Question

I use Jason Davies' wordcloud library for d3 (https://github.com/jasondavies/d3-cloud) and my problem is that the words in the cloud overlap.

I am aware that there are already questions regarding this issue on stack overflow (and other sites), but none of these helped in my case.

In the following example I use the example cloud from Jason Davies' site and altered only a few things:

  • I read my words and their sizes from an external file.
  • I set rotation to 0. Rotation angle does not seem to make a difference though.
  • I commented out the "Impact" font, to rule out any issues with loading the font. (It makes no difference either though.)

Here is my code:

<!DOCTYPE html>
<meta charset="utf-8">
<body>
<script src="d3.js"></script>
<script src="d3.layout.cloud.js"></script>
<script>
   d3.tsv("testdata.txt", 
  function(error, data) {

  var fill = d3.scale.category20();



  d3.layout.cloud().size([300, 300])
      .words(data)
      .padding(1)
      .rotate(function(d) { return 0; })
  //    .font("Impact")
      .fontSize(function(d) { return d.size; })
      .on("end", draw)
      .start();

  function draw(words) {
    d3.select("body").append("svg")
        .attr("width", 300)
        .attr("height", 300)
      .append("g")
        .attr("transform", "translate(150,150)")
      .selectAll("text")
        .data(words)
      .enter().append("text")
        .style("font-size", function(d) { return d.size + "px"; })
    //    .style("font-family", "Impact")
        .style("fill", function(d, i) { return fill(i); })
        .attr("text-anchor", "middle")
        .attr("transform", function(d) {
          return "translate(" + [d.x, d.y] + ")rotate(" + d.rotate + ")";
        })
        .text(function(d) { return d.word; });
  }
  }
  )

</script>

The testdata looks like this (the color information is not used in the example):

word    size    color
der 39  #a9a9a9
die 37  #a9a9a9
und 30  #a9a9a9
athenischen 29  #a9a9a9
Die 29  #a9a9a9
eine    28  #a9a9a9
,   27  #a9a9a9
einer   26  #a9a9a9
attischen   26  #a9a9a9
liberalen   26  #1e90ff
zur 25  #a9a9a9
athenische  24  #a9a9a9
christliche 23  #a9a9a9
attische    23  #a9a9a9
_START_ 22  #a9a9a9 
reinen  22  #a9a9a9
englischen  21  #a9a9a9 
oder    21  #a9a9a9
--  21  #a9a9a9
radikalen   21  #a9a9a9
Q*M 21  #a9a9a9
Q*M 21  #a9a9a9
christlichen    20  #a9a9a9
schöne  20  #1e90ff
repräsentativen 20  #a9a9a9
sozialen    20  #a9a9a9
hellenische 19  #1e90ff
modernen    19  #a9a9a9
radikale    19  #a9a9a9
griechische 19  #a9a9a9
-   18  #a9a9a9
schönen 18  #1e90ff
alle    18  #a9a9a9
radicalen   18  #a9a9a9
als 17  #a9a9a9
neuen   17  #a9a9a9
perikleischen   16  #a9a9a9
bürgerlichen    16  #a9a9a9
Namen   16  #1e90ff

If I run the js script with the test data my word cloud comes out with overlaps. Sometimes it only happens after a few reloads, but it is fairly frequent.

Other people reported the same issue and found that it was related to using web fonts or skipping the rotate parameter. This does not apply in my example.

I suspect that it might be related to the fact that there are to many words for the canvas size, however, I also did tests where I signficantly increased the canvas size and it still happened (though less frequently, as the random placement of the words made it less likely). In addition to that, you can see that several words are not shown at all due to the small canvas size. Why leave some out and create overlap for others? So I think the issue lies elsewhere.

Any ideas?

Thanks!

Was it helpful?

Solution

I ended up asking Jason Davies himself and it was actually a pretty simple mistake: You have to specify the text accessor function in the first statement (not only in the "draw" function). It works if you add one line like this:

d3.layout.cloud().size([300, 300])
  .words(data)
  .padding(1)
  .rotate(function(d) { return 0; })
//    .font("Impact")
  .text(function(d) { return d.word; }) // THE SOLUTION
  .fontSize(function(d) { return d.size; })
  .on("end", draw)
  .start();

OTHER TIPS

The fix by the author didn't work for me. What did work was to specify the .font() in the cloud layout setup and declare the same font-family in the draw code - so ironically, the bit that the author above has commented out. This gives the code the font to calculate dimensions from. Without it, it takes a best guess.

d3.layout.cloud().size([800, 400])
  .words(words)
  .font('Impact') // <-- what mattered
  .fontSize(d => d.size)
  .on('end', draw)
  .start()

You'll need to also specify the font-family in the draw() function so that the cloud renders the words in the correct font that matches what you declared above :

.attr('font-family', 'Impact')

I have tried a sample for you to fiddle with, Please have a look. wordcloud without overlap

essentially:

<div id="cloud"></div> 

// First define your cloud data, using `text` and `size` properties:


var fill = d3.scale.category20();
var words = {
"Battery Related": "52382",
"Billing": "52412",
"Break Related": "52490",
"Chain Related": "52471",
"Clutch Related": "52468",
"Dealer attitude": "52488",
"Electrical Related": "52352",
"Engine Related": "52446",
"Handle Bar Related": "52486",
"Happy": "52472",
"Jerking": "52325",
"Jerking Problem": "52325",
"Low Mileage": "52489",
"Noise": "52462",
"Poor Pickup": "52406",
"Running Off": "52242",
"Service Quality": "52488",
"Silencer Problem": "52468",
"Starting Trouble": "52490",
"Suspension Related": "52365",
"Vehicle Noise": "52467",
"Vibration": "52463",
"Washing": "52488"
};
var max_freq = 52490;
var cloudwords = ["Battery Related", "Billing", "Break Related", "Chain Related", "Clutch Related", "Dealer attitude", "Electrical Related", "Engine Related", "Handle Bar Related", "Happy", "Jerking", "Jerking Problem", "Low Mileage", "Noise", "Poor Pickup", "Running Off", "Service Quality", "Silencer Problem", "Starting Trouble", "Suspension Related", "Vehicle Noise", "Vibration", "Washing"];
var url = 'http://xxx.yyyy.zz.ww/?q=abc/';
var width = 800,
height = 800;

var leaders = cloudwords
.map(function(d) {

return {
  text: d,
  size: 5 + (words[d] / max_freq) * 0.9 * 30 // *the size of the "box" occupied by each word. has no relation to text size.
};
})
 .sort(function(a, b) {
 return d3.descending(a.size, b.size)
});

var leaderScale = d3.scale.linear().range([1, 20]); // *scale range to plot the relative sizes of the words.

leaderScale.domain([d3.min(leaders, function(d) {
return d.size;
}),
d3.max(leaders, function(d) {
return d.size;
})
]);

 // Next you need to use the layout script to calculate the placement, rotation and size of each word:

 d3.layout.cloud().size([width, height])
.words(leaders)
.padding(0) //fiddle with padding here, does not really have any effect    on overlap.
.rotate(function() {
 return ~~0; //to keep the words horizontal 
 })
 .font("Impact")
 .fontSize(function(d) {
 return d.size;
 })
 .on("end", drawCloud)
 .start();

 function drawCloud(words) {
 d3.select("#cloud").append("svg")
 .attr("width", width)
 .attr("height", height)
 .attr("text-align", "center")
 .append("g")
 .attr("transform", "translate(" + [width >> 1, height >> 1] + ")")    //for transalting words to their different postions.
 .selectAll("text")
 .data(words)
 .enter().append("text")
 .style("font-size", function(d) {
  return leaderScale(d.size) + "px"; //used scale to resize words to a linear scale.
 })
 .style("font-family", "Impact")
 .style("fill", function(d, i) {
  return fill(i);
 })
 .attr("text-anchor", "middle")
 .attr("transform", function(d) {
  return "translate(" + [d.x, d.y] + ")rotate(" + d.rotate + ")";
 })
 .text(function(d) {
  return d.text;
 })
 .on("click", function(d, i) {
  window.open(url + d.text);
 });
 }

 // set the viewbox to content bounding box (zooming in on the content, effectively trimming whitespace)

 var svg = document.getElementsByTagName("svg")[0];
 var bbox = svg.getBBox();
 var viewBox = [bbox.x, bbox.y, bbox.width, bbox.height].join(" ");
 svg.setAttribute("viewBox", viewBox);
Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top