Tormod Landet

  1. Installing R in WSL

    Short note for posteriority: the documentation for R tells you to install the key for the Ubuntu 18.04 R repository via the following command:

    sudo apt-key adv --keyserver --recv-keys E298A3A825C0D65DFD57CBB651716619E084DAB9

    Using apt-key in this way does not work on Ubuntu 18.04 on Windows 10 (in WSL, Windows Subsystem for Linux) due to use of an unsupported IPC mechanism for communicating with dirmngr (at least with Windows 10 version 1709). The command gives the following error:

    Executing: /tmp/apt-key-gpghome.Sf0LlrpiEu/ --keyserver --recv-keys E298A3A825C0D65DFD57CBB651716619E084DAB9
    gpg: connecting dirmngr at '/tmp/apt-key-gpghome.Sf0LlrpiEu/S.dirmngr' failed: IPC connect call failed
    gpg: keyserver receive failed: No dirmngr

    Instead run the following command to add the key:

    curl -sL "" | sudo apt-key add

    How did I get the number 0x51716619E084DAB9? Add the repository and run apt update:

    sudo add-apt-repository 'deb bionic-cran35/'
    sudo apt update

    It will complain:

    Err:2 bionic-cran35/ InRelease
      The following signatures couldn't be verified because the public key is not available: NO_PUBKEY 51716619E084DAB9


    The full set of commands to install R from the repositories:

    curl -sL "" | sudo apt-key add
    sudo add-apt-repository 'deb bionic-cran35/'
    sudo apt-get update
    sudo apt-get install r-base

    I'm not really much of an R user, but I needed it this one time and I prefer using WSL when I have to use Windows.

  2. Ocellaris and my PhD project

    Ocellaris, the free surface two phase Navier-Stokes solver I have been building as a part of my PhD work at the Department of Mathematics at the University of Oslo is finally ready for the world! It is built using an exactly divergence free unstructured DG FEM numerical method that uses special velocity slope limiters to stabilise the higher order shape functions near the factor 1000 jump in momentum at the air/water free surface, which enables higher order simulations of free surface flows such as ocean waves. I just released version 2019.0.1, which has greatly improved the user guide and general documentation.

    You can see an example simulation below, more examples are available on the Ocellaris Blog.

    The PhD thesis and the included papers are all soon ready as pre-prints. Looking forward to delivering this winter!

  3. Cleaning cruft in a Ubuntu installation

    I have been running Linux on my laptops for as long as I have been owning laptops. My current one has been through quite a few versions of Ubuntu Linux. There are normally no big changes from version to version, but it steadily improves. In that process some software packages that previously were installed by default are replaced or become obsolete. Most of this software is still updated, but it is no longer present on new installations and it may have no value to me.

    Every time I update to a new version of Ubuntu I have a feeling that the cruft is building up, but there is no obvious way that I have found to remove software that has no longer been found worthy of being part of the default install. So, this time I decided to write a script to identify these packages. The script takes a list of software packages to keep, i.e the software that I use directly when I use my machine. From this list it finds all the packages that must be present for my prefered software to work, and lists all the other that can in princible be removed.

    In reality, it is hard to come up with a list of all the software that is required for a Ubuntu installation to work properly, so I also made the script read the newest list of default packages. Any of these that happen to be installed are also kept along with their dependencies.

    The script itself, can be found on Bitbucket. It does not uninstall any packages by itself, but suggest packages that you can uninstall by running apt purge PCKGNAME or similar. I put it online and write about it here in the hope that it will be useful to someone else.

    I started out with the following list of packages to keep,

    python3 ubuntu-desktop krita thunderbird firefox ubuntu-minimal digikam gimp darktable gedit Downloads/ubuntu-16.10-desktop-amd64.manifest -v

    and after some trial and error I ended up with the following by noticing packages suggested for removal that I in reality wanted to keep and adding them to the command line invocation,

    python3 ubuntu-desktop krita thunderbird firefox ubuntu-minimal digikam gimp darktable gedit clang bcmwl-kernel-source systemsettings enfuse ffmpegthumbs flashplugin-installer gstreamer1.0-nice gstreamer1.0-plugins-bad gstreamer1.0-plugins-bad-faad  gstreamer1.0-plugins-bad-videoparsers gstreamer1.0-plugins-ugly gstreamer1.0-plugins-ugly-amr hugin hugin-tools inkscape gimp-ufraw git  gmsh gphoto2 gstreamer1.0-libav gimp-plugin-registry  ttf-mscorefonts-installer mercurial kipi-plugins gimp-gmic sshfs Downloads/ubuntu-16.10-desktop-amd64.manifest -v

    This allowed me to get rid of approximately 4 GB of cruft without any noticeable problems afterwards. Note that I did not blindly remove all packages suggested by the above final invocation, but I removed maybe 90% of them. The script is not smart, it can easily suggest that you remove your network driver, so use with caution!


    Hoarfrost / Yay!

  4. 3D visualization in IPython notebook

    The IPython notebook is a great tool for interactive computing. Below I show a short Python example on how to work interactively with simple 3D models in the IPython notebook.

    To begin with, we need a simple way to represent a 3D model in Python. We will here restrict ourselves to geometries made up of plane 2D polygons in 3D:

    class PolygonCollection(object):
        def __init__(self, nodes, polygons):
            A collection of plane polygons in 3D
            Nodes is a list of 3-tuples of coordinates
            Polygons is a nested list of node indices for each polygon, one
            list of indices for each polygon
            self.nodes = nodes
            self.polygons = polygons
        def extents(self):
            Calculate the size of the 3D polygon model
            xmin, ymin, zmin = 1e100, 1e100, 1e100
            xmax, ymax, zmax = -1e100, -1e100, -1e100
            for x, y, z in self.nodes:
                xmin = min(x, xmin)
                ymin = min(y, ymin)
                zmin = min(z, zmin)
                xmax = max(x, xmax)
                ymax = max(y, ymax)
                zmax = max(z, zmax)
            return xmin, xmax, ymin, ymax, zmin, zmax

    We then populate our 3D model with some polygons. The polygon <-> node/vertex connectivity is simply represented as indices into the list of nodes/vertices:

    nodes = [(0.0, 0.0, 0.0), (1.0, 0.0, 0.0), (1.0, 1.0, 0.0), (0.0, 1.0, 0.0),
             (1.0, 0.0, 0.0), (1.0, 0.0, 1.0), (0.0, 0.0, 1.0),
             (0.0, 1.0, 0.0), (0.0, 1.0, 1.0), (0.0, 0.0, 1.0)]
    elements = [(0, 1, 2, 3), (0, 4, 5, 6), (0, 7, 8, 9)]
    polygons = PolygonCollection(nodes, elements)

    IPython notebook allows us to embed images, Latex equations, HTML, Javascript and more. In this case we will use Javascript to render our 3D model. This can be done as follows:

    from IPython.display import display, Javascript
    js = polygons2js(polygons)

    The end result can be seen below. In a reasonable modern browser you should be able to rotate the 3D model. If it does not work it may also be that the Javascript libraries imported to do the rendering have changed since the time when this was written. I do not believe the libraries have stable APIs just yet. As of May 2014 it should work in the latest versions of Firefox, Chrome and Internet Explorer.

    Update March 2017: fixed the javascript (and the below Python code) so that it works with the latest (Nov 2016) version of JSModeler. The python code below should produce the same Javascript code that is running on this page, but it has not been tested, so there might be a typo somewhere.

    The "only" thing needed to make this work is a function polygons2js(polygons) that converts the 3D model to a set of javascript commands that will run inside the notebook. This can be done quite easily with the help of JSModeler and three.js. A simple implementation will look something like this:

    import random
    def js_wrap(js, js_libraries):
        Wrap javascript commands in code that preloads needed libraries
        lines = ['function getMultipleScripts(scripts, callback) {',
                 '  if(scripts.length == 0)',
                 '    callback();',
                 '  jQuery.getScript(scripts[0], function () {',
                 '    getMultipleScripts(scripts.slice(1), callback);',
                 '  });',
                 'var scripts = [']
        lines += ['"%s",' % script for script in js_libraries]
        lines += ['];',
                  'getMultipleScripts(scripts, function() {',
        return '\n'.join(lines)
    def polygons2js(polygons):
        # Find the size of the 3D model. We use this later to make the default camera and rotation point work
        xmin, xmax, ymin, ymax, zmin, zmax = polygons.extents()
        length = max([xmax-xmin, ymax-ymin, zmax-zmin])
        xmean = (xmin + xmax)/2
        ymean = (ymin + ymax)/2
        zmean = (zmin + zmax)/2
        # Generate the Javascript code
        # see for details
        canvas_id = 'js3dcanvas_%d' % random.randint(0, 1e10)
        canvas_style = 'style="border: 1px solid black;" width="800" height="600"'
        js = ['var widget = jQuery(\'<canvas id="%s" %s></canvas>\');' % (canvas_id, canvas_style),
              'var viewerSettings = {',
              '  cameraEyePosition : [-2.0, -1.5, 1.0],',
              '  cameraCenterPosition : [0.0, 0.0, 0.0],',
              '  cameraUpVector : [0.0, 0.0, 1.0]',
              'var viewer = new JSM.ThreeViewer();',
              'viewer.Start(widget, viewerSettings);',
              'var body = new JSM.Body();']
        # Add node coordinates
        coords = []
        for coord in polygons.nodes:
            coords.append('[%f, %f, %f]' % ((coord[0]-xmean)/length,
            #print coord, coords[-1]
        js.append('var coords = [%s];' % ', '.join(coords))
        js.append('for (var i = 0, len = coords.length; i < len; i++) {')
        js.append('  body.AddVertex(new JSM.BodyVertex('
                  'new JSM.Coord(coords[i][0], coords[i][1], coords[i][2])));')
        # Add elements
        polys = [repr(list(poly)) for poly in polygons.polygons]
        js.append('var elems = [%s];' % ', '.join(polys))
        js.append('for (var i = 0, len = elems.length; i < len; i++) {')
        js.append('  body.AddPolygon(new JSM.BodyPolygon(elems[i]));')
        js.append('var meshes = JSM.ConvertBodyToThreeMeshes(body);')
        js.append('for (var i = 0; i < meshes.length; i++) { viewer.AddMesh (meshes[i]); }')
        # Wrap final js code in a library loader to make sure JSModeler and three.js are available
        js = '\n'.join(js)
        libraries =  ['',
        return js_wrap(js, libraries)

    I have found this to be quite handy for inspecting small 3D models and modifying the interactively. The PolygonCollection class can also be updated to automatically display the 3D model in the notebook if it objcts of the class are left on the last line of an IPython input cell, or if you call display() on them. For this to work you will have to extend the PolygonCollection class with a _repr_javascript_ metod that returns polygons2js(self) (as a string). IPython will automatically look for this method and call display(Javascript(polygons._repr_javascript_())) for you.

Page 1 / 9 »