by Sebastian Witte

Fun with mpd

If you're not familiar with mpd: it is a music player daemon called Music Player Daemon. I have been using it for years and today, a Debian update destroyed my setup. At first I was wondering why there was no music showing up in ncmpcpp, which is my client of choice, and started investigating. A short search for mpd in htop immediately revealed to me, that some rogue user, whose name is mpd, now runs mpd on the port that I used to run it on as myself. Well, Debian probably enabled that service for me in some mysterious way. (This is actually one of the few things I do not like about Debian.) Well, a quick

systemctl disable mpd
systemctl stop mpd

should disable it, right? -- WRONG!

Warning: Stopping mpd.service, but it can still be activated by:

At least it told me that. Anyway, what should I do about it? Remove the socket file? Change the port of my usual mpd instance? To have some music playing right away, I did the latter first and immediately realized that I do not want to start ncmpcpp with a port parameter all the time. Also, it is quite nice to have a socket activated mpd.

My first shot at solving this problem

My first intuition was to do this:

gpoupadd mpd
gpassw -a saep mpd
gpasswd -a mpd mpd
chown -R saep:mpd ~/Music

and change the music_directory entry in /etc/mpd.conf to my home music folder and now I have a socket activated mpd running.

The actual solution

However, while writing this post I realized a few things. Nobody but me has access to the laptop. Also, nobody but me in my family would listen to that music and additionally, I would have to fix permissions for every new file I add there. That stinks and is overly complicated. So, I can just symlink my usual ~/.mpd/mpd.conf to /etc/mpd.conf and be done with it.

rm /etc/mpd.conf
ln -s ~saep/.mpd/mpd/conf /etc/mpd.conf
systemctl restart mpd


by Sebastian Witte

Building a prototype

This is the second post describing my progress on this project. I'm actually enjoying this experience more than I anticipated. While starting to move things around and creating the first set of modules, I actually started to change some of the general structure of the project.

I changed my mind

I don't insist on limiting myself to a git repository as I found the filestore library. Now git, darcs and mercurial can be chosen as blog entry backends. Well, at least the brief look at the documentation showed that it provides me with the information I need. If someone wants to implement it, other repository types could work, too. The library provides a convenient interface to retrieve the version information I need, so it seems to be a perfect fit for the job.

I also ditched the notion of a required repository layout. As those version control systems can handle the number of files I'm expecting in a somewhat wort-case analysis quite well, I will just scan the whole repository for supported files on startup. This should suffice for this early development stage.

I played around with CSS

I also started working on my personal style sheet. It's definitely not done, but it renders my previous post pretty enough for a first draft. I even had to google around for hacks to make the navigation bar span the whole display with while keeping the content smaller. I already start to fear the days when I want to make this application look good on my mobile phone... I think I'm getting carried away now.

The project structure -- Part 2

I now have essentially three modules. One is just a nice wrapper with some extensible data types to store configuration data. Another one is the entry serving the blog, which also has some configuration data structures as well as some drafty data types to do filtered searches on blog content. I think that I'm getting carried away again... The third module is an HTML template for the blog part of this web application. There is also the executable, but I may some day replace it with a dyre replacement. But for now

cabal run saeplog --entry-path "~/git/myblog" --resources "~/git/blog/resources

suffices for my development needs.

Next steps

I hope to get a reasonably structured project up and running very soon, so that I can upload a working prototype on GitHub.

by Sebastian Witte

Writing this blog

This blog post is the first that I actually do and it is about writing the blog software that I am (hopefully) publishing this post from. I did try to do this some time ago, but I never really got around to actually do it.

This post is structured in a few different sections. First I will describe my intentions for writing this piece of software. Then I will describe the rationales behind the basic tools and libraries that I chose. In the passing years, these might change, though. Then I will describe the basic project structure that I am planning to use.


I like using my good old vim for text editing and I really hate those limited browser-based editing windows. I realize that I do not gain a lot by using vim for a task such as writing a blog post, but I just want to do it that way! I also want to strife away from my part-time day job for which I mainly write Java code and I think that I miss these cute Compile-time errors that free me from a lot of debugging. The only things that I currently do in Haskell are the occasional project Euler riddles and some work for my master thesis. A major part of my motivation is to try out some of the Haskell libraries for everyday tasks, such as JSON-processing, file conversion, web development, testing and project design. To combine all these, I need a non-trivial extensible project and that is what a web presentation, which might initially just be a blog, can fulfill.

The workflow of writing blog posts should be kept quite simple. I want to write (pandoc's) markdown or eventually literate Haskell into a simple text file and commit it to a git repository. I kind of did such a year or two back with pyblosxom, but I wasn't really happy with it.

Another motivation for writing blog posts is my recent transition to the Colemak keyboard layout, which I might blog about if I've gotten used to it a bit more.

Tools and libraries

The web framework

Initially, I wanted to use the Yesod Web Framework as it seemed to be high-quality piece of software with incredible benchmarks and it includes all the bells and whistles you might need. There are however a few key moments that kept me away from using it. At two points in time somewhere in the last 4 years, I was tempted to write a web application and tried using Yesod. I was quite inexperienced and I couldn't get it to install the first time within an hour and then I gave up. The second time, I was able to set it up, but the book diverged to far from the library and it was to annoying to synchronize that. Hence I gave up after an hour again. The urge to actually create a web application war probably not big enough.

The other alternatives that I looked at were Snap and Happstack. Snap kind of scared me away with yet another piece of a web-specific language, namely Heist.

Happstack seemed to be the simpler one to learn and so I chose it. The site seemed eledant and yet simple and the tutorial link thew tiny numbers at me:


A good starting point is the happstack-lite tutorial. It's less than 2000 words long, yet contains everything you need to create a full-featured website: [..]

Only 2000 words! That is marketing I like. And another important appraisal was this sentence from the title page:

If you are looking for a lightweight, simple solution we recommend starting with happstack-lite. happstack-lite provides everything you need to implement a web application with out relying on template haskell, external preprocessors, or complex types. It is possible to seamlessly transition into from happstack-lite into the full happstack ecosystem -- so you are not limiting yourself by starting with happstack-lite.

If the lite version isn't enough, I can just upgrade. That seems to be great.

The file conversion

Pandoc has become the tool of choice concerning document conversion for me. It supports a lot of typical markup languages as well as literal Haskell to quickly create documents in the format I want. I used Emacs with the fabulous orgmode which is a great piece of software and it was second in the tools to use, but as I do not use the org-mode anymore, especially since I moved back to vim, it fell to the second place. I do not (yet) have any use cases for the advanced features of orgmode except for throwaway TODO-lists whose elements are not worthy to be raised in the existence of a ticket in a ticket system like Redmine. For these, a small markdown list is sufficient enough anyway.

The persistence

For the blog entries I plan to use a simple Git repository with a specific layout. The top-level folders are the years in which I wrote the entries. In these folders are all entries of the year. I don't expect to write a lot of these, so indexing should be almost instantaneously. I first thought about doing a full folder hierarchy down to days, but that seems to be overkill for the occasional blog post that I will probably write. As I sometimes remove typos or fix broken links in documentation at work on a very regular basis, using the timestamps of last edit and such are not really useful, so I need some metadata somehow. As I started this section, I thought about generating a JSON metadata file together with the content file, but as I am already using git to store the files, I can also just use the metadata stored within the repository. It at some point, I want to support comments, I should probably create a folder for each blog post I make. Alternatively, I could use a data base for that. I think I will go for databases as it is another technology to test out in Haskell and it is still a feature I haven't planned yet. As nobody may be reading these posts anyway, a manual email is probably a reasonable starting point.

The things I did not think about

Yeah, I don't think about a lot of things. Whole libraries are actually filled with it! ;-)

Project structure

I suppose the term modular should be sufficient to describe the project structure. As I'm testing out various libraries, I should be able to simply replace a module and be done with it. In this case, I might at some point start to write factor things out into separate cabal modules. The whole project should be usable as a library in the end, so I will have to think about the exported functions and decide how configurable the software should be. As blogging should just be a starting point for this project, every component should essentially be a composable object and hence a configurable library function or module. To start of, I will hence need a Template module for the blog entries and its metadata that will be inserted into a highly personal general web layout. As I am just starting and nothing is release-worthy, I will keep everything within one cabal project for.