pattern, I had to mirror the left face horizontally and the top face vertically so that they could all start at the nearest vertex. This has negligible effect on symmetrical patterns such as gridlines, but asymmetrical ones have to be either pre-flipped or the matrices must be changed.

With this useful shape, I drew a variety of illustrations including these:

Template:quadratic_function_graph_complex_roots.svg 30 Sep 2016[edit]

Visualisation of the complex roots of y = ax2 + bx + c: the parabola is rotated 180° about its vertex (orange). Its x-intercepts are rotated 90° around their mid-point, and the Cartesian plane is interpreted as the complex plane (green).[1]
Graph of y = ax2 + bx + c, where a and the discriminant b2 − 4ac are positive, with
  • Roots and y-intercept in red
  • Vertex and axis of symmetry in blue
  • Focus and directrix in pink

A few days ago, I found that Wikipedia lacked an illustration showing the key values of the general univariate quadratic function y = ax² + bx + c and so drew the graphic on the left. Just as I thought my work was done, I found out about a method to graphically find the complex roots of such a function without real roots – eons since I left high school!

Back to the drawing board, I modified my earlier graphic to visualise the method and created the template on the right for easy inclusion in Wikipedia articles.

The trickiest bit was creating the text objects containing mathematical expressions. I employed a neat device in which a fraction is constructed from

  1. A set of tspans containing the numerator,
  2. Another tspan resetting the x-coordinate and containing as many "_" characters as needed – harking back to the original function of the underscore on typewriters!
  3. A set of tspans resetting the x-coordinate, incrementing the y-coordinate by 1em and containing the denominator. For a middle-aligned text object, Firefox and Chrome browsers middle-align the entire denominator, but the Wikimedia renderer middle-aligns only its first tspan. To get "2a " to align correctly, the x-coordinate had to be specified for each tspan.

For example, b/2a is created with the following SVG code (note the wrapping of the line before the ">" character to avoid introducing meaningful whitespace):

 <text transform="translate( 130, 40)" x="0" y="0"
  ><tspan>&#8722;</tspan><tspan font-style="italic">b</tspan><tspan x="0">__</tspan
  ><tspan x="-10" dy="1em">2</tspan><tspan x="10" font-style="italic">a</tspan></text>

I would have preferred the minus sign outside the fraction: −b/2a , but that would have required the extra hack below...

For the coordinates (b/2a, −(b²−4ac)/4a) , I had to manually create a separate "( , )" text object in larger font (with some stretching in the y direction to avoid the brackets looking too thick) to wrap around the two functions. Is there a better way? Let me know! 8 Aug 2016[edit]

Stereographic (top) and equirectangular (bottom) projections of the Last Supper sculpture at Via Crucis in Esino Lario during Wikimania 2016

I thought I'd do a little part of documenting Wikimania 2016 and the village of Esino Lario by shooting 360° spherical panoramas of notable places using the Photo Sphere feature of the native Android camera app of my Nexus 7:

The page also features Dschwen's Pano360 panoramic viewer and "Tiny Planets" generated by the same app.

For optimum results, one should really use a tripod or a specialised camera, as doing it by hand results in many stitching errors – nevertheless, the poor man's method doesn't require much beyond an Android phone or tablet with gyroscopes, and lots of patience!

As I discussed with user:Pigsonthewing on my Commons talk page, in case anyone wants to shoot some, here are some techniques for better results:

  • Choose scenes far from the camera to reduce parallax – best are landscapes, but at least stand up and stay away from railings.
  • If possible, choose scenes without obvious geometric patterns e.g. large floor tiles and roof beams – grass, rocks, foliage etc are best.
  • Avoid taking a shot when moving people are in the scene to avoid half bodies.
  • Pick an identifyable point on the ground and try to keep the camera over it.
  • Pivot around the camera rather than standing still and turning the body; I grasp the device around the lens with one stationary hand and use the other to swing it around the pivot.
  • Reposition or wait until the sun is not directly visible, to avoid overly high dynamic range and lens artifacts.

Shame I discovered them only as I went along, but I suppose one gets better with practice...

Template:Tallest towers in the world.svg 17 Jul 2016[edit]

Comparison of the CN Tower with the world's seven tallest towers
The collapsed pyramid of Meidum shaded

I noticed editors adding translations into File:Phase_diagram_of_water.svg and learnt that, using the switch tag and systemLanguage parameter, an SVG file can contain multiple languages. While visiting Milan's Pinacoteca di Brera after Wikimania, I realised that I could hack this mechanism to highlight different parts of an illustration in different articles. Back home, I successfully tested it on File:comparison_of_pyramids.svg as on the left. I'll use another example as I'd written on the pyramids SVG below.

In its SVG, each highlighted version is amended as follows: <switch><use xlink:href="#id" systemLanguage="CODE"/></switch> where CODE is one of these codes. As codes must be valid, I picked ones which look similar, yet are relatively rare to not conflict with real use: from left to right, ts, ca, cu, or, os, ml and kl.

To use the SVG, add lang=CODE in the image's Wikitext and it appears as on the right: [[File:Tallest_towers_in_the_world.svg|thumb|lang=cu|Comparison of the CN Tower with the world's seven tallest towers]].

I create the template:Tallest towers in the world.svg to make it easier to use. The first parameter is the code (set to blank for no highlighting) and the second is a custom caption snippet (set to blank to use the name of the page it appears on, a trick described below): {{tallest towers in the world.svg|CODE|CAPTION SNIPPET}}.

SMIL missile demo zh-hant.svg 9 Jul 2016[edit]

Missile game in Chinese
"Wheel of fortune" demo

I had the privilege to share accommodation with user:Shangkuanlc at Wikimania. While chatting about my adventures in dynamic SVG, he was fascinated with this silly little game I made to showcase what could be done with SMIL, and very kindly translated it to Chinese. We talked about other ideas for SVG games – including an Angry Birds parody I named "Anti-Birds", a skeet shooting game where the player tries to snipe, in mid-air, birds about to crash into the usual porcine architecture!

I realised, however, was that games often have random elements but SMIL and SVGs without JavaScript are normally deterministic. After discussion with user:Niedzielski, I thought of a way to simulate random events by rapidly rotating an invisible pinwheel of triggers in front of a button. Due to its rapid rotation and invisibility, the trigger that is clicked is practically random. I created this "wheel of fortune" prototype to demonstrate the concept ("Show triggers" toggles the visibility of the triggers).

The translation effort also revived the idea to rewrite an updated SVG translation tool. I had previously used Jarry1250's SVG Translate but it no longer works, and also requires the user to manually fill in all the fields. I learnt that Pywikibot is a good platform to use, and attended a training session and spoke with user:Yuvipanda. Let's see what I can come up with one of these weekends...

Wikimania 2016 22 Jun 2016[edit]

... so I'm giving a talk and training at Wikimania 2016. Come join us on Saturday 25 June 2016 if you're in town. For a preview, below is my slide deck...

SdKfz 250-3 I (Neu) (3D-animated).gif
About this image

1 Techniques
  1.1 Animation
  1.2 Tooltips
  1.3 Hyperlinks
  1.4 Progressive disclosure
2 Applications
  2.1 Interactive timelines
  2.2 Simple 3D viewer
3 GIF animation to SVG converter
4 Best practice

5 Cheers
Bid Wikimania Esino Lario 2016 slides 07 SMIL.svg
About this image

German tank problem graphs.svg 12 Apr 2016[edit]

Graphs of estimated population size, N for number of samples, k and largest sample serial number, m using frequentist (dotted lines) and Bayesian analysis (solid line shows mean and shading shows range from minimum possible value to mean plus 1 standard deviation). The example shows if 4 racquets are observed and the highest serial number is 60, frequentist analysis predicts N = 74 whereas Bayesian analysis predicts a mean of 88.5 and standard deviation of 138.72 − 88.5 = 50.22, and a minimum of 60 racquets. In the SVG file, hover over a graph to highlight it.

An advantage I thought Perl had over Python was the ability to interpolate or embed variables directly into a string, such as

print "Hello $name";

where Python uses format or % :

print "Hello %s" % (name)
print "Hello {0}".format(name)

which gets clumsy for long strings or a large number of variables.

That was until I learnt that I could do this (as long as name is a local variable):

print "Hello %(name)s" % locals()
print "Hello {name}".format(**locals())
print "Hello {name:s}".format(**locals())
print "Hello {name!s}".format(**locals())

I wondered if it would be even better if name could be an arbitrary expression and not just a variable e.g. name.upper() to capitalise the whole string or len(name) for its number of characters (zero-padded to eight hexadecimal characters). And so I wrote the following:

import re
def fmt(string): ## string.format(**vars()) using tags {expression!format} by CMG Lee
 def f(tag): i_sep = tag.rfind('!'); return (re.sub('\.0+$', '', str(eval(tag[1:-1])))
  if (i_sep < 0) else ('{:%s}' % tag[i_sep + 1:-1]).format(eval(tag[1:i_sep])))
 return (re.sub(r'(?<!{){[^{}]+}', lambda m:f(, string)
         .replace('{{', '{').replace('}}', '}'))

which lets me do this:

print fmt("Hello {name.upper()}")
print fmt("characters:{len(name)!08x}")

If there is no format string, 's' is assumed, but if the expression contains !, such as {'differ' if 1 != 2 else 'equal'!s}, the format string must be specified. To escape braces, repeat it: fmt("{{this is not a tag}} {'but this is'}")

It's far from perfect: if the expression contains }, such as {{'a':1,'b':2}['a']!d}, it can't tell which } the closing brace of the interpolation is, as regular expressions cannot count nested brackets, and error messages are quite ugly.

The example graphic illustrates the German tank problem dating back to WW2. As paraphrased for a sporting mate:

Earth lights vs population density.png 26 Mar 2016[edit]

Earth Lights vs Population Density

9.3 MP (4320×2160) 6 MB PNG

RGB.svg Red: dense population Purple: dense population + highland Blue: highland, shallow water or ice cap
Yellow: dense population + lights Green: lights Cyan: lights + highland

Earlier this year, Cambridge hosted an e-Luminate festival. Among artistic displays of light were talks highlighting the inaccessibility or unaffordability of home lighting in some developing countries.

I thought it would be useful to show the disparity of light usage on a world map, and decided to encode different dimensions of data in different channels of a colour image, as I had done in this heightmap. I found assorted bitmaps in equirectangular projection at NASA, and rescaled and copied them into different channels in The GIMP:

  • Lights, having finest structure, in the green channel, the channel which appears brightest
  • Population density in the red channel
  • Elevation in the blue channel, the channel which appears darkest and is also reminiscent of water
  • Coastlines and country borders in the alpha channel, which can be turned on and off in some image viewers

Elevation was derived from topography and bathymetry maps. After some experimentation, I found that each could be independently normalised, with low elevation being dark and vice versa, and added together. This resulted in a sharp boundary between land and sea, and also darkness in lowlands where there is more population and lights. Coastlines and country borders were rendered from a modified blank Commons map after setting the fill and stroke of each SVG path to contrasting colours.

To avoid the Chukchi Peninsula in east Russia from being chopped in half, the map was shifted by 10° west so that it wraps around at the 170th meridian west — I later found that the peninsula is still broken up and could have chosen 168.98°W, passing between the Diomede Islands.

The finished piece shows the stark contrast between developed and developing countries: USA and Europe mostly swathed in green, and India, China, Indonesia and Nigeria in red – as others have pointed out, there is a large shift just across the border between the two Koreas. Large cities, especially in Europe and Japan, glow yellow from both high population density and light output.

As expected, most habitation is in the lowlands, but there are lights in the Rockies and Andes (cyan) and dense population around the lofty shores of Lake Victoria and Ethiopia (purple).

And that's my new wallpaper!

Template:Diagonal split header 7 Feb 2016[edit]

An HTML/CSS holy grail that I had been seeking for several years
is the ability to split a table cell diagonally into 2 triangular halves,
so that all the row and column headers can be compactly labelled,
as in this simplified version of the template:Table of phase transitions:
Solid Liquid Gas
Solid Solid-solid transformation Melting Sublimation
Liquid Freezing Boiling/evaporation
Gas Deposition Condensation
Unfortunately, HTML doesn't actually provide such a construct.
Workarounds include using non-rectangular tables or arrows:
Solid Liquid Gas
From Solid Solid-solid transformation Melting Sublimation
Liquid Freezing Boiling/evaporation
Gas Deposition Condensation
  To →
From ↓
Solid Liquid Gas
Solid Solid-solid transformation Melting Sublimation
Liquid Freezing Boiling/evaporation
Gas Deposition Condensation

I had tried using a CSS background image of a diagonal line, but this doesn't scale well with arbitrary text.

Then one day, I suddenly thought of using a background gradient, as I had done for the table in Electromagnetic spectrum. The gradient runs from the bottom-left to the top-right of the cell, and is the default header-cell grey everywhere except from 49% to 51% of the way where it transitions to a darker grey:

background:linear-gradient(to top right,#f2f2f2 49%,#aaa,#f2f2f2 51%)

The darker colour and transition width were chosen so that, for typical cell sizes, it looks like a diagonal line running across the cell, similar in appearance to regular cell borders. The cell is really still one entity, so the spacing of the text labels has to be adjusted to straddle the line. Though it works on Chrome, Internet Explorer and Firefox, two limitations are that the labels easily overrun the "line" if the user is not careful, and that the background of the header cell cannot be easily changed. An alternative I tried is to have a div completely fill the cell and overlay a transparent-dark grey-transparent gradient, but my tests resulted in small gaps between the diagonal line and the cell borders, ruining the effect.

To use the template, create a header cell using ! followed by {{diagonal split header|HEADER-OF-ROW-HEADERS|HEADER-OF-COLUMN-HEADERS}}. Pad the header text with &nbsp; to straddle the line. Adding rowspan or colspan attributes allows it to span multiple rows or columns, as in this example for Chinese New Year:

Dynamic SVG for Wikimedia projects 24 Jan 2016[edit]

Interactive SVG map using CSS and SMIL

So I've submitted an application for a scholarship and talk at Wikimania 2016 to present my ideas for interactive and animated SVG, and meet potential collaborators who might be interested to further develop them. I wrote up the article

Dynamic SVG for Wikimedia projects

See what you think!


SEA Games logo.svg 31 Dec 2015[edit]

The Southeast Asian Games Federation logo
A log-lin vapor pressure chart for various liquids

A year ago, my computer died on me and I thought I had lost a useful little tool which shows coordinates of any point in an SVG file under development to allow the user to quickly trace a shape from a bitmap, even when zoomed in. I discovered that I had saved a copy as a comment in File:Descriptive_geometry_lines.svg and took the opportunity to update it with new things I've learnt about JavaScript:

 <script type="text/ecmascript">
  /// SVG coordinate picker by CMG Lee (Dec 2015)
  /// based on and 610406
  function _r(p, c) { return p.replace(/~/, c); }
  function pickCoords(evt, doSave) {
   var pos = document.getElementsByTagName('svg')[0].createSVGPoint();
   pos.x = evt.pageX; pos.y = evt.pageY;
   var ctm =, ROUND_TO = 10;
   if (ctm) { pos = pos.matrixTransform(ctm); }
   var x = Math.floor(pos.x / ROUND_TO + 0.5) * ROUND_TO,
       y = Math.floor(pos.y / ROUND_TO + 0.5) * ROUND_TO,
       log = [x,y, x,y].reduce(_r, '#~,~ x="~" y="~"');
   if (doSave) { location.href += log; } else { history.replaceState(null, '',
           location.href.substring(0, location.href.lastIndexOf("#")) + log); }
   document.getElementById('cursor'   ).setAttribute('transform',
    [pos.x,pos.y, ROUND_TO].reduce(_r,'translate(~,~) scale(~)'));
    [    x,    y, ROUND_TO].reduce(_r,'translate(~,~) scale(~)'));
   pos = null;
 <path id="crosshair" d="M9999,0H-9999V9999H0V-9999H9999 M0.5,0.5H-0.5V-0.5H0.5
         M1,1H-1V-1H1" fill-rule="evenodd" fill-opacity="0.5" fill="#99cc00"/>
 <circle id="cursor" cx="0" cy="0" r="0.1" fill-opacity="0.5" fill="#669900"/>
 <circle cx="0" cy="0" r="99999" fill-opacity="0" style="cursor:none"
         onmousemove="pickCoords(evt)" onmouseup="pickCoords(evt, true)"/>

Now it not only shows the cursor's coordinates in the address bar in Firefox (sadly I couldn't update the window title by setting window.document.title ) but saves points clicked (actually when the mouse button is released). By changing the ROUND_TO value, one can change the granularity of the values, which is reflected by an overlay — leaving the inner 4 squares moves to a different value, of which centre point is the corner of the larger square of the quadrant that will be moved into.

Eager to try it out, I found a relatively simple logo on commons:Top_200_images_that_should_use_vector_graphics and managed to fix the non-uniformity in the rings in no time. As another test, I vectorised a graph from commons:Top_200_graph_images_that_should_use_vector_graphics to avoid the user having to look up the legend.

Not a bad diversion on Christmas week ୯ ͡°  ͜১͡° ੭

Bahtinov mask principle.svg 4 Jul 2015[edit]

Illustration of the operation of a Bahtinov mask for refracting optics, showing that rotating it by 180° reverses the direction of the pattern.

On a visit to an astronomy observatory, I learnt about the use of a Bahtinov mask to focus a telescope. Invented only 7 years ago by an amateur astronomer, it ingeniously exploits what is normally considered an undesirable artifact: diffraction spikes. As I couldn't find a diagram on the Internet which explains its operation, I decided to draw one while awaiting the airport bus back to Cambridge.

The next time I had some time to kill waiting for another flight to board, I drew the graphic below, which describes the effect of the struts holding up a reflector's secondary mirror on the spikes.

Perhaps I've found a new way to make better use of down time at the airport!

Comparison of diffraction spikes for various strut arrangements of a reflecting telescope

cmglee Cambridge Science Festival 2015 Menger sponge.jpg 13 Apr 2015[edit]

A model of a tetrix viewed through the centre of the Cambridge Level 3 MegaMenger at the 2015 Cambridge Science Festival
Stellar evolution of low-mass (left cycle) and high-mass (right cycle) stars, with examples in italics

I was delighted to see a model of a Menger sponge made of interlocking folded business card at this year's Cambridge Science Festival, so had to take this interesting perspective through its middle. Learning more about the fractal, I added several tidbits to its article, including an image of its curious cross-section through its middle.

Since the festival had been lacking its own article for many years, I also decided to write it up and include some of my more representatitve shots: Cambridge Science Festival

Inspired by a slide in one of the talks, I drew up this graphic of stellar evolution based on an unannotated rendering from NASA.

SVG animation 1 Feb 2015[edit]

Wikipedia, Wikibooks and Wikiversity are excellent in describing things textually and graphically. However, some concepts are better explained interactively, for example, comparing members of a set or investigating the components of a system, but we are currently limited to video clips and GIF animations.

Now that Scalable Vector Graphics (SVG) is becoming well-supported in web browsers, I believe that we should explore ways to enhance articles with dynamic SVG. JavaScript or ECMAScript allows almost limitless ability for interaction and animation.

A major obstacle is that uploads with JavaScript are barred, understandably for security reasons — on a popular site like Wikipedia, cross-site scripting issues are a concern. Fortunately, SVG provides two other techniques for interactivity and animation: SMIL and CSS.

My project explores them, and aims to create demonstrations and tutorials to allow users to create dynamic content. Here are some demonstrators developed so far:

Animated (non-interactive) SVG
SMIL demonstrating change in transformation and CSS attributes
SMIL demonstrating motion along a path and simulation of 3D
SMIL demonstrating morphing of shapes
Interactive SVG
SMIL map showing change through time
SMIL graphic with expandable and collapsible parts
CSS graphic highlighting features of interest on hover and linking to Wikipedia on click


Template:Classical mechanics derived SI units 16 Nov 2014[edit]

At the 24th Cambridge meetup, Charles showed us the use of templates to create navigation boxes, so I decided to create one which I would have found really useful when I studied mechanics:

Linear Angular
→: × m
↓: ÷ s
→: × rad
↓: ÷ s
× kg (linear),
kgm2rad−1 (angular)
→: × m
↓: ÷ s
→: × rad
↓: ÷ s

The left table has translational properties, common symbols and units, while the right table has rotational ones. They are classified according to the SI base units from which they are derived, according to the scheme on the right. My only gripes about the template are what to title it (it's really more a linear-angular analogy) and the wasted whitespace especially on wide browser windows.

Pete (the wub) also showed me File:London_Underground_Overground_DLR_Crossrail_map.svg which highlights London Underground lines on mouse-click, and I've been experimenting with SVG interactivity lately, but that will have to wait for my next update...

territorial disputes in the South China Sea 2 Nov 2014[edit]

Summary of disputes
Area of dispute
The nine-dash line area
Maritime boundary along Vietnamese coast
Maritime boundary north of Natuna Islands
Maritime boundary north of Borneo
Maritime boundary off Palawan and Luzon
Maritime boundary & islands in Luzon Strait
Islands in southern South China Sea
Islands in northern South China Sea
Land in eastern Sabah
Summary of popular destinations

Medieval town
Night life
Close to Athens
Airplane access
Luxury hotels

I attended an educational Cambridge Festival of Ideas talk yesterday and was intrigued by the complexity of the territorial disputes in East and Southeast Asia.

Back home, I found that though Wikipedia's Territorial disputes in the South China Sea article describes the ongoing ones, it was difficult to see which countries were involved in which disputes. I thus decided to add a table to summarise them.

I recycled an idea I'd used on this page on the Greek Islands in Wikivoyage (the table on the right): some column headers are rotated 90° so that the table isn't too wide. I've tested it on Chrome, Firefox, Opera and Internet Explorer. Credit goes to Timwi who was instrumental in helping me developed it.

It's a shame that this technique requires manual tweaking, such as adjusting the width and height of table cells' CSS to avoid the sideways text overflowing its cell. Pity, too, that the 'BLACK FLAG' U+2691 (⚑) and 'CROSSED SWORDS' U+2694 (⚔) Unicode characters don't display correctly on my browsers — either would have made a much more appropriate icon than the tick!

animal hearing frequency range.svg 26 Oct 2014[edit]

Lin-log graph of the hearing ranges of some animals

Once in a while, I hear or read a comment on the exceptional hearing range (in terms of frequency) of bats, dogs and other animals. Inspired by this interesting BBC article, I decided to compare the ranges of animals for which I could find sources, including humans. For comparison, I added C (musical note) of octaves in the range, calculated from A440 (pitch standard). It seems that while bats can hear very high frequencies, their overall range (in terms of octaves) is poorer than ours. Humans aren't actually too bad, but the overall leaders are porpoises and cats, as well as, surprisingly, ferrets and cows — who would've expected that?!

This was also my first attempt at (almost) polyglot code combining SVG and Python (programming language). I had previously appended the generator Perl and Python scripts in my SVG contributions as a comment. I think that including generator code in the SVG avoids its loss and keeps it up-to-date (storing it on, say, its Commons file description page is liable to cause it to become unsynchronised, as I found with file descriptions).

However, where the static SVG contained embedded images such as File:Volcanic eruption map.svg, this increased the file size horribly. To minimise file size, I divided the SVG into two parts:

  1. Static SVG — SVG hardcoded into the Python script, such as CSS and object definitions
  2. Dynamic SVG — SVG generated by Python, such as plotted graphs

Unfortunately, I couldn't make the SVG completely valid Python code as it must start with <?xml which is invalid in Python. My poor-man's solution was to instruct the user to delete the first line. The following is an outline of the SVG file.

The first line ends with an SVG comment to comment out the start of the Python comment plus the note to the user on the second line. When the first line is deleted, the Python script immediately starts with a block comment to comment out all the SVG code:

<?xml version="1.0" encoding="utf-8"?><!--
""" To recover the Python script to generate this SVG, delete the line above -->
<svg version="1.1" xmlns="" xmlns:xlink="" width="100%" height="100%" viewBox="[x] [y] [width] [height]">


 <use xlink:href="#main"/>
<!-- Please retain this and other comments, which contain Python code to generate this SVG. """

The BEGIN_DYNAMIC_SVG and BEGIN_DYNAMIC_SVG sentinel values tell the Python script where to insert the SVG generated dynamically. The top-level group with id="main" is "used", and the SVG is wrapped up with the Python code proper as an SVG comment.

Anything which the Python script wishes to dynamically output is stored in a string array called outs, which is later "joined" into a string. The final line closes the SVG comment to make it valid a SVG file.

# Store output as a group with id="main" in string array "outs"
import re

# Code to compile everything into a .svg file follows...
double_dash = '-' + '-' # because SVG comments cannot have 2 consecutive '-'s
myself      = open(__file__, 'r').read() # the contents of this very file
file_out    = open(re.sub(r'\..*?$', '', __file__) + '.svg', 'w') # *.* -> *.svg
try: # use try/finally so that file is closed even if write fails
 file_out.write('''<?xml version="1.0" encoding="utf-8"?><!%s
%s''' % (double_dash,
                r'\1%s\2' % (''.join(outs).strip('\n')),
                myself))) # replace dynamic SVG block with contents of "outs"
# SVG-Python near-polyglot framework by CMG Lee (Oct 2014) -->

All this seems unnecessarily convoluted; if you can think of a better way, do let me know!

cmglee Cambridge Wikimedia Meetup 23 tour Chesterton Tower.jpg 23 Sep 2014[edit]

Brian Josephson before the tour

I had the honour of meeting 1973 Nobel physics prize winner Emeritus Professor Brian Josephson on Saturday at a walking tour in northeast Cambridge led by Charles Matthews, one of the pioneers of Wikipedia. In conjunction with Wiki Loves Monuments, Charles decided to hold a tour of lesser-known listed buildings in Cambridge, viz. the Cambridge Museum of Technology and Chesterton Tower, before the regular Cambridge meetup.

We also saw the Riverside Bridge, swift tower, Logan Meadow nature reserve and Elizabeth Way Bridge, before returning to CB2 cafe for hot chocolate. Perhaps due to the rain, only 3 people turned up. I suggest that we have more walks, but after meetups, so that attendees may be persuaded to join in!

Malaysia tree diagram.svg 8 Sep 2014[edit]

Redrawn as SVG
Original PNG

User:Wdcf had contributed an insightful graphic giving an overview of the complicated history of Malaysia. Over three years ago, I coloured the background stripes and added a map to show the locations of these political entities.

After learning SVG, I thought that it could be redrawn as in vector graphics and finally set about the task this weekend. I took the opportunity to add recent developments viz. the formation of the Federal Territories. I also enlarged the text so that smaller thumbnails are legible. Balancing font size, letter spacing and fitting everything in its respective column was quite a challenge! Sadly, MediaWiki has trouble rendering centre-aligned tspan elements on the same line, as can be seen in the Fourier series square wave circles animation.gif animations below, so I had to use uppercase for the place names instead of red text (at least they don't look like links to be clicked).

equidistribution theorem.svg 9 Aug 2014[edit]

Illustration of filling the unit interval (horizontal axis) with the first n terms using the equidistribution theorem with 4 common irrational numbers, for n from 0 to 999 (vertical axis). The 113 distinct bands for π are due to the closeness of its value to the rational number 335/113.
(click for detailed view)
Illustration of the filling of the unit interval (x-axis) using the first n terms of the Van der Corput sequence, for n from 0 to 999 (y-axis)

Greetings from Wikimania 2014!

These plots were inspired by a question User:Alex_Great asked me on my talk page: how to automatically generate random colours for graphs etc. After referring him to , I became curious about low-discrepancy sequences, and learnt about the van der Corput sequence and the equidistribution theorem.

As I was curious about what the sequences look like on a number line, using a visualisation technique User:Timwi had used on the irregularity of distributions article, I wrote a Python script to render the graphs. It's my first attempt at using Python to output SVG (I had previously used Perl). The unexpected result was that of π, which had distinct bands, because milü, the rational number 355/113 is very close to its value.

I also discovered a 2-dimensional version named the Halton sequence, but instead of plotting a graph, which User:Jheald had already done, I decided to show the evolution of the sequence:

Illustration of the first 8 points of the 2,3 Halton sequence

comparison optical telescope primary mirrors.svg 7 Jul 2014[edit]

Comparison of nominal sizes of apertures of some notable optical telescopes

I had been reading about the Very Large Telescope, European Extremely Large Telescope and Overwhelmingly Large Telescope (when will they run out of adverbs?!) but couldn't picture how much larger each progressively gets, so I drew this graphic comparing the nominal sizes of the primary mirrors of some large existing and future optical telescopes with various objects like tennis and basketball courts, and the Arecibo radio telescope.

The telescopes are colour-coded by country in which they are located, and whether they currently exist. Their dimensions are as found on the Web, though holes in most mirrors are not shown due to insufficient data. Dotted circles show mirrors with equivalent light-gathering ability.

I'm amazed how enormous the cancelled Overwhelmingly Large Telescope would have been, and conversely, how relatively small the space telescopes (even the JWST). Also note that the hole in the E-ELT is as big as the entire mirror of the largest current telescopes!

SVG-wise, this is my first time using CSS in an original SVG. Quite a neat idea — I must do it more often...

Very nice picture! But I notice that the segmented mirrors show the central aperture, while all the circular mirrors are shown as solid The LSST, for example, has an extremely large central aperture (5116 mm hole in 8360 mm outer diameter), which greatly reduces its collecting area. Likewise, OWL had a very large (33m) central aperture due to the decision to use a flat secondary. Would you like a list of sizes for the various telescopes? (talk) 01:12, 10 July 2014 (UTC)
Thanks for your comment and compliment (sorry I didn't get your name). I would have certainly preferred to show the apertures, though as I noted in the file description, I could show only those for which I could find published data (Wikipedia discourages original research). If you could link me to the missing ones, I'd most appreciate it! cmɢʟeeτaʟκ 17:57, 25 July 2014 (UTC)
Yes, sorry, I saw that after I wrote the above comment and didn't want to delete it two days later. I'll see what I can do. (talk) 15:20, 26 July 2014 (UTC)
A list of central hole sizes for all the depicted solid mirrors is now at File talk:Comparison optical telescope primary mirrors.svg (talk) 09:28, 28 July 2014 (UTC)

UPDATE 23 SEP 2014: User: brought to my attention this very graphic appearing in Gizmodo/Sploid. Cheers!

Titanic casualties.svg 14 Jun 2014[edit]

Breakdown of survivors and victims of the RMS Titanic (click for more detail)
Visualisation of some British currency terms before decimalisation (click for more detail)

I recently decided to explore a sometimes-derided type of statistical chart: Pictograms.

Often considered eye-candy, particularly when used where a bar chart is more appropriate, pictograms are usually relegated to infographics in mass media.

I wondered in which use cases pictograms were superior. One example I could think of was where the reader might want to count the number of icons, instead of looking up a scale. This fits well with small discrete numbers, especially where each icon naturally represents a single unit.

To that end, I drew these pictograms, which I think gives a better picture (mind the pun!) of the subject than any other type of chart. Any thoughts?

Fourier series square wave circles animation.gif 4 May 2014[edit]

This video has an excellent visualisation of the effect of taking progressively more terms in a Fourier series. I wish I had seen it when I was learning about Fourier transforms etc, and so, to help future students, I decided to compare four functions using the same technique:

The animations were made by first creating SVGs of their 60 frames and uploading them to Wikimedia Commons to render them into PNGs, which unfortunately centre-aligned the first tspan (instead of the entire row) of each denominator. They were then downloaded, and sliced and combined into GIF animations with these ImageMagick commands:

convert.exe Fourier_series_square_wave_circle_animation.svg.png -crop 1x60@ +repage +adjoin Fourier_series_square_wave_circle_animation_%%03d.png
convert.exe -delay 10 -loop 0 -page +0+0 ^
 Fourier_series_square_wave_circle_animation_000.png ^
 Fourier_series_square_wave_circle_animation_001.png ^
 Fourier_series_square_wave_circle_animation_002.png ^
 Fourier_series_square_wave_circle_animation_003.png ^
 Fourier_series_square_wave_circle_animation_004.png ^
 Fourier_series_square_wave_circle_animation_005.png ^
 Fourier_series_square_wave_circle_animation_006.png ^
 Fourier_series_square_wave_circle_animation_007.png ^
 Fourier_series_square_wave_circle_animation_008.png ^
 Fourier_series_square_wave_circle_animation_009.png ^
 Fourier_series_square_wave_circle_animation_010.png ^
 Fourier_series_square_wave_circle_animation_011.png ^
 Fourier_series_square_wave_circle_animation_012.png ^
 Fourier_series_square_wave_circle_animation_013.png ^
 Fourier_series_square_wave_circle_animation_014.png ^
 Fourier_series_square_wave_circle_animation_015.png ^
 Fourier_series_square_wave_circle_animation_016.png ^
 Fourier_series_square_wave_circle_animation_017.png ^
 Fourier_series_square_wave_circle_animation_018.png ^
 Fourier_series_square_wave_circle_animation_019.png ^
 Fourier_series_square_wave_circle_animation_020.png ^
 Fourier_series_square_wave_circle_animation_021.png ^
 Fourier_series_square_wave_circle_animation_022.png ^
 Fourier_series_square_wave_circle_animation_023.png ^
 Fourier_series_square_wave_circle_animation_024.png ^
 Fourier_series_square_wave_circle_animation_025.png ^
 Fourier_series_square_wave_circle_animation_026.png ^
 Fourier_series_square_wave_circle_animation_027.png ^
 Fourier_series_square_wave_circle_animation_028.png ^
 Fourier_series_square_wave_circle_animation_029.png ^
 Fourier_series_square_wave_circle_animation_030.png ^
 Fourier_series_square_wave_circle_animation_031.png ^
 Fourier_series_square_wave_circle_animation_032.png ^
 Fourier_series_square_wave_circle_animation_033.png ^
 Fourier_series_square_wave_circle_animation_034.png ^
 Fourier_series_square_wave_circle_animation_035.png ^
 Fourier_series_square_wave_circle_animation_036.png ^
 Fourier_series_square_wave_circle_animation_037.png ^
 Fourier_series_square_wave_circle_animation_038.png ^
 Fourier_series_square_wave_circle_animation_039.png ^
 Fourier_series_square_wave_circle_animation_040.png ^
 Fourier_series_square_wave_circle_animation_041.png ^
 Fourier_series_square_wave_circle_animation_042.png ^
 Fourier_series_square_wave_circle_animation_043.png ^
 Fourier_series_square_wave_circle_animation_044.png ^
 Fourier_series_square_wave_circle_animation_045.png ^
 Fourier_series_square_wave_circle_animation_046.png ^
 Fourier_series_square_wave_circle_animation_047.png ^
 Fourier_series_square_wave_circle_animation_048.png ^
 Fourier_series_square_wave_circle_animation_049.png ^
 Fourier_series_square_wave_circle_animation_050.png ^
 Fourier_series_square_wave_circle_animation_051.png ^
 Fourier_series_square_wave_circle_animation_052.png ^
 Fourier_series_square_wave_circle_animation_053.png ^
 Fourier_series_square_wave_circle_animation_054.png ^
 Fourier_series_square_wave_circle_animation_055.png ^
 Fourier_series_square_wave_circle_animation_056.png ^
 Fourier_series_square_wave_circle_animation_057.png ^
 Fourier_series_square_wave_circle_animation_058.png ^
 Fourier_series_square_wave_circle_animation_059.png ^
 -coalesce -depth 3 -colors 32 -dither FloydSteinberg Fourier_series_square_wave_circle_animation.gif

nucleosynthesis periodic table.svg 31 Mar 2014[edit]

Periodic table showing the origin of elements

As Neil deGrasse Tyson movingly put it [1], an astounding fact is that we are star dust — most of the elements in our bodies (barring some recent radioactive-decay products) were created in stars, by cosmic rays or during the Big Bang itself.

At a talk at the Cambridge Institute of Astronomy, the speaker presented a slide similar to this periodic table showing the origins of the elements. I was surprised to find Wikipedia without an equivalent graphic, so decided to make one!

Simplified Git data flow

The SVG itself is straightforward, the only interesting bit being the Appley skeuomorphic gradient shading I "invented" when making the Git chart on the left:

<linearGradient id="grad_shade" x1="0%" y1="0%" x2="0%" y2="100%">
 <stop offset="10%" stop-color="#ffffff" stop-opacity="0.5"/>
 <stop offset="90%" stop-color="#000000" stop-opacity="0"/>

Applying this to a shape makes its top 10% half-transparent white, and its bottom 10% black but fully transparent. The middle region has a curious gradient which, in my opinion, looks quite attractive when overlaid on a copy of the shape filled with the intended solid colour.

comparison ISS HST orbits globe centered in Cape Verde.svg 28 Feb 2014[edit]

Comparison of ISS and Hubble Space Telescope orbits

I caught Gravity on a flight and marvelled at the amazing special effects and way-above-average accuracy of its portrayal of space travel, particularly how the alarm sounds came back when the hallucination turned up the air pressure and how the Soyuz righted itself during reentry.

Nevertheless, one of the popular gripes is how easy it was to travel between the Hubble, ISS and Tiangong. Of course, one could imagine an alternate universe where they are put into almost identical orbits. When the film first came out, I drew this graphic to show how different the real ISS and Hubble's orbits are. Even if they were at the same height and almost the same longitude/latitude, their orbital inclinations would require huge change in delta-v which I doubt a backpack could provide.

As for the creation of the image, I faced the classic problem of "drawing Saturn with its rings": How to draw a circle with an ellipse both behind and in front of it?

The solution I used was to erase half the ellipse using an SVG clip-mask, and draw it over the globe which already has the full ellipse drawn under it. Because of anti-aliasing, the twice-drawn half is slightly thicker than the other half, but it's not noticeable at normal resolutions. I could have made the ellipse below also a half-ellipse but thought that any gap in the joint would look even worse...

Burj Khalifa floors.svg 14 Jan 2014[edit]

Burj Khalifa with floors colour-coded by function

This dimetric projection of the tallest building in the world is my first SVG animated using CSS3. I had previously used JavaScript/ECMAScript to animate SVG but the Wikimedia uploader rejected it as JavaScript was considered a security risk. Using CSS3 is more limited, allowing noninteractive animation of various attributes, the transform attribute in this case.

Using the technique to simulate 3D rotation below, I rotated each different floor plan, replicated it in the appropriate colour as many times as needed, and scaled the whole group vertically by half. The illusion was reinforced by the sheer number of floors. The only niggles were...

  1. The framerate of the animation: each floor being made of 12 elements (6 for each fill and stroke — combining both resulted in an ugly redundant line at the apex of each lobe) and there being over 200 floors meant that there were over 1200 elements to animate in each frame,
  2. The sharp angles of the base and spire "floors" giving a jagged appearance in silhouette, and
  3. The Mediawiki renderer ignoring the CSS3 — it would have been perfect if it rendered an animated GIF (animated PNGs being less well-supported).
Firefox 26.0 breaks animation

I was so satisfied that I retroactively applied it to the Absolute World Tower model and was even more impressed with the quality of the rendering. As there were fewer elements and the floors were more rounded, both the animation and the silhouette were much smoother. That the Mediawiki renderer ignores CSS3 was used to show the rightmost text label only in a static thumbnail.

Sadly, browser support for CSS3 is still patchy: the test animation of falling snowflakes works in Chrome, but not in Firefox 26.0. At time of writing, noone has suggested a fix or workaround for this bug :-(

Incidentally, it's now been 2 years since I started this behind-the-scene look at my work with Wikipedia/Wikimedia. Does anyone actually look at this? If so, I'd love to hear from you — cheers!

UPDATE 26 JAN 2016: user:RaphaelRBarbosa reminded me that the snowflakes animation now works in Firefox — thanks, Raphael!


coupon collector problem.svg 28 Dec 2013[edit]

Graph of number of coupons, n vs the expected number of tries needed to collect them all, E (T )

I play a game which has characters making various sculptures, each attempt generating 1 of 12 possible sculptures at random. Every time I complete a "collection" of 12, I get a bonus and get to start a new collection.

I wondered how many tries I need, on average, to complete a collection from scratch. A search introduced me to the Coupon collector's problem. Paraphrasing the article,

In probability theory, the coupon collector's problem describes the "collect all coupons and win" contests. ... Given n coupons, how many coupons, E (T ) do you expect you need to draw with replacement before having drawn each coupon at least once?

The answer uses a series which I had read about but never known an application for: the Harmonic series, in the form

Hn = 1/1 + 1/2 + 1/3 + 1/4 + 1/5 + ... + 1/n

Mathematical induction gives E (T ) = n Hn , which I've plotted for n from 1 to 60 (to cover the case of 52 playing cards — I'd have preferred 100, but that makes the smaller bars too hard to see).

I added dotted lines to show the number of tries needed for each coupon. Though this value increases, it increases rather slowly, as E (T ) being linearly proportional to n logn implies.

Back to my game, my 12 sculptures would thus require 38 attempts, on average. In practice, I found that it takes fewer tries, perhaps because it wasn't as random as I'd thought.

By the way, I love the elegant proof that the harmonic series diverges (goes to infinity as n approaches infinity), unlike a geometric series with a ratio between -1 and 1 (not inclusive). I'll consider drawing a proof without words of it one day...

rolling shutter effect.svg 10 Nov 2013[edit]

Simulated effect of a rolling shutter on a spinning disc (Click for animation)
My, what flexible rotors you have!

I used to be baffled by photographs like the one of the helicopter on the left, wondering what on Earth could have caused such distortion on the blades! I recently learnt that it is an effect of the rolling shutter on many digital cameras, especially those on mobile phones.

Unlike film cameras which expose the entire frame at the same time (actually they have rolling blinds too, but the effect is only apparent at very fast shutter speeds), cameras without a mechanical shutter expose and read off a line at a time. This means that by the time the last row is captured, a fast-moving subject will have moved significantly from where it was when the first row was captured.

To illustrate the warping more precisely, I came up with a sequence of 24 frames of a rotating disc and simulated a rolling shutter scanning a row at a time. In each frame, the left half shows the state of the disc and the position of the captured row, while the right half shows the image captured until that point. Sadly, there is a bug in Mediawiki (not in Firefox or Chrome) that renders some artifacts in the top rows; I guess it didn't like my 24 levels of nesting to generate the captured image sequence, the most I've used to date.

Nevertheless, it's now clear to me that the part of the disc moving in the direction of the scan (the right side) is smeared out into large shapes while those moving the other way (the left side) is compressed. The small number of rows cause the jagged appearance; more rows in a real camera lead to smoother curves.

To complete the simulation, I sliced the image into 24 layers in Gimp and stitched them into GIF animation.

I later found this video which explains it even better. I might make a similar animation in the future...

UPDATE 31 MAR 2014: Really proud to see my animation not only used on Vsauce, but host Michael Stevens even built on it to explain that our view of the Andromeda Galaxy suffers from a rolling-shutter effect.

Hipparcos Catalogue equirectangular plot.svg 21 Oct 2013[edit]

File:Hipparcos_Catalogue_equirectangular_plot.svg marked the completion of my trio of charts of objects in the night sky, all of right ascension against declination in an equirectangular grid:

Constellations ecliptic equirectangular plot.svg The first plot is of the 88 modern constellations. Surprisingly, constellations are now defined as regions of the sky rather than groups of stars, so the graphic shows their boundaries rather than the traditional lines-between-stars. I also added the ecliptic and a negative bitmap to show the Milky Way. User:Timwi kindly helped me shade the regions using the four-colour theorem, which help tell adjacent ones apart.
Messier objects equirectangular plot.svg The next graphic is of the fuzzy things catalogued in the 18th century by astronomer Charles Messier. As science progressed, it was discovered that they were very different objects, some being clouds of gas while others being clusters of stars. Yet others were whole galaxies far, far away... (but that was a long time ago!)
Hipparcos Catalogue equirectangular plot.svg The finale is a chart of the 2000 or so brightest stars more recently catalogued by the Hipparcos project. That was an opportunity to revive the lines-between-the-stars, which was obtained from data files of the excellent virtual planetarium, Stellarium. I'm tempted to draw them on the first plot, too, but that will have to wait for another day...

cmglee London FCO Grand Staircase.jpg 28 Sep 2013[edit]

The Grand Staircase of the Foreign and Commonwealth Office

Another year, another Open House London!

This time, I managed to win the ballot for a place on the Arriva Heritage Bus Tour, which used the last classic Routemaster, RM2217 to take us to 5 landmarks:

  1. The Royal Society,
  2. Apothecaries' Hall,
  3. Maughan Library of King's College London,
  4. Middle Temple, and
  5. One Bishops Square.

I also visited the Foreign and Commonwealth Office (FCO), and the Brunel Museum and its recently-reopened entrance shaft of the Thames Tunnel.

Of course, my trusty camera came along, so here's another batch of photos for your viewing pleasure...

million award logo.svg 27 Aug 2013[edit]

Logo for Million Award
Million award logo.svg This user has not won a Million Award but is nevertheless proud to have designed a logo for it!

Today, I got my first Wikimedia request for a logo design (a very old hobby) for the so-called Million Award, which I think is a good idea — much as I enjoy drawing pretty pictures of my interests, I've long thought that my efforts benefit society most when spent on popular topics which are somehow lacking.

The requester, User:Khazar2 had adapted my visualization of 1 million graphic (see below), so I continued in that direction. Befitting the theme of "million", the middle cube has 1 millionth the virtual volume of the large cube, while the CMYK colours evoke the idea of "articles" in the traditional printing sense. "1M" is hopefully language-independent and the star is a common motif for awards such as the Barnstar.

The skewed text is also my first use of SVG matrix affine transformation, and I was delighted that the bottom-right edge of the star (simply Unicode character U+272A) lined up with the bottom-right side of the cube. It was also my first attempt to specify a particular font, in this case, one with tapering (not slab) serifs so that the "1" and "M" look more elegant and reminiscent of printed text.

All in all, it was a fun little project for the day!

cmglee Horniman London skyline.jpg 25 Jul 2013[edit]

I revisited the Horniman Museum, one of my favourite museums in London because of their approach to education by comparing artifacts from different cultures. I also love their skeleton-skin preparations, which show just the skeleton in half of a stuffed animal, as it shows how the bones (and consequently, muscles and internal organs) fit in the body. These are some examples:

I also reshot the London skyline from Forest Hill to show a completed Shard. Sadly, the summer air was quite hazy, but some image-processing helped. In the style of Horniman, see how the city has changed in 1½ years!

London skyline from Forest Hill in July 2013
London skyline from Forest Hill in Jan 2012

More photos are at Commons

Template:Alcohol metabolism formulae 25 Jun 2013[edit]

A CrossCountry train passing behind the new building of the Laboratory of Molecular Biology in the Cambridge Biomedical Campus, view from the southwest on the Cambridgeshire Guided Busway bridge on 22 June 2013.

The Laboratory of Molecular Biology opened its doors to the public on Sat, 22 June, so my friends and I attended some talks, exhibitions and a tour of their new home. (I also used the opportunity to take some photos!)

One of the speakers talked about a mutation causing possibly the greatest cancer risk in the world: ALDH2*2

With China rapidly adopting a drinking culture, about 25% of the East Asians (perhaps 300 million people) face increased risk of esophageal cancer. If I understood correctly, the risk of an affected person drinking, compared to an unaffected one, is greater than that of lung cancer from smoking (of course, the quantity matters...)

I couldn't find a graphic of the breakdown of alcohol in the alcohol flush reaction page, so tried to create one. I could have done it in SVG, but thought it was an interesting experiment to do it in the LATEX-like maths "equation editor". I had to tweak the spacing quite a bit and couldn't draw diagonal double bonds, but managed the following:

Metabolism of alcohol (ethanol) to acetaldehyde (ethanal) and then acetic acid (ethanoic acid)

Whatjsay? :-)

perpetual calendar cycle.svg 14 Jun 2013[edit]

Table showing the 400-year cycle of the 14 possible calendars in the Gregorian calendar, arranged in columns to show a semi-regular 28-year pattern. The days of week denote the day of week that January 1st of the numbered year falls on. "+" and darkened cells indicate leap years.

Timwi's image at got me interested in calendars again — I must be getting old!

His most amazing find was that the 400-year cycle of the Gregorian calendar (the one in everyday use today) has a multiple of 7 days, meaning that the cycle repeats exactly every 400 years. But interesting and colourful as his graphic was, it didn't help with my understanding of how the 14 possible calendars actually change, so I decided to draw my own...

While on holiday, I wrote a quick program to output into a simple text file as one line separated by spaces:

  • The day-of-week of January 1st of each year from 2001 to 2400 (unlike Timwi's choice of 2000 to 2399, I started with 2001 so that the rare 400-year leap on 29 Feb 2000 after 1 Jan wouldn't disrupt the pattern), and
  • Whether the year was a leap year.

Changing the width of my text editor window let me see in how many years the pattern more-or-less repeated.

The answer was 28, not surprisingly the lowest common multiple of 7 (days in a week) and 4 (almost every 4th year is leap), the pattern disrupted only by the non-leap years 2100, 2200 and 2300. 28 columns would, however, be too wide, so I decided to go down then across, instead of across then down.

The pattern appeared more regular now, but the non-leap century years still played havoc, until I got the idea of skipping the necessary number of cells until the pattern resumed. There are actually 2 possible places each non-leap century year can go. I chose the later one so that 2300 doesn't end up at the top of its column.

This resulted in the graphic on the right. It's now much clearer to me how the calendars cycle — one day forward each non-leap year, and two days forward each leap year, except for the non-leap century years which leave large gaps.

P.S. I redrew the graphic in Wikitext below, if anyone wants an HTML copy.

Day of week of Jan 1 Leap year Gregorian calendar for years from 2001 to 2400
Monday   2001 2029 2057 2085 2125 2153 2181 2221 2249 2277 2317 2345 2373
Tuesday   2002 2030 2058 2086 2126 2154 2182 2222 2250 2278 2318 2346 2374
Wednesday   2003 2031 2059 2087 2127 2155 2183 2200 2223 2251 2279 2319 2347 2375
Thursday 2004 2032 2060 2088 2128 2156 2184 2224 2252 2280 2320 2348 2376
Saturday   2005 2033 2061 2089 2101 2129 2157 2185 2225 2253 2281 2321 2349 2377
Sunday   2006 2034 2062 2090 2102 2130 2158 2186 2226 2254 2282 2322 2350 2378
Monday   2007 2035 2063 2091 2103 2131 2159 2187 2227 2255 2283 2300 2323 2351 2379
Tuesday 2008 2036 2064 2092 2104 2132 2160 2188 2228 2256 2284 2324 2352 2380
Thursday   2009 2037 2065 2093 2105 2133 2161 2189 2201 2229 2257 2285 2325 2353 2381
Friday   2010 2038 2066 2094 2106 2134 2162 2190 2202 2230 2258 2286 2326 2354 2382
Saturday   2011 2039 2067 2095 2107 2135 2163 2191 2203 2231 2259 2287 2327 2355 2383
Sunday 2012 2040 2068 2096 2108 2136 2164 2192 2204 2232 2260 2288 2328 2356 2384
Tuesday   2013 2041 2069 2097 2109 2137 2165 2193 2205 2233 2261 2289 2301 2329 2357 2385
Wednesday   2014 2042 2070 2098 2110 2138 2166 2194 2206 2234 2262 2290 2302 2330 2358 2386
Thursday   2015 2043 2071 2099 2111 2139 2167 2195 2207 2235 2263 2291 2303 2331 2359 2387
Friday 2016 2044 2072 2112 2140 2168 2196 2208 2236 2264 2292 2304 2332 2360 2388
Sunday   2017 2045 2073 2113 2141 2169 2197 2209 2237 2265 2293 2305 2333 2361 2389
Monday   2018 2046 2074 2114 2142 2170 2198 2210 2238 2266 2294 2306 2334 2362 2390
Tuesday   2019 2047 2075 2115 2143 2171 2199 2211 2239 2267 2295 2307 2335 2363 2391
Wednesday 2020 2048 2076 2116 2144 2172 2212 2240 2268 2296 2308 2336 2364 2392
Friday   2021 2049 2077 2117 2145 2173 2213 2241 2269 2297 2309 2337 2365 2393
Saturday   2022 2050 2078 2118 2146 2174 2214 2242 2270 2298 2310 2338 2366 2394
Sunday   2023 2051 2079 2119 2147 2175 2215 2243 2271 2299 2311 2339 2367 2395
Monday 2024 2052 2080 2120 2148 2176 2216 2244 2272 2312 2340 2368 2396
Wednesday   2025 2053 2081 2121 2149 2177 2217 2245 2273 2313 2341 2369 2397
Thursday   2026 2054 2082 2122 2150 2178 2218 2246 2274 2314 2342 2370 2398
Friday   2027 2055 2083 2100 2123 2151 2179 2219 2247 2275 2315 2343 2371 2399
Saturday 2028 2056 2084 2124 2152 2180 2220 2248 2276 2316 2344 2372 2400

Dewan Rakyat 2013 Equal Area.svg 9 May 2013[edit]

Results of the Malaysian Dewan Rakyat based on the 2013 general election, showing parliamentary constituencies represented by equal-area hexagons with approximate geographic locations.

Once again, I'm honoured to see this cartogram of the parliamentary constituency results of the Malaysian general election, 2013 appearing on the main page of the Chinese Wikipedia.

To be honest, most of the credit should go to User:Hytar who made the cartogram for the Malaysian general election, 2008. I just fixed a few minor cosmetic issues, tidied up the SVG code and updated the data to reflect the 2013 results. Due to his or her judicious use of CSS, the task was pretty straightforward.

I must give kudos to Hytar, too, for being so supportive and polite — a model Wikipedian if there was one!

The next job is to update File:Dewan_Undangan_Negeri_2008_Equal_Area.svg with this year's results. Hytar referred to some spreadsheets going around. Obtaining one would make updating the 505 hexagons much easier.

Template:Blakey_65moll.jpg_K/T_impact_site 9 Apr 2013[edit]

{{Blakey_65moll.jpg_K/T_impact_site}} Today, I discovered a new technique to annotate images, using the Annotated image template. I had previously converted bitmaps to Base64 and embedded them in SVG files. With the template, I could add labels to any image directly in the Wiki markup.

For example, the code below produces the graphic on the right. The template also let me crop the sides of the Mollweide projection, to make the graphic fit inside text and yet allow detail on the map to be clearly seen. To isolate the label from the bitmap, I repeated it twice in black with slight offsets to simulate an outline — an alternative was to use a CSS shadow but CSS support for it is dicey.

I was surprised to find plenty of maps of the Chicxulub crater, the location of the Cretaceous–Paleogene_extinction_event, but no world maps showing this location on Wikimedia. Of course, the world looked a bit different 65 million years ago!

{{ Annotated image
 | image   =
 | caption = K/T impact site on a contemporary world map (65 mya)<ref>[ K/T extinction]</ref>
 | width   = 300 | height = 190 | image-width = 400 | image-left = -45 | image-top = -3
 | annotations = {{Annotation|74|64|* Impact site|font-weight=bold|color=black}}
                 {{Annotation|76|66|* Impact site|font-weight=bold|color=black}}
                 {{Annotation|75|65|* Impact site|font-weight=bold|color=white}}

UPDATE 26 JAN 2016: It seems the maps have been removed due to copyright violation; I'll see if I can redraw them in the future.

chain hoist.svg 10 Mar 2013[edit]

Example of a differential pulley

Ever since I learnt about the differential pulley, I have been intrigued by the mechanism. Although it superficially resembles a block and tackle, at least two of its pulleys are sprockets, which engage with a chain, hence its other name, "chain hoist". Additionally, these two pulleys are joined together, and the ratio of their diameters (or radii) determines its mechanical advantage.

Drawing a graphic of it was an interesting exercise in SVG-hacking:

First, the teeth of the sprockets were simply the dotted stroke (outline) of the circle representing the pulley, set to the same colour as its fill, and with the duty cycle, phase and width adjusted to fit.

Next, I needed a path that looked like a chain, and found a way to do it with three dashed lines of different duty cycles, phases, widths and colours, one of them in the background colour to "erase" the hole in the links. This looked fine until I overlaid the chain on the pulleys, making the background colour incorrectly appear over the pulleys. If only clip paths had an inverse function which let one ignore pixels inside the path, rather than outside!

I left the problem for over a year before discovering masks. I could now draw the chain without the holes and overlay the whole thing on any background. The experiment on the left, a world map with borders replaced with chains, proved that this technique was feasible as long as turns were not too sharp. The hack was successfully applied to the differential pulley graphic (right), though I later found that Google Chrome had trouble with my mask. Oh well...

commons:Category:Historical images of Penang 3 Mar 2013[edit]

Mount Erskine in 1824

I met User:Charles Matthews, User:Deryck Chan and User:Magnus Manske at the Cambridge Wikimedia Meetup Number 17 again yesterday and they convinced me that artwork from 1850 and before is globally considered public domain. I've therefore completed upload of 50 historical paintings and drawings of Penang housed in the Penang Museum and Art Gallery. The works show a much quieter island compared to the bustling city of today, but likely presents a European colonist's romanticised perspective. Nevertheless, it serves as a good historical record of the island.

There are 11 other pieces dating from after 1850, and 2 with dates unstated. Perhaps someone can upload them as they come out of copyright in the coming years...

sexagenary cycle years.svg 9 Feb 2013[edit]

Relationship between the current Sexagenary cycle and Common Era years
An optical illusion similar to Rotating Snakes by Kitaoka Akiyoshi

As several Asian cultures celebrate the lunar new year tomorrow, I made a couple of somewhat related illustrations.

I have sometimes wondered how the Chinese decide that a particular year is the Year of the Water Snake (for example). I knew that the 12 animals of the Chinese zodiac go in sequence but the element was a mystery. The Sexagenary cycle article makes it clear: The element cycles in the order wood, fire, earth, metal and water, changing every two years (the first year being yang and the second, yin). With that in mind, I drew the graphic on the right, adding in the numeric years. As it wasn't immediately obvious, I also added a tortuous line to show the progression — hope it's not too messy... I initially laid the years out in a table before realising that they wrap around in both X and Y axes: The surface of a toroid! Alas, drawing a toroidal surface isn't that easy, so the best I could manage was to make an annulus (mathematics) and wrapping the element radially.

The graphic on the left is based on a clever optical illusion which tricks the brain into seeing movement where there is none. The original, named Rotating Snakes, by psychologist Kitaoka Akiyoshi is an example of peripheral drift illusion which works on the viewer's peripheral vision. For best effect, one should use as large an image as possible, such as this one. An even more stunning illusion is found on the cover of the indie album Merriweather Post Pavilion.

Happy lunar new year to all who celebrate it!

comparison of pyramids.svg 31 Jan 2013[edit]

English version with collapsed pyramid at Meidum
German version with Bosnian "Pyramid of the Sun"

I've just updated my comparison of some pyramidal buildings, adding the collapsed pyramid at Meidum to replace the Bosnian "Pyramid of the Sun" which was forcefully removed due to the misconception that I had tried to promote the Bosnian hill as a building. That was never my intention — I thought that including it in the graphic only served to highlight how ridiculously enormous it was compared to other pyramids. Nevertheless, an editor removed it, leaving an unsightly gap in the labels.

Meanwhile, I had been thinking about including the Meidum pyramid but felt that my SVG skills were inadequate at the time. Now that I had more experience with Bézier curves and a gap to fill, I traced this illustration and added it to the graphic, renumbering the pyramids in the process (I had checked that no pages that used it referred to the numbers).

Guess you can't win every battle, but you can make the most of the outcome!

UPDATE 16 NOV 2014: I've hyperlinked the label and profile of each "pyramid" so that hovering highlights them and clicking opens its Wikipedia article.

Template:Distance from Sun using EasyTimeline 2 Jan 2013[edit]

Today marks the Perihelion of the Earth, the moment in a year that the Earth is closest to the Sun, paradoxically near the coldest time of year in the Northern Hemisphere!

A while back, I discovered the ability to render graphics in an article by just writing some code, instead of uploading an image (such as SVG) and transcluding it: EasyTimeline.

Originally designed to create timelines, it can be repurposed to draw bar charts in general. My first use of it is the chart below graphically illustrating the range of distances some bodies of the solar system can be from the Sun. It can be seen that the inner planets have almost circular orbits whereas the outer dwarf planets and Halley's Comet have very eccentric orbits.

Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Astronomical unit Halley's Comet Sun Eris (dwarf planet) Makemake (dwarf planet) Haumea (dwarf planet) Pluto Ceres (dwarf planet) Neptune Uranus Saturn Jupiter Mars Earth Venus Mercury (planet) Astronomical unit Astronomical unit Dwarf planet Dwarf planet Comet Planet

Distances of selected bodies of the Solar System from the Sun. The left and right edges of each bar correspond to the perihelion and aphelion of the body, respectively, hence long bars denote high orbital eccentricity. The radius of the Sun is 0.7 million km, and the radius of Jupiter (the largest planet) is 0.07 million km, both too small to resolve on this image.


Mesoamerican Long Count Calendar visualization.svg 21 Dec 2012[edit]

Visualisation of the Mesoamerican Long Count calendar
Visualisation of powers of ten from 1 to 1 million

I'm pleasantly suprised at the lack of media coverage of the 2012 phenomenon, considering the frenzy when the 2012 movie was released — it's certainly nothing like what accompanied the Y2K scare.

Nevertheless, I was curious to find out the relevance of today's date in the Mayan calendar and read about their units of time. 21st December 2012 marks the end of the 13th (written 12 in their dates as they start counting at zero), and the start of the 14th b'ak'tun, a unit of time 144000 days long (around 400 years).

Using the idea I developed to visualise largish numbers at the same scale (one million in the illustration on the right, and one billion in this one for high-resolution screens), I drew similar blocks representing the Mesoamerican time units up to a piktun (around 8000 years), in principle representing almost 3 million small cubes in the largest block! I added colours and a scale showing equivalent dates on the Western calendar. To reduce the width of the image, I flipped my trimetric projection horizontally.

From this image, I can see that while the date is interesting, it's not exceptional as it occurs once every 394 years. Moreover, nothing major happened at the last few changeovers: 1618, 1224 and 830.

On that note, I'll just wish everyone, "Happy new b'ak'tun!"

Heathrow Airport map with third runway.svg 17 Dec 2012[edit]

Map of London Heathrow Airport showing proposed extension and third runway
Fukushima I and II Nuclear Accidents Overview Map as of 15 March 2011

The map on the left is my first attempt to combine an SVG export of OpenStreetMap data with annotation to show the terminals of London Heathrow Airport and its proposed extension, although I had previously embedded bitmaps of OpenStreetMap maps in the illustration of the Fukushima Daiichi nuclear disaster (right).

Though I have used Heathrow Airport for many years, I had no idea about its layout, other than the Central Bus Station and Underground station serving Terminals 1 and 3 (and formerly Terminal 2) while Terminal 4 had its own stations. The map now makes it clear how the buildings are laid out with respect to the runways and one another. I took the opportunity to illustrate the controversial third runway and extension, which has garnered plenty of press coverage in the UK.

Technical-wise, I found adding annotations straightforward but modifying existing SVG objects (even just changing their colours) quite difficult, particularly since there were so many objects that I had to hunt down the target object by colour, which was specified as, for instance, rgb(80%,60%,100%) instead of the more common #cc99ff format. I also thought the SVG was extremely inefficient in terms of file size. I would have, for example, grouped all objects of the same style and applied the styling to the group, rather than style each object individually. If I ever need to annotate an exported OpenStreetMap SVG again, I'll consider using an editor like Inkscape.

parallax barrier vs lenticular screen.svg 30 Nov 2012[edit]

Comparison of parallax-barrier and lenticular autostereoscopic displays
Principle of operation of a pound lock

Another thing that we talked about at the CUWPS gathering was three-dimensional (3D) displays, such as those in the Nintendo 3DS. I showed the group the drawing on the left to illustrate the principle, and the subject changed to drawing SVG.

We looked at the source code and discussed use of groups, transforms and defined blocks to reduce the effort of producing one, while also saving on file size.

This drawing is one of my forays into simulating a 3D scene in a 2D medium, ironic considering the subject of the drawing is technology to simulate 3D with 2D!

The projection used is the approximate isometric projection used in isometric graphics in video games and pixel art, actually a dimetric projection. It simplifies mapping between 3D and 2D coordinates in this case, so I could work out the coordinates by hand, but traditionally, it allowed pixels to fall neatly into the raster graphics matrix without needing antialiasing. It looks slightly more aesthetic than true isometric to me, too, as more of the sides is visible than the top, just like how I would sketch a 3D object.

Another drawing done with the same projection is the sequence on the right is showing how some canal locks work.

Penang Harbour HMS Magpie 1884 Penang map.jpg 26 Nov 2012[edit]

Detail of an 1884 map showing Penang Island and part of Province Wellesley
1799 map of early George Town, Penang

I have always been interested in the small island of Penang and was over the moon when I found this scan (left) of an ancient map dating from 1799. Though it was only 13 years since Penang's founding, the map had landmarks such as Fort Cornwallis, Penang Road and the grid of streets near Beach Street. It was fascinating that while land to the southeast had been reclaimed to build Weld Quay, land to the north had been lost to the sea, which may explain why the Esplanade was built.

It then occurred to me that Penang being a British colony meant that there must be extensive historical records of early Penang within a stone's throw at the Cambridge University Library. I found that very book from which the 1799 map came, and also a more recent map in the Map Room (right) — how Penang had grown in 85 years!

I spoke of my experience with the Cambridge University Wikipedia Society at their gathering yesterday and discussed with President Deryck about expanding GLAM to include the University Library archives. I know I'd certainly be interested in their ancient maps ୯ ͡°  ͜১͡° ੭

Cambridge University colleges timeline.svg 29 Oct 2012[edit]

Timeline of the colleges of the University of Cambridge compared with some events in British history.

I attended the 16th Cambridge Wikimedia meetup and inaugural Cambridge University Wikipedia Society meeting yesterday and had the pleasure of meeting several enthusiastic Wikimedians/Wikipedians and learning about the projects and ideas of User:Charles Matthews and User:Magnus Manske. I'm looking forward to the planned collaboration with the University of Cambridge and local museums.

Speaking of Cambridge University, on the right is a chart I made a while ago to illustrate the complex history of the colleges of the University of Cambridge. It shows the years of formation of the colleges, when they changed their names, and most importantly, when they started admitting women (or men, in the case of Girton College and Hughes Hall). Obtaining all this data from disparate sources was quite a challenge, and I'm still missing one crucial date:

When St Edmund's College became mixed.

Repeated emails to various parties in the college were unanswered, so I've shaded the transition to the range of years this could have happened. Anyone got any idea?

UPDATE 28 NOV 2012: User:Deryck Chan managed to get a statement from St Edmund's Emeritus Fellow Dr Michael A Hoskin that he remembered it as 1969. Pending confirmation of historical records, I've tentatively indicated 1969 as the year St Edmund's became mixed.

UPDATE 17 DEC 2012: Thanks to the tireless effort by the St Edmund's College Combination Room Women’s Officer, Miss Maryam Farooq, the year women could be admitted to St Edmund's College, as postgrads, has been established to be 1965.

Container City massing model.svg 28 Sep 2012[edit]

Container City II.
Illustration of the structure of Container City showing how the containers are stacked.
Interior of Studio 12 in Container City I.

Last weekend, hundreds of establishments in London welcomed visitors for Open House London, and so I took the opportunity to visit a pioneering work of shipping container architecture at Trinity Buoy Wharf: Container City.

The structure consists of two buildings named Container City I and Container City II, made out of 52 old 40-foot shipping containers. Even the lift shaft and stairwell are containers standing on end. I really like the balconies made from the containers' open doors, and the bright colours.

Unlike the exterior, its interior looks pretty much like ordinary flats except for the portholes and occasional glimpses of the exterior, and tall people may find the low ceiling claustrophobic! Being an open house, a studio was open to the public, though I cynically think it's just a ploy to get people to rent it :-)

To better understand its structure, I reused my patented (not!) trimetric projection to depict the containers, one layer at a time. The SVG code was wholly done in Notepad++, believe it or not, though the hardest part was finding out what colour each container was painted!

using watch and sun as compass.svg 24 Sep 2012[edit]

A method to identify north and south directions using the sun and a 12-hour analogue clock or watch set to the local time, 10:10 a.m. in this example.

Continuing the theme of clock faces, I recently found out about a method scouts and similar groups learn to locate north and south using the sun and an analogue clock or watch:

As the sun appears to complete one rotation around the sky in 24 hours while the hour hand of a 12-hour clock takes 12 hours, bisecting the angle lets one identify the cardinal directions.

Unfortunately, modern timekeeping complicates matters with time zones and daylight saving. Most countries (except perhaps China and Kiribati) have time zones roughly corresponding to the local time. Daylight saving normally move clocks forward by an hour.

Using this information, the more detailed description at Cardinal_direction#Watch_face and the infamous marketing time, I illustrated these four scenarios:

  1. In the northern hemisphere, with daylight saving.
  2. In the northern hemisphere, without daylight saving.
  3. In the southern hemisphere, with daylight saving.
  4. In the southern hemisphere, without daylight saving.

The SVG itself is rather straightforward, the only difficulty being calculating the rotational transforms. What I'm a little proud of, though, is the sun rendered with 8 quadratic Bezier curves and a radial gradient!

Template:comparison four face clocks.jpg 25 Aug 2012[edit]

Comparison of some notable four-face clocks at the same scale.
Top-left: Metropolitan Life Insurance Company Tower
Bottom-left: Allen-Bradley Clock Tower (previous record holder)
Middle: Abraj Al Bait
Top-right: Palace of Westminster clock tower
Bottom-right: Kremlin Clock
The Abraj Al-Bait towers

I had read about the Abraj Al-Bait Towers (left) and was curious to know how large its new-world-record clock actually was. The statistics, such as its 46-metre-diameter clock faces don't really do it justice, so I decided to create a photomontage of notable four-face clocks and clock towers, all at the same scale.

I chose the Allan-Bradley (the previous record-holder), Big Ben clock tower and Met Life Tower clocks, and later added the Kremlin one, as I felt that they gave a fair representation of famous large four-face clocks.

Orthorectified negative (top) and positive (bottom) representations of the picture, partially obscured.

Next was sourcing photographs of these clocks. I used the crop tool in Photoshop to orthorectify them, so that they appear like they were taken head-on. This process was simplified by the fact that the faces were nearly two-dimensional and I could assume that they were circular. I had had recent practice orthorectifying images, such as one of The Great Picture on the left.

I then had to find a way to get all the images to the same scale. Using the more common "diameter of dial" was troublesome, as the extent of the dial varied from clock to clock. Instead, I used the published lengths of their minute hands as it is easy to measure and one of the longest measurements specified, to reduce uncertainties.

Composite the montage, write a caption, create a template to present it uniformly, and that was basically it...

Now one can see how enormous the new clock is; the entire Big Ben clock tower (to be named the Elizabeth Tower for the British Queen's 60th year at the throne) could fit in the Abraj Al-Bait spire, and if the new clock had numbers on the dial (it has ticks), each number would be almost as big as one Big Ben clock!

Turning Torso structure.svg 19 Aug 2012[edit]

Simplified plan and massing model of Tower 1 of Absolute World residential condominium in Ontario.
Turning Torso by night.
Illustration of the general structure of the Turning Torso by Santiago Calatrava. (1) shows a typical floor plan, where the grey circle denotes the core and blue shapes denote the steel framework. (2) shows the way the nine segments fit around the core, and (3) is a dimetric projection of the tower.

I met an architect who is a fan of Santiago Calatrava and we started talking about his Turning Torso. I found it difficult to visualise its shape, especially because most photograph show only the view taken from the ground (far left).

A little Googling later uncovered plans at and .

With a trick I learnt from Adobe Flash to simulate 3D shapes with 2D ones (rotate then scale horizonatally and vertically by different amounts, and rotate again if needed), I drew this illustration (near left). I can now see the building's shape much more clearly! By the way, guess what influenced the choice of colours...

UPDATE 10 MARCH 2013: The idea was reused for the illustration of Absolute World on the right.

comparison convolution correlation.svg 10 Jul 2012[edit]

One corner of a squircle (blue) compared with that of a rounded rectangle (red)
One corner of a squircle (blue) compared with that of a rounded rectangle (red)
Visual comparison of convolution, cross-correlation and autocorrelation.

I had wondered how to draw the intersection of two SVG shapes, such as the purple areas in the image on the right. One simple method is to set the opacity of the upper (drawn later) shape between 0 and 1, to make it semitransparent. However, this also causes the non-overlapping area to be semi-transparent.

It occurred to me that I could draw that shape twice. Calling the shapes A and B, we can draw A and then B fully opaque, and finally draw a semitransparent version of A on top. The opaque A ensures that the non-overlapping areas remain opaque. In theory, this is slightly different from a single opaque A due to anti-aliasing at its edges, but I haven't found any problems in practice.

The same idea was used in the graphic comparing convolution and correlation functions on two signals. This image additionally uses a more advanced technique with clip paths to shade the common areas under the graphs, but a description of it will have to wait until another day...

Template:CiviltàValleIndoMappa-en 28 Jun 2012[edit]

Location of Cmglee in the Indus Valley and extent of Indus Valley Civilization (green).

I've just discovered a way to embed the name of the current page in a transcluded template, using the {{PAGENAME}} system variable.

For example, the caption of the image on the right should read, "Location of Cmglee in the Indus Valley and extent of Indus Valley Civilization (green)." (unless someone has changed the template!)

Of course, that doesn't make much sense! That's because the name of this page is Cmglee (its full name being User:Cmglee).

When the CiviltàValleIndoMappa-en template is used in the articles for the cities labelled, such as Mohenjo-daro and Harappa, it does its magic. For instance, in the Mohenjo-daro article, the caption reads, "Location of Mohenjo-daro in the Indus Valley and extent of Indus Valley Civilization (green)."

This allows the same image with an appropriate caption to be placed in multiple articles without having to copy and paste the same wikitext and then amend the city name in all of these pages. Moreover, updating the template instantly updates all pages using it. Nice!

2012 Summer Olympics torch relay map.svg 14 Jun 2012[edit]

Simplified route of the 2012 Summer Olympics torch relay.
Illustration of use of cubic Bezier splines to smooth a polyline, The different colours indicate different curviness values (black lowest and red highest). Triangles and squares indicate control points for the Bezier spline of the corresponding colour.

One of the frustrating limitations of SVG is the lack of a feature to automatically draw a smooth curve through a set of points. Sure, one can draw Bézier curves and elliptical arcs between pairs of points, but the coordinates of control points or arc radii must be specified.

Based on an algorithm at Stack Overflow, I developed a Perl function to automatically calculate the control points of cubic Bezier splines to smooth a polyline.

Each point (except the end points) can take up to two curviness values which determine how curvy or straight the curve on each side of the point is. Zero curviness results in a simple polyline. Applied to individual points, this can be used to create sharp corners. Negative curviness is used to create loops. The test curves on the left illustrate its use.

One of the first applications of this function (adapted to gradually change the colour of the curve along its length) is the creation of the simplified torch relay route map to the right. The actual route is far more tortuous than this. To make it easy to trace at this scale, I just wanted the curve to smoothly pass through all the marked places. To make crossings clearer, the curviness of certain points were tweaked until visually acceptable.

Another application, which I shall use more often as the need arises, is the drawing of the smooth curve of many mathematical functions. Currently, I have to generate hundreds of points to get a reasonably smooth curve. With this function, far fewer points will be required.

cmglee ArcelorMittal Orbit.jpg 7 Jun 2012[edit]

ArcelorMittal Orbit viewed from Stratford High Street (A118).

I was pleasantly surprised and honoured to see my photograph of ArcelorMittal Orbit appearing on the Wikipedia main page on 17 May.

It was almost an accident that I shot this photo. On that fateful day, I had taken a coach from Cambridge to London to catch a London Prepares test event at the Olympic Park. I had expected to alight at Stratford, as I had done several times before. I was aghast that National Express had changed the 010 coach schedule to bypass Stratford, forcing me to get off at Bow and walk back to Stratford.

I remembered a great view of the Olympic Stadium and Orbit visible from the flyover above the A118 roundabout — I had uploaded a much lower-quality version of the same scene from a coach to London a few months before, as can be seen from the file's history. So, I decided to climb it.

Though there were no signs forbidding pedestrian entry, the overpass was mighty hairy, with nary a footpath to avoid the fast-moving traffic. Nevertheless, I was rewarded with this magnificent view — if you could call the Orbit "magnificent" ;-)

And that's the story behind the picture... Keep tuning in for more behind-the-scenes!

comparison gender life expectancy CIA factbook.svg 26 May 2012[edit]

Comparison of male and female life expectancy at birth for countries and territories as defined in the 2011 CIA Factbook, with selected bubbles labelled. The dotted line corresponds to equal female and male life expectancy. The apparent 3D volumes of the bubbles are linearly proportional to their population, i.e. their radii are linearly proportional to the cube root of the population.
Earthquakes of moment magnitude 8.0 and greater since 1900. The apparent 3D volumes of the bubbles are linearly proportional to their respective fatalities.

This BBC News article reminded me of the graphic on the right, which I made to compare the life expectancies (at birth) of males and females in different countries. As is widely known, women generally outlive men, particularly in the Eastern European countries labelled. The BBC article discusses possible reasons for this. There is, however, a reversal of this trend in sub-Saharan Africa. It is even more extreme in Montserrat, where (if the figures are to be believed), men live a whole 3¾ years longer than women, on average.

I initially found it a challenge to depict the size of the countries' populations, ranging from about 5000 to almost 7 billion. I could have used a logarithmic scale, but decided to imagine the markers as 3-dimensional "bubbles" with volumes proportional to their population. That let me cover the 6 orders of magnitude with radii covering 2 orders of magnitude, so that if, for example, the smallest bubble were 1 pixel wide, the largest would be a manageable 100 pixels wide.

Additionally, the transparent bubbles let multiple bubbles overlap and still be visible. A small dot at their centres allow the reader to read off the life expectancies in years.

The same idea was reused for the graphic on the left, showing fatalities in earthquakes of magnitude 8 and greater since 1900.

comparison of highest mountains.svg 9 Apr 2012[edit]

Comparison of the heights of the Eight-thousanders, Seven Summits and Seven Second Summits
Mark I
Mark I
Mark II
Mark II
Diagram comparing the height of Burj Khalifa to other buildings and structures

A recent exchange I had with RacerX11 reminded me of the benefits of constructive cooperative editing.

I (and presumably some others) have wondered how the famous tall mountains around the world, such as Aconcagua and Kilimanjaro compare with the giants of the Himalayas. Of course, I could look up their heights to the metre in various tables, but there's nothing like a scale image to provide instant understanding, as in the comparison of skyscrapers on the left.

So I decided the compare the Eight-thousanders, Seven Summits and Seven Second Summits in the top-right graphic. Sadly, I had to use triangles as I did not have the actual elevation profiles of the mountains.
Neither RacerX11 nor I were satisfied with my use of symbols to show which list each mountain belonged to. There were actually 5 lists:

  1. The Eight-thousanders
  2. Bass's Seven Summits
  3. Bass's Seven Second Summits
  4. Messner's Seven Summits
  5. Messner's Seven Second Summits

After throwing ideas at each other, I realised that I could group their labels as in the bottom-right graphic. RacerX11's finishing touch was to use symbols to distinguish just the 4 mountains in the Bass and Messner lists.

It would have been superb to add the Volcanic Seven Summits and Volcanic Seven Second Summits, too, but the image was already a bit cluttered, and there is controversy on the membership of the list itself.

UPDATE 31 JAN 2013: I just found out that Thomas Laussermair has written on very kind words about the image and a mod I made for him — thanks, tlausser/visualign!

Cambridge Kings Hedges Jenny Wren bus stop RL.jpg 29 Feb 2012[edit]

Stereo image
Left frame 
Cambridge Kings Hedges Jenny Wren bus stop RL.jpg
The Jenny Wren pub and bus stop with an approaching Citi 1 bus.

I borrowed a Fujifilm FinePix Real 3D W3 and went around taking photos of Cambridge in the summer of 2011.

I then developed the stereoscopic image template to let the reader view the pictures in various arrangements — though cross-eye viewing lets one view much larger images, some people find parallel viewing more comfortable. The template supports side-by-side sterescopic images in both parallel and cross-eye formats.

The first stereoscopic photograph I've uploaded to Wikimedia shows a Stagecoach Citi 1 bus in King's Hedges, a residential suburb to the north of Cambridge city centre.

ideal projectile motion for different angles.svg 25 Feb 2012[edit]

Trajectories of projectiles launched at different elevation angles but the same speed of 10 m/s in a vacuum and uniform downward gravity field of 10 m/s2. Points are at 0.05 s intervals and the length of their tails is linearly proportional to their speed. t = time from launch, T = time of flight, R = range and H = highest point of trajectory (indicated with arrows).

Being a markup language, SVG code can be easily generated using a program or script, as contrasted with, say, a Microsoft PowerPoint or Adobe Illustrator drawing.

As I'm already familiar with Perl (the less-abstruse constructs, at least!) I wrote scripts to automatically generate SVG code which would otherwise need lots of calculator work, such as drawing graphs. Labels and other decoration can then be manually added to complete the illustration. So that a future editor can easily tweak the code, the Perl code appends itself to the SVG as a comment.

An example is this diagram which shows how a projectile behaves when launched at different angles of elevation, in the absence of air resistance.

As is often taught in high school, a 45° angle indeed gives the greatest range. What is less known is that the time of flight increases with the angle. Since there are two complementary angles giving the same range (other than the maximum), it is possible for artillery to fire a shell at the larger angle, then another at the smaller one after an appropriate delay, so that both shells hit the target at the same time — a basic form of Multiple Rounds Simultaneous Impact.

moon names.svg 17 Feb 2012[edit]

SVG with English labels
Lunar nearside with major maria and craters labeled.
SVG with German labels
Erdzugewandte Seite des Mondes mit den größten Maria und Kratern gekennzeichnet.

One of the features of SVG which initially attracted me was the ability to embed a bitmap, such as a JPEG or PNG image.

This allows labels to be relatively easily updated or translated into different languages, by simply replacing the English terms in its SVG file with a text editor, as Timwi has done with the lower image on the right.

Original JPEG with "burnt-in" labels.

Before I learnt this, I had to tediously edit the image on the left to correct "sea of storms" to "sea of crises" (unable to find a matching font, I manually cut and pasted individual letters from other labels).

Even after doing so, saving the file slightly reduced the image quality, due to the lossy compression JPEG uses.

UPDATE 31 JAN 2013: provides a useful facility to convert an image to Base64 which can then be embedded in an SVG as follows (replace png with jpeg, gif etc as needed):

<image x="0" y="0" width="WIDTH_IN_PIXELS" height="HEIGHT_IN_PIXELS" xlink:href="data:image/png;base64,IMAGE_BASE64_DATA"/>

tensegrity simple 3.gif 21 Jan 2012[edit]

Stereo image
Left frame 
Tensegrity simple 3 RL.png
Animation Each compression member (green) is symmetric with the other two, and symmetric from end to end. Each end is connected to three cables (red) which provide compression and define the position of that end.
The simplest tensegrity structure

I had once toyed with the POV-Ray raytracer. Similar to SVG, one writes code to render an image, but in this case in 3D!

As I thought that the tensegrity article could benefit from an animation to better show the spatial relationship of structural parts, I decided to create a GIF animation of a rotating structure, as on the left.

Alas, a reader found the animation distracting, so we settled on a static image with a link to the animation. It then occurred to me to use the stereoscopic image template I had once written to display the static image, as on the right...

Herbert Backstage Pass cmglee 65.jpg 2 Jan 2012[edit]

Group photograph at the end of Herbert Backstage Pass.

Following the success of Derby Backstage Pass, the UK chapter of the Wikimedia Foundation and Herbert Art Gallery and Museum, Coventry organised Herbert Backstage Pass on 1 October 2011. The first major Wikipedia-related event I've attended, it seemed a perfect collaboration: The museum giving Wikimedia members access to parts not normally accessible to the public as well as time and attention of curators and other staff, while the members contribute to Wikipedia articles relevant to the collection.

This photograph (right) was taken at the end of the day, sadly after several attendees had left. More photos of the event can be viewed here.


UPDATE 15 OCTOBER 2013: My visit to the Coventry Transport Museum to see ThrustSSC on that Coventry trip has just resulted in my third (as far as I can tell) image to appear on the Wikipedia main page (today's). ☻✌

symmetrical 5-set Venn diagram.svg 1 Jan 2012[edit]

Five-set Venn diagram using congruent ellipses in a radially symmetrical arrangement devised by Branko Grünbaum. Labels have been simplified for greater readability; for example, A denotes ABcCcDcEc (or A ∩ ~B ∩ ~C ∩ ~D ∩ ~E), while BCE denotes AcBCDcE (or ~ABC ∩ ~DE).

This image marked the start of my regular contribution of SVG drawings to Wikipedia. I found using SVG to mark up images (especially by hand or with a script) rather fascinating, albeit at times frustrating due to

  1. Some design decisions of the SVG syntax (e.g. specifying label text colour with fill but line colour with stroke), and
  2. Limitations of the Wikimedia SVG renderer (e.g. not properly supporting markers making arrows hard to code, and differences in text size and alignment between the rendered PNG and browser render, as can be seen by the image's long editing history).

Making this image taught me various SVG basics, such as setting the viewport, creating and transforming primitives, and using defined blocks. A trick I found is how to properly outline overlapping shapes with semi-transparent fills (the naive method causes strokes (outlines) to be overlaid, giving outlines of inconsistent colour):

Make the collection of shapes a block, and use it once with zero stroke-opacity to paint the fills, then again with zero fill-opacity to paint the strokes.

The Venn diagram itself is an aesthetic solution to the puzzle of how to represent, with just ellipses, all possible combinations of five sets with each combination appearing exactly once.[2]

Do you still have the code you used to generate the 5 way Venn diagram? I would like to know the center of the ellipsis; hence I'm asking. Thank you. Carstensen (talk) 10:00, 16 April 2013 (UTC)

Kudos for this, capturing all 31 possibilities (2^5-1). @Carstensen the code is closer than you might think. Go to the svg file and use your browser to View | Source. The bulk is the text markers. The essence is six lines: one <ellipse> and three <use>s. The center of rotation appears to be offset 17.5% of the major radius, and 22.5% of the minor radius. Very smartly crafted SVG code. Bob Stein - VisiBone (talk) 19:23, 28 October 2014 (UTC)

Sweet! Thanks! Carstensen (talk) 11:09, 10 December 2014 (UTC)

UPDATE 18 APRIL 2016: Appeared in The Guardian[3] on John Venn's 180th birthday!

  1. ^ "Complex Roots Made Visible – Math Fun Facts". Retrieved 1 October 2016. 
  2. ^ Grünbaum, Branko, "Venn Diagrams and Independent Families of Sets." Math. Mag. 48, 12-23, 1975.
  3. ^ "John Venn Google doodle: the right way to do a Venn diagram". The Guardian. Retrieved 18 April 2016.