We are reaching limits of what is feasible with imperative languages and we should move to declarative languages.
When applications written in imperative languages grow, the code becomes convoluted. Why? Imperatively programmed applications contain statements such as if X do Y else do Z. As Y and Z contain invisible side-effects the correctness of the program relies on some implicit invariant. This invariant has to be maintained by the programmer or else the code will break. Thus each time a new feature is added to an application or a bug is fixed the code for the application gets more complex as keeping the invariant intact becomes harder. After a while the code becomes spaghetti-code and bugs are introduced as the programmer fails to maintain the invariant. This is going to happen despite the best intentions of the programmer to keep things clean. Why is this?
People tend to only understand what they can see. For most people it is
difficult to grasp more abstract matters without somehow visualizing them.
Software is an example of such an abstract matter. Let us visit the process of
developing software through a comparison with developing a building.
With this script it should be possible to install and build the JCU app in a
local directory. It does not build the UHC for you. If that is wanted the
option could be build in of course!
For my Experimentation Project at Utrecht University I ported the “JCU”
application to Haskell. The JCU application is used to give Dutch High school
students the opportunity to taste Prolog.
The project uses the Utrecht Haskell Compiler and its JavaScript backend. The UHC
translates Haskell to Core and then translates this Core language to JavaScript.
For more information on this see the blog of the creator of the
UHC JavaScript backend.
Please read my report on this
project. The project is hosted on GitHub in the following repositories:
update 28-01-2012:
The keyword jscript in the UHC has been changed to js in order to avoid
association with Microsoft’s JScript.
Also new Object syntax is now available in the foreign import directives.
Today I’ve launched my new blog. It is based on Octopress
and works by statically generating the pages and then syncing them with the
server.
If you are for example on OS X Lion and installed XCode 4.2 and you run into
weird errors like a missing gcc-4.2, and Homebrew throws errors like this:
Error: The linking step did not complete successfully
The formula built, but is not symlinked into /usr/local
Please install the gcc package from this nice fellow:
osx-gcc-installer
And if you are getting nagged by rb-fsevent. Change
Gemfile
1
gem'rb-fsevent'
to
Gemfile
1
gem'rb-fsevent',"0.9.0.pre4"
Update
The comments have been exported with the Wordpress plugin to Disqus. I’m
currently looking at how to highlight code within Disqus comments.
On several occasions I noticed that when performing a cabal update that the
index was being downloaded at the rate of plus min 300 KB/s. Finally I got around to do
something about this. I’ve set up a caching server located in Utrecht, The
Netherlands. It is a caching proxy for the hackage repository. If you want to
use it, add the following to your ~/.cabal/config file. (Or equivalent on
Windows.)
Be sure to comment out the already existing remote-repo. Otherwise, cabal will
download both indexes and merge them, and we don’t want this.
The funny bit
Apparently this only helps if your machine is fast enough to process the index
(untarring and all extra administration cabal performs).
Plainly getting the file from the cache:
1234
> wget http://spockz.nl:12080/packages/archive/00-index.tar.gz
Saving to: 00-index.tar.gz
100%[======================================>] 3.394.821 10,9M/s in 0,3s
2011-11-13 00:12:35 (10,9 MB/s) - 00-index.tar.gz saved [3394821/3394821]
And running cabal update with my cache as source:
1
cabal update -v3 7,50s user 0,21s system 99% cpu 7,736 total
And then finally, with the original repository:
1
cabal update -v3 7,57s user 0,25s system 28% cpu 27,372 total
So here we see that the user time is roughly the same but you spent almost three
times more seconds waiting for your coffee to get cold. Any further speed
improvements for cabal update will probably require optimalisation of the code.
The caching server
I’m using Varnish
to cache the request to hackage. And here is my config file. Please shoot if you
see any improvements.
I’ve just released a new version of lhs2TeX-hl: 1.4.5 and it includes some new features:
Added support for recursively traversing includes of .lhs files
The program now doesn’t fail completely when haskell-src-exts fails to parse
a file. An error is reported and the program continues. :) A fmt file is
still generated.
functions that aren’t functions but constants are now given the tag `constant’
Removed a faulty command from the list.
Cleaned up some code. (Probably introduced other ugly code)
Binary operators are now typeset better. (I hope :))
Note the installing of librsvg twice, this is in accordance with https://github.com/mxcl/homebrew/issues/4970. If you don’t get an error after running the first install of librsvg you might have a new update brew of librsvg.
People have asked me how to launch a program through the terminal in the
background but with it’s own terminal screen similar to the way the “start”
command in the Windows CMD prompt works. Well here it is. Add the following two
files (start.applescript) and start to your path, and make them executable
(chmod +x {start.applescript,start}).
One point of notice is to always encapsulate strings with
spaces that should be one parameter in quotes.