Biswajit Banerjee

Reading VTK particles in Javascript

Parsing simulation output data

In our previous article we discussed the process of calling XML functions in C++ to write data produced by particle simulations in XML format files. These output files can then be viewed in powerful tools such as VisIt or Paraview.

Modern browsers have become powerful in recent years and can now render 3D graphics quite efficiently. The next few blog posts will discuss our experience in developing a basic user interface with standard web tools to visualize simulation output data. In this article, I will explain how the input process would work in such a tool.

The VTK XML particle data format

Output data from our simulations are written to disk every few timesteps. At each timestep, we get a data file like the one listed below.

<?xml version="1.0"?>
<VTKFile type="UnstructuredGrid" version="0.1" byte_order="LittleEndian" compressor="vtkZLibDataCompressor">
      <DataArray type="Float64" Name="TIME" NumberOfTuples="1" format="ascii" RangeMin="0" RangeMax="0">
    <Piece NumberOfPoints="412" NumberOfCells="0">
        <DataArray type="Float64" Name="Radius" NumberOfComponents="3" format="ascii" RangeMin="0.83645083538" RangeMax="1.7075128111">
          1.15 0.5 0.69 1.15 0.5 0.69
          1.15 0.5 0.69 0.575 0.5 0.345
          1.15 0.5 0.69 0.575 0.5 0.345
          0.575 0.5 0.345 1.15 0.5 0.69
        <DataArray type="Float64" Name="Axis a" NumberOfComponents="3" format="ascii" RangeMin="1.923869419" RangeMax="3.8462926819">
          2.250054 1.570796 0.6792575 2.536687 1.570796 0.9658907
          0.8923825 1.570796 0.6784139 1.576513 1.570796 0.005716666
          1.523425 1.570796 0.0473713 2.090659 1.570796 2.62173
          2.128077 1.570796 0.5572808 0.0678937 1.570796 1.502903        
        <DataArray type="Float32" Name="Points" NumberOfComponents="3" format="ascii" RangeMin="29.109286806" RangeMax="99.675090711">
          49.731739044 0 0.99367249012 50.900959015 0 5.3205289841
          30.811420441 0 3.8980329037 54.012958527 0 5.2524261475
          71.06136322 0 7.7233600616 65.397392273 0 6.3970627785
          42.516990662 0 11.963620186 48.560340881 0 2.5312030315
        <DataArray type="Int64" Name="connectivity" format="ascii" RangeMin="1e+299" RangeMax="-1e+299">
        <DataArray type="Int64" Name="offsets" format="ascii" RangeMin="1e+299" RangeMax="-1e+299">
        <DataArray type="UInt8" Name="types" format="ascii" RangeMin="1e+299" RangeMax="-1e+299">

For now, let us consider only data that have been written in ASCII format.

Reading the particle XML data

We will read in this data into a code that provides a user interface to visualize the particle data.

For now, the details of the user interface are not important, but it is worth noting the following :
* it is written in Typescript
* it uses the Vue framework for user interactions and Vuex for data management
* Typescript type definitions are from npm and av-ts
* the code is transpiled to ES6 using tsc and babel, and
* packaged using webpack before it is run on a browser

First let us examine the actual reader code:

// Function: parseAndSaveVTKXML
// Input:   xmlDoc : XMLDocument (The current version of jquery.d.ts declares the return type of parseXML as "any")
// Store:   pointData : any {} (A key-value object containing arrays of particle data)
public parseAndSaveVTKXML(xmlDoc : any) {
    // Use jquery to read the xmlDoc
    let $xml = $(xmlDoc);
    // Check that the file is actual a VTK XML file
    let fileType = $xml.find("VTKFile").attr('type');
    if (fileType != "UnstructuredGrid") {
      console.log("Invalid file type" + fileType);
    // Read time stamp after trimming the string
    let timeStr = $xml.find("FieldData").find("DataArray").text().trim();
    let time = parseFloat(time);
    // Read the number of points in the data set
    let numPtsStr = $xml.find("Piece").attr('NumberOfPoints').trim();
    let numPts = parseFloat(numPtsStr);
    // Read the state data one by one
    var pointData:any = {};
      function () {
        let data = $(this);  // "this" is now pointing to the data XMLDocument
        let key = data.attr("Name");  // Use the name of the variable as the key
        let type = data.attr("type");
        let numComponents = data.attr("NumberOfComponents"); 
        if (numComponents) {  // More than one component
          let numComp = parseFloat(numComponents);
          let dataArray = convertTo1DArray(data);
          let arrayOfVec : any = [];
          while (dataArray.length) {
            arrayOfVec.push(dataArray.splice(0,numComp));  // Convert into 2D array        
          pointData[key] = arrayOfVec; // Save the data                      
        } else { // Only one component
          pointData[key] = convertTo1DArray(data);

    // Read the position data
    let points = $xml.find("Points").find("DataArray");
    let numComponents = points.attr("NumberOfComponents");
    let dataArray = convertTo1DArray(points);
    let numComp = parseFloat(numComponents);
    let arrayOfVec : any = [];
    while (dataArray.length) {
    pointData["Position"] = arrayOfVec;
    // Save the data
    Store.commit('SET_VTK_PARTICLE_DATA', pointData); // Save to a Vuex store

The convertTo1DArray method has the form

private convertTo1DArray(data : string) : number[] {
  let dataArray =
    data.text().trim()                 // Remove leading/trailing white spaces
        .replace( /\s\s+/g, ' ' )      // Replace multiple spaces with one space
        .replace(/(\r\n|\n|\r)/gm," ") // Replace newline characters with space
        .split(" ")                    // Convert into 1D array
        .map(parseFloat);              // Convert into floats
  return dataArray;

Note that we have not saved the time in this simplified version. The JSON representation of the saved data is of the form

  "Position": [[1,2,3],[4,5,6],....],
  "Radius": [[0.1,0.2,0.3],[0.4,0.5,0.6],....],
Asynchronous IO and reading the file

The previous section assumes that a file has been read in and parsed using parseXML. Let us now see how the file is actually read from disk. We will assume that a File object is available and has been obtained using the <input> tag in the HTML document. The FileReader feature of HTML5 makes it possible to read files easily from disk without using the fs library from node.js.

The HTML input tag inside the appropriate Vue template has the form

<input type="file" v-on:change="readVTKXMLParticleFile">

The readVTKXMLParticleFile function is called after a file has been selected from the list of files.

public readVTKXMLParticleFile(event: any) {
  // Choose first file in selected list
  let file =[0];
  // Read the file as text and store data
  this.readXMLFile(file, "vtk-xml");

The actual asynchronous read is defined in readXMLFile:

import $ = require("jquery"); // Use jQuery for the parsing
  public readXMLFile(file: any, type: string) {

    if (file) {
      let reader = new FileReader();
      // Use a closure to pass on the actual parser function
      reader.onload = (
        function(parserFunction: any) {
          return function(event: any) {
            let xmlData = reader.result;      // The text that has been read in
            let xmlDoc = $.parseXML(xmlData); // Use jQuery to do the parsing
      )(this.parseAndSaveVTKXML); // This is where the actual parsing is done

    } else {
      console.log("Unable to load file ",;


Once the data have been stored, we can start the process of visualization. Future articles will discuss our experience with vtk.js and three.js. We will also discuss the potential and pitfalls of Typescript and Vue.

If you have questions/comments/corrections, please contact banerjee at parresianz dot com dot zen (without the dot zen).