Mat Alan Gilbert: IT Notes - Software Development and Other Scribbles

Going back to a previous directory in the Linux terminal

2013-11-28T13:33:00.000+00:00

A little cd trick to cd back in Linux terminal history:

 cd -

So useful I had to share!

Adding Syntax Highlighting to Linux less

2013-11-27T13:20:00.001+00:00

Install source-highlight:

 sudo apt-get install source-highlight

Add the following two lines to~/.bashrc

 export LESSOPEN="| /usr/share/source-highlight/src-hilite-lesspipe.sh %s"  
 export LESS=' -R '

Now source files opened with less will be nicely syntax highlighted. source-highlight supports many languages - see

 source-highlight --lang-list

for a full listing.

Using Nautilus file manager with Enlightenment window manager

2013-05-07T10:30:00.000+01:00

I'm a fan of Bodhi Linux, which uses the Enlightenment window manager and is pretty speedy on all kinds of hardware. Unfortunately however, the built in file manager isn't yet fully functional in my opinion. As Dropbox has a dependency on Nautilus, I've integrated that into Enlightenment as the primary file manager.

To do this, we need to start Nautilus with the --no-desktop option. Start by making moving /usr/bin/nautilus to /usr/bin/nautilus.original, and create a new script at /usr/bin/nautilus containing the following command:

 #!/bin/bash  
 nautilus.original --no-desktop $@ &

The $@ expands any passed arguments, allowing the use of the script as if it were the original Nautilus binary.

To ensure Enlightenment uses only Nautilus, we need to disable the integrated file manager from the Modules settings (Run Everything -> Modules -> Files). Disable EFM (Starter) and EFM Operation Info and close the settings window.

If you use the Places module, either on the desktop or in a panel, you'll also want to ensure that this launches Nautilus instead of the integrated file manager. To do this, right-click on the gadget, then go to Settings. Finally, select Use a custom file manager and enter nautilus in the box.

And we're done - we're now using Nautilus as a replacement for the Enlightenment File Manager.

References:[Crunchbang Forums: Make nautilus --no-desktop default.]

Scanf in Go lang on Windows

2013-04-06T21:15:00.001+01:00

A little gotcha when using fmt.Scanf in Go:

fmt.Scanf currently relies on the Unix line ending \n, while Windows uses \r\n. This means that when using fmt.Scanf, you must include the \n if you're building your Go program for Windows. For example:

 fmt.Scanf("%d", &num)

will work in Unix but not in Windows. You're most likely to notice this when using fmt.Scanf multiple times - either in a for loop or simply sequentially. fmt.Scanf will likely appear to run twice as often as you expect, leading to unexpected program flow.

To resolve this issue, just make sure to include the \n in your use of fmt.Scanf, as demonstrated below.

 fmt.Scanf("%d\n", &num)

Incidentally, if you're trying to track down errors like this one, you can always check function error codes, as demonstrated below.

  n, err := fmt.Scanf("%d", &num)  
     if err != nil {  
       fmt.Println(n, err)  
     }

In the above case, we get this error:

 20 unexpected newline

Which gives us the line number, and the issue from that line.

[References] Stack Overflow: parallel processing: How do I use fmt.Scanf in Go

Linux hibernate hangs on resume - Solved

2013-03-06T14:42:00.000+00:00

If s2disk successfully hibernates your machine, but on powering up the resume screen appears to hang, after it has appeared to stop working try pressing Alt-SysRq-E. This ends all processes in the virtual terminal used by s2disk, and, for me at least, allows me to resume working in my hibernated session.

Odd, but this might help others close to giving up on hibernation in Linux.

Fixing plugin problems with Banshee music player in Linux

2013-03-04T08:13:00.000+00:00

After installing Banshee on my minimalist Bodhi Linux install, trying to play music lead to Banshee asking to install plugins, and then hanging. Installing the following packages using the command below resolves the issue.

 sudo aptitude install gstreamer0.10-plugins-bad gstreamer0.10-plugins-good gstreamer0.10-plugins-ugly

Converting Latex to HTML using Pandoc

2013-02-21T16:16:00.000+00:00

Install Pandoc using your disto's package manager, or by following the installation instructions here.

Then use:

 pandoc input.tex -o output.html --mathjax

Nice and straightforward!

Linux Remote Desktops: A Comparison

2013-02-20T16:06:00.003+00:00

This is the final post in a three part series exploring Ubuntu-based remote desktop options. See posts one and two for an exploration of NoMachine and ThinLinc.

My primary motivation for testing these remote desktop solutions was my desire to find a remote desktop solution that worked as Windows Remote Desktop - in terms of speed and functionality. While Linux and Windows do have differing user paradigms - Windows being designed around a single user logon, I was looking for the ability to log on to my system both at work and at home, resuming my session from wherever I was. My work machine is behind a VPN, and I need my sessions to be responsive enough to work as if I was working on the local machine.

Productivity therefore is of primary importance to me, and closely tied to this is the latency of a connection. I find that a slow, lagging connection impairs not only my ability to work in the graphical environments I require, but also affects my desire and motivation to try to continue to work remotely. For these reasons, I've found that using NoMachine, via the FreeNx client is most suitable for my purposes - there was just too much latency within a Cendio Thinlinc connection during my testing. To be clear, your mileage may vary - the latency I experienced may be down to a specific interaction between something in my setup and Thinlinc, but during testing over a number of days, at a variety of times, under different network condition, the NoMachine sessions consistently delivered a near desktop speed experience.

I'm interested to hear about the experience of others, using either of the Linux remote desktop solutions I tested - feel free to add them in the comments below.

[Image Credit] lifeskillsfromthedojo.com

Another Linux Remote Desktop Solution: Cendio ThinLinc

2013-02-20T15:32:00.000+00:00

This is the second post in a three part series exploring Ubuntu-based remote desktop options. See posts one and three for further exploration and conclusions.

After the popularity of my earlier post reviewing FreeNx, Pierre Ossman, a developer at Cendio, informed me of their product, which provides an alternative remote desktop/thin client option for Linux. This post details my comparison of FreeNx and Cendio ThinLinc, as options for remote desktop solutions that rival Windows remote desktop.

Installation and use of Cendio ThinLinc

Obtaining and installing the software is relatively easy - you do have to give Cendio your email address before they will email you the download link, but the installation instructions are straightforward, and the installer does a good job of pulling in any additional required packages. The installation instructions simply consist of unzipping the downloaded package and executing an install script.

For me, the installer did seem to pull a a significant number of additional packages - these packages may just be required as ThinLinc provides a little more functionality than Nx. Overall, the ThinLinc installation took longer and felt a little less friendly than Nx, requiring slightly more configuration and interaction during the install process.

ThinLinc administration page

Once installed, the ThinLinc server is administered via a webpage at http://localhost:1010

Downloading and installing the client is also a simple process, and connecting to the server is straightforward, as can be seen in the box below:

It took me a bit of time to figure out how to disconnect from a session without logging off - the ability to resume a session from multiple locations is a primary driver of my remote desktop use. After some digging around in the documentation however, I found the relevant information - pressing F8 while logged into a remote session pops up a menu that provides the option to disconnect from the session, without ending it.

Now for the best bit - downloading and running the client on the same machine as the server allows you to start a ThinLinc session on that machine. Working in this session then allows connection from another location - using the ThinLinc client at work, for example, leaving programs running and then using the client from home to connect to that same machine gives the same experience as Windows Remote Desktop - the one feature I've missed ever since I've moved to using Linux full time.

Finally, it's worth noting that neither ThinLinc or NoMachine support the full Unity desktop that that is default in Ubuntu versions later than 11.04. Because of this, you need to ensure that gnome-session-fallback is installed:

 sudo apt-get install gnome-session-fallback

The Comparison
The results of my comparison can be found in the final post of this series: Linux Remote Desktops: A Comparison.

[Image credit] RoomMaster2000

Juniper Network Connect in Bodhi, Ubuntu and Linux Mint

2013-02-20T15:13:00.000+00:00

The Juniper Network Connect client simply wouldn't launch for me under any of the Linux distributions listed in the title of this post, even after trying multiple Java version installations.

Thankfully, Paul Smith at Mad-Scientist.us has put together a great script which allows connection to a Juniper WebVPN from a Ubuntu-based distro. He has a great write up on his blog, so I won't repeat his instructions here, other than to say he makes it incredibly simple to get connected to your institution WebVPN.

I followed his instructions on a reasonably fresh install of Bodhi Linux*, and ended up running into a "missing /etc/resolv.conf" error. This can be fixed by running the following command:

 sudo dpkg-reconfigure resolvconf

Answer Yes to all the prompts and restart your machine once the tool finishes.

Launching Paul's script (hosted here at Github, should you want to check out the code), should now present you with a working GUI, allowing you to connect you to your WebVPN. He also provides the option of connecting via the command line, detailed further in his blog post

As usual, any issues or comments in the box below!

* If you haven't checked this distro out, I highly recommend it - it absolutely flies even on older hardware, and is Ubuntu-based giving access to the huge Ubuntu software repositories

[References]
Mad-Scientist.us
Bohdi Linux Forums

LaTeX Error: File `algorithm.sty' not found

2013-01-21T15:49:00.002+00:00

To resolve this error in a Debian based Linux distribution (Debian, Ubuntu, Mint, etc), the fastest way to continue working is to install the texlive-science package, using:

 sudo apt-get install texlive-science

Printing from the Linux command line

2012-12-06T08:04:00.000+00:00

To print a file from the Linux command line, use the simple command:

 lpr FILENAME

It's that easy!

I came across this while trying to print all pdf files in a directory. To do this, the simple bash script you need is:

 #!/bin/bash  
 for file in *.pdf; do  
   /usr/bin/lpr "$file"  
 done

Save this into a file - I called mine print-all-pdfs.sh. Then make it executable using

 chmod +x print-all-pdfs.sh

Copy it to the directory you want to run it in, then execute with

 ./print-all-pdfs.sh

For an additional challenge, can anyone adapt the above script to print all .pdfs from directories recursively? Answers in the comments below!

[References] Bash script adapted from a LinuxForums post.

Notes on using Heroku

2012-12-05T22:06:00.000+00:00

These are a few common things I find myself looking up when working with Heroku - I update this post if I find myself repeated searching for how to do the same tasks.

Running commands from a folder not linked to Heroku

Simply append --app APP_NAME to any relevant Heroku command:

HEROKU_COMMAND --app APP_NAME

Deployment

If trying to push to Heroku:

 git push heroku master

but get a publickey error:

 Permission denied (publickey).  
 fatal: The remote end hung up unexpectedly

You need to add your public key to Heroku.

 heroku keys:add path_to_publickey

If you don't have a public key, Heroku can prompt and generate one for you - simply use the command:

 heroku keys:add

Resetting the application database

 heroku pg:reset DATABASE_URL

To reset the database without a validation prompt, use:

 heroku pg:reset DATABASE_URL --confirm APP_NAME

Rake

To run a rake task on Heroku, create the task in app/lib/tasks (Rails), test locally then deploy. To run, use:

 heroku run rake TASK_NAME

If you have any useful commands you regularly use while working with Heroku, let me know in the comments below!

How to add a menu item for SPSS in Linux

2012-12-04T12:30:00.000+00:00

I keep forgetting where SPSS installs itself - this .desktop file may be of help to others too.

Create .desktop file for SPSS:

 sudo leafpad /usr/share/applications/spss.desktop

Add this text to that file:

 [Desktop Entry]  
 Name=SPSS  
 GenericName=SPSS  
 Comment=Statistics program  
 Exec=/opt/IBM/SPSS/Statistics/19/bin/stats  
 Icon=gnome-monitor  
 Terminal=false  
 Type=Application  
 Categories=Application;Education;

Save, then log out and back in again, or restart the desktop using

 xfdesktop --reload

if using an XFCE desktop.

SPSS will now be available under Applications -> Education.

Find all files and move to a target directory from the Linux command line

2012-11-26T08:16:00.000+00:00

For this example, we're going to be moving all .mp3 files from the subdirectories of the current directory into current directory.

To do this, we need to combine two different commands. To find all the files in the current directory and its subdirectories, we use the find command:

 find -name \*.mp3

The \ escapes the *, so it's interpreted as a wild card on the command line. The -name argument allows us to check only the file names.

To move files around on the command line, we use the mv command, and for our example, we add the -t argument, which lets us define the target directory:

 mv -t <target-directory> <file-to-move>

where target-directory is where we want to move the file to, and file-to-move is the file we want to move.

With a bit of command-line magic, we can combine these commands using the -exec switch on find, to give us our final command:

 find -name \*.mp3 -exec mv -t . {} \+

-exec takes the command we want to execute, and {} \+ appends each find result to the command, which in our case is mv. So the above line will find all files end in .mp3 and move them to the current directory.

Fail Fast - A Lesson In Not Wasting Time

2012-11-19T08:30:00.000+00:00

Around a year ago, I had an idea for a small browser based application that had the potential to smooth and streamline a workflow within universities around the country. The concept was simple and wouldn't have taken a huge amount of time to develop. Given my position working in a university setting, I had the ideal environment to test and prototype, but due to constraints on my time I didn't take the concept any further.

Fast forward to last week. I'd finally found the inspiration, motivation, and time to solidify and explore this idea. It took me less than an hour to map out the specifics of my concept, and email a few academic contacts to see what they thought of the idea. Feedback was positive; I'd found a niche. Unfortunately, the very same week, the university have announced that they are trialling a new, much larger system which implements a substantial feature list, and includes the subset of functionality I was aiming to provide.

Why do I tell you this story? Well, there's lessons here. Despite my initial disappointment, this is a perfect example of failing fast and learning quickly. While there was a part of me that felt I shouldn't release my idea into the world until I had built the application and had a full, perfect implementation to demonstrate, had I done this I would have committed hours, days, or weeks of my time to building a system that had zero chance of ever being used. Here's what this experience taught me:

Talk about your ideas. It's natural to feel protective of your concepts, and worry that they might be stolen or plagiarised (or laughed at - see 3.). In my experience, it's highly unlikely that they will, and if they do, talking to enough people makes it clear where the idea originated. The knowledge and insights you stand to gain from sharing knowledge and ideas far outweigh any perceived risks of collaboration.
Act Early. If you have an idea, or discover a niche you could occupy, act on it early. If I'd had the confidence to explore my concept a year earlier, I'd have been first to market, so to speak, and would have been in a much better position to compete with the larger system that has come along now. Who knows, I might have gained enough traction to have become that larger system.
Fight your inner perfectionism. The world is imperfect, and so your contributions to it can be too. Don't spend forever perfecting designs, ideas and implementations. However scary it might be, put yourself and your ideas out there, and you'll be pleasantly surprised at the what happens.

So there we have it, it took me less than a day to establish that, despite a market need, my good idea wasn't commercially viable in current circumstances. My time is now free to explore my next project. Next time you find yourself thinking "maybe there's a need for this... one day I'll have the time..." - take the time to explore, find feedback and fail fast or fly.

[Image credit] http://thumannresources.files.wordpress.com

Finding out what Matlab tooboxes are available for use

2012-11-16T14:08:00.000+00:00

Finding out what toolboxes are available to a Matlab installation is actually relatively simple - it's just a case of typing:

ver

at the Matlab prompt.

On it's own, however, this function doesn't supply much useful information, as it doesn't tell you what licences are available to the installation, and hence which toolboxes you can actually use (which is probably what you care about when you're trying to find out what toolboxes are available).

It is possible to find out what licenses are available to the installation, however it takes a bit more work.

First, we need to define a feature string, which is just a list of all the toolboxes potentially installed:

 featureStr = {'Aerospace_Blockset'; ...  
        'Aerospace_Toolbox'; ...  
        'Bioinformatics_Toolbox'; ...  
        'Communication_Blocks'; ...  
        'Communication_Toolbox'; ...  
        'Compiler'; ...  
        'Control_Toolbox'; ...  
        'Curve_Fitting_Toolbox'; ...  
        'Data_Acq_Toolbox'; ...  
        'Database_Toolbox'; ...  
        'Datafeed_Toolbox'; ...  
        'Dial_and_Gauge_Blocks'; ...  
        'Distrib_Computing_Toolbox'; ...  
        'Econometrics_Toolbox'; ...  
        'EDA_Simulator_Link_DS'; ...  
        'Embedded_Target_c166'; ...  
        'Embedded_Target_c2000'; ...  
        'Embedded_Target_c6000'; ...  
        'Embedded_Target_MPC555'; ...  
        'Excel_Link'; ...  
        'Filter_Design_HDL_Coder'; ...  
        'Filter_Design_Toolbox'; ...  
        'Fin_Derivatives_Toolbox'; ...  
        'Financial_Toolbox'; ...  
        'Fixed_Income_Toolbox'; ...  
        'Fixed_Point_Toolbox'; ...  
        'Fixed-Point_Blocks'; ...  
        'Fuzzy_Toolbox'; ...  
        'GADS_Toolbox'; ...  
        'IDE_Link_MU'; ...  
        'Identification_Toolbox'; ...  
        'Image_Acquisition_Toolbox'; ...  
        'Image_Toolbox'; ...  
        'Instr_Control_Toolbox'; ...  
        'Link_for_Incisive'; ...  
        'Link_for_ModelSim'; ...  
        'Link_for_Tasking'; ...  
        'Link_for_VisualDSP'; ...  
        'MAP_Toolbox'; ...  
        'MATLAB'; ...  
        'MATLAB_Builder_for_dot_Net'; ...  
        'MATLAB_Builder_for_Java'; ...  
        'MATLAB_Distrib_Comp_Engine'; ...  
        'MATLAB_Excel_Builder'; ...  
        'MATLAB_Link_for_CCS'; ...  
        'MATLAB_Report_Gen'; ...  
        'MBC_Toolbox'; ...  
        'MPC_Toolbox'; ...  
        'NCD_Toolbox'; ...  
        'Neural_Network_Toolbox'; ...  
        'OPC_Toolbox'; ...  
        'Optimization_Toolbox'; ...  
        'PDE_Toolbox'; ...  
        'Power_System_Blocks'; ...  
        'Real-Time_Win_Target'; ...  
        'Real-Time_Workshop'; ...  
        'RF_Blockset'; ...  
        'RF_Toolbox'; ...  
        'Robust_Toolbox'; ...  
        'RTW_Embedded_Coder'; ...  
        'Signal_Blocks'; ...  
        'Signal_Toolbox'; ...  
        'SimBiology'; ...  
        'SimDriveline'; ...  
        'SimElectronics'; ...  
        'SimEvents'; ...  
        'SimHydraulics'; ...  
        'SimMechanics'; ...  
        'Simscape'; ...  
        'SIMULINK'; ...  
        'Simulink_Control_Design'; ...  
        'Simulink_Design_Verifier'; ...  
        'Simulink_HDL_Coder'; ...  
        'Simulink_Param_Estimation'; ...  
        'SIMULINK_Report_Gen'; ...  
        'SL_Verification_Validation'; ...  
        'Spline_Toolbox'; ...  
        'Stateflow'; ...  
        'Stateflow_Coder'; ...  
        'Statistics_Toolbox'; ...  
        'Symbolic_Toolbox'; ...  
        'SystemTest'; ...  
        'Video_and_Image_Blockset'; ...  
        'Virtual_Reality_Toolbox'; ...  
        'Wavelet_Toolbox'; ...  
        'XPC_Embedded_Option'; ...  
        'XPC_Target'};

That's a lot of typing, so I'd just copy and paste from above, into your Matlab prompt.

Next, we want to check that licenses exist for each of the items in the above list:

 index = cellfun(@(f) license('test',f),featureStr);  
 availableFeatures = featureStr(logical(index));

Licences that exist are now stored in availableFeatures, so typing that at the prompt gives us a list of licences that are available to the installation:

 availableFeatures

To see if a single license exists, we can also simply supply license('test',<feature_name>) with the single feature string we care about. For example:

 license('test', 'Neural_Network_Toolbox')

This is normally as much information as we need, but if we need to know if a licence is available to checkout at the present time, we can continue as below.

NB: Using license('checkout',<license>), as described below, is not normally a good idea when using network licences, as the license will not be released until Matlab is shut down. Usually, licenses are checked out automatically, as and when they are required by a toolbox.

So far, we've found out whether or not a license exists, but not if one is presently available for us to check out. To actually checkout a license, such as the Signal Toolbox, we can do:

 license('checkout','Signal_Toolbox')

Where Signal_Toolbox comes for the feature string list at the top of this post.

If Matlab returns:

 ans = 1

Then the license checkout was successful. We can also look at the licenses we currently have checked out out using:

 license('inuse')

As usual, I hope this has been helpful, and feel free to add any notes or observations in the comments below.

References:

Fixing Print Screen in Linux Mint XFCE

2012-11-06T00:47:00.000+00:00

To enable Print Screen key functionality in Linux Mint XFCE:

Ensure xfce4-screenshooter is installed:

 sudo apt-get install xfce4-screenshooter

Then, in the Menu, go to Settings -> Keyboard then Application Shortcuts -> Add. In the Command field, enter xfce4-screenshooter and then click OK, before pressing the Print Screen key on the keyboard, to set that shortcut.

Remote Desktop in Ubuntu 12.04

2012-11-06T00:23:00.000+00:00

This is the first post in a three part series exploring Ubuntu-based remote desktop options. See posts two and three for further exploration and conclusions.

In my opinion, both of the mainstream remote desktop/windowing approaches in Linux do not compare with the ease of use, and the speed of Windows remote desktop. VNC is inherently insecure, and setting up VNC over an SSH tunnel isn't the most user friendly of experiences. X forwarding over SSH is better, but in my experience there's simply too much latency when using it outside a LAN. Enter NX, or NoMachine - the ~~only~~ Linux remote desktop solution ~~I've come across~~ used that isn't a pain to get up and running, and is more than fast enough to be useable.

Here's what you need to do to get going with NoMachine. Below, target machine refers to the machine you want to connect to, and client machine refers to the machine you are connecting from.

On the target machine:

If it's not already installed, add openssh server to your machine:

 sudo apt-get install openssh-server

Add the FreeNX repository to your system using:

 sudo add-apt-repository ppa:freenx-team

Then install FreeNX with:

 sudo apt-get update  
 sudo apt-get install freenx

As Unity3D is too heavy to use over a WAN connection, allow fallback to a gnome session:

 sudo apt-get install gnome-session-fallback

On the client machine:
Use the NX client of your choice - ~~personally I used Remmina, with the Remmina NX plugin:~~ Update: While Remmina works well, the offical NX client gives a lot more options - see below for more.

 sudo apt-get install remmina
 sudo apt-get install remmina-plugin-nx

Start Remmina, and create a new connection. You should now be able to choose to connect using the NX protocol, as seen below.

Enter the target machine's IP address into the server field, and the username and password of the user account you wish to connect to. Hit connect and after a few moments your remote desktop will open.

You can also use the official NX client, and if you're connecting from Windows or Mac, that's probably easiest. For Linux, the official client download and install instructions are currently here.

I've not yet had the chance to experiment, but it may be possible to run the NX client from a different desktop on the target machine - if so, then this would make NX a complete, drop in replacement for Windows remote desktop, as far as I can see. Which is no bad thing, as other than the obvious proprietary graphics issues Linux can suffer from, a decent RDP implementation was the only only key feature I've noticeably missed since moving to Linux full-time (yes, there's X forwarding over SSH, but even with compression there's still too much lag to be usable over my connections). Update: It looks like the official NX client has a shadowing mode - again, I've not yet tested it, but it may be useful for the above.

Update Nov-16-2012: As noted in the comments, Cendio also offer a Linux remote desktop solution - after testing and comparing, my thoughts can be found in post two.

Installing the latest version of R into Debian Stable

2012-11-05T14:33:00.001+00:00

The R packages available in the Debian repository are are little out of date - to install the latest (stable) version we need to add a backports repository. Just follow the steps below!

Open the sources list with:

 sudo nano /etc/apt/sources.list

Then add this line to the bottom of the file (use your favourite mirror if desired):

 deb http://cran.ma.imperial.ac.uk/bin/linux/debian squeeze-cran/

The Debian backports are signed with the key of "Johannes Ranke (CRAN Debian archive) <jranke@uni-bremen.de>", with key ID 381BA480 . Import this using:

 sudo apt-key adv --keyserver subkeys.pgp.net --recv-key 381BA480

Then, update the packages cache - this should complete without errors.

 sudo apt-get update

Finally, install R and the R recommended packages using:

 sudo apt-get r-base r-recommended

To check the version of R that is now installed, use:

 R --version

To start R, simply type

at a command prompt. As usual, let me know of any issues or problems in the comments below.

Problems with R and RGtk2

2012-09-27T15:46:00.001+01:00

I ran across this issue while installing Rattle, a data mining library for the statistical programming language R, on to a Debian system. When installing the package RGtk2, seeing the below error means that your GTK version is not the version required by RGtk2.

 * installing *source* package ‘RGtk2’ ...  
 ** package ‘RGtk2’ successfully unpacked and MD5 sums checked  
 checking for pkg-config... /usr/bin/pkg-config  
 checking pkg-config is at least version 0.9.0... yes  
 checking for INTROSPECTION... no  
 checking for GTK... no  
 configure: error: GTK version 2.8.0 required  
 ERROR: configuration failed for package ‘RGtk2’

The required version can be found in libgtk2.0-dev, so install this using your preferred package manager - for example using apt-get:

 sudo apt-get install libgtk2.0-dev

Then reload R and:

 install.packages('RGtk2')

Now, after the installation has finished, you should see:

 * DONE (RGtk2)

As usual, let me know if it works for you in the comments below.

If you like this blog, and use R for statistical processing, you can help support my writing by purchasing a book from my bookstore. For both beginning and more advanced R users, I'd recommend Kenneth Knoblauch.

Install the latest version of R in Ubuntu

2012-09-20T16:25:00.001+01:00

The R version in the Ubuntu repositories is not the latest, so here's how to install R into Ubuntu 12.04:

Add the SSL key and import into apt:

 gpg --keyserver keyserver.ubuntu.com --recv-key E084DAB9  
 gpg -a --export E084DAB9 | sudo apt-key add -

Add the repository to your sources list, by doing:

 sudo nano /etc/apt/sources.list

Then add this line to the bottom of the sources file:

 deb http://cran.ma.imperial.ac.uk/bin/linux/ubuntu precise/

Then, to install the latest version of R, do:

 sudo apt-get update  
 sudo apt-get install r-base

Hit yes at the prompts, and voila, the latest version of R is installed!

Great gEdit Theme Repository

2012-09-10T15:37:00.000+01:00

I'm a fan of light on dark backgrounds when writing or coding, and Githubber kalmi maintains a great collection of gEdit (the default text editor in Gnome) themes - they can be found here: https://github.com/kalmi/gedit-themes

To use, clone the repo into a temporary directory, using:

 git clone https://github.com/kalmi/gedit-themes.git

Then copy them into a styles folder in the gEdit directory, using the install script from the repo:

 cd gedit-themes
 ./install.sh

To use them, open gEdit, then go to Edit -> Preferences -> Fonts and Colours. The themes should now be available in the Colour Scheme scroll box.

Enjoy!

Mounting an executable CD Drive in Linux

2012-07-19T17:14:00.000+01:00

I ran into the below problem while installing Matlab onto a Linux machine. Hopefully this will help someone else!

On most modern Linux systems, CD drives are automatically mounted. The problem comes when they are mounted without any exec permissions - this means that any scripts, etc on the CD, or .iso, will be unable to run. To resolve this, follow the steps below:

Find the name of your mounted CD drive by using:
```
 mount  
```
Then remount the device with the exec bit set:
```
 mount -o remount,exec $DEVICE
```

Where $DEVICE is the name of the CD drive.

You'll then be able to run all executables on the drive - in my case, the Matlab executable was then launched using the below commands:

 $ su  
 $ /path/to/mounted/cdrom/./install

Remember to logout as root when the installer has finished!

 logout

Betty and George - an overview

2012-07-18T11:24:00.000+01:00

This is the first in a series of posts detailing my latest hobby project. The bulk of the work was completed in under 10 hours, with minor styling tweaks taking a little longer to get right. The project is currently hosted here, and the source code is on Github here. As usual, please post any comments or opinions in the comments below, or tweet me directly @matalangilbert.

Introduction

Betty and George are Artificial Intelligence entities - they chat with each other in real time, and, somewhat incredibly in my opinion, make sense most of the time! They're sometimes witty, sometimes poignant, and sometimes inane, but they keep talking, no matter what!

Technical Overview

The application is split into two main parts. The backend is a Ruby program managing two Cleverbot instances (http://www.cleverbot.com), juggling responses between each instance. There's not too much to this, but I will go over the technical details in a future post.

Each conversational response is posted to a Rails application endpoint by the Ruby program, and then the front-end takes over. The frontend is backed by a Rails application, which receives data from the Ruby program, and displays it on the live site. There's a couple of neat techniques used on the main application page, which update the page in the background. The first is a background AJAX request, which hits the server every 5 seconds, retrieving new responses and adding them to the top of the scrolling conversation on the site. The second is an "Endless River of News" style endless scroll implementation. This invisibly loads more of the conversation as the user scrolls down the page, in much the same way as Twitter and other sites do. Individually, neither of these techniques are particularly new on their own, but combining them seamlessly was technically satisfying. I'll talk more about these techniques in a dedicated future post.

Hosting

At present, both the conversation program and the Rails application are hosted at Heroku. A future post will detail a few of the technical subtleties I discovered while deploying the live Betty and George website.

Now visit Betty and George!

OK, so this was just a brief overview of the Betty and George architecture - now go check them out, and let me know what you think in the comments below!