The question was a simple one: given a string (e.g.) “examplesgnome” consisting of two substrings (in this case, “examples” and “gnome”), how can we swap the two substrings in-place? In the interview, we covered the third method in more detail, but touched upon the other two. Afterwards, I calculated their computational efficiency, and determined while they all have the same upper bound, one has a lower average running time.

Method one: reversal

This is one Dr. Kelly suggested, which is really rather neat. It consists of two steps: reversing the entire string, then reversing both substrings. For example:

1. examplesgnome
2. emongselpmaxe
3. gnomeexamples

If we take to equal the total number of characters, to equal the number of characters in the smaller substring, and to equal the number of characters in the larger substring, this has:

swaps, so we may as well say it’s of linear complexity .

Method two: cycles

This one was also suggested by Dr. Kelly, and involves taking the first character of the larger substring out to a temporary location, then cyclically moving characters places along into the generated space. Once the strings have been swapped ( swaps have taken place), we put the stored character back in the final space. For example:

1. examplesgnome
2. _xamplesgnome
3. gxamples_nome
4. gxa_plesmnome
5. gxamplesmno_e
6. gxampl_smnoee
7. g_amplxsmnoee
8. gnamplxsm_oee
9. gnam_lxsmpoee
10. gnamelxsmpoe_
11. gnamelx_mpoes
12. gn_melxampoes
13. gnomelxamp_es
14. gnome_xamples
15. gnomeexamples

Since each character moves to its final position in one move, the algorithm is intuitively of complexity — but we can go further and say it’s of absolute complexity , taking the steps to remove the first character and fill the last space into account.

Method three: reduction

This is the method I came up with during the interview and, with some help, made work. It basically involves swapping the first smaller substring into its final position, then applying the same algorithm to the resulting subsubstrings of the second substring. For example:

1. examplesgnome
2. gnomelesexamp
3. gnomeampexles
4. gnomeexpamles
5. gnomeexmaples
6. gnomeexamples

This is quite a neat one to implement, and will perform swaps in the worst case (when the smaller substring is only one character long, as one character moves to its final position with each swap).

However, what if we consider the case “foobar”, where the substrings are “foo” and “bar”? In this case, method three only takes 3 swaps. Obviously, this method is limited in the upper bound to swaps, but it seems it can do the operation in far fewer — as few as .

I haven’t got a concrete proof for this, but I believe this is actually of complexity:

as the lengths of the substring for each iteration depend on the remainder of for the th iteration. This complexity can be manipulated to prove that it’s also of complexity :

All-in-all, I’m quite pleased with the interview: even if I’m not successful, it’s given me an interesting problem to consider. Corrections and further thoughts on the problem are welcomed.

It seems to be release week for me. Here’s Almanah Diary 0.5.0. Of note with this release is the support for text formatting (bold, italic and underline — nothing particularly fancy), persistent window dimensions, the ability to set the encryption key, and better recovery from database corruption/errors (although it’s still not perfect).

Packagers should note that the GTK+ dependency has been bumped from 2.12 to 2.14, and the GConf dependency is no longer optional — though the GtkSpell dependency now is (as spell checking support is now optional).

I’ve now also completed the application name migration to “Almanah Diary”, and as such, the source is now at http://svn.gnome.org/svn/almanah/, and localisation statistics are at http://l10n.gnome.org/module/almanah/.

Plans for future versions include:

• Proper accessibility support
• Automatic link types (yes, this is a stupid name), so that (for example) past Evolution appointments would be shown for each day
• Documentation
• Export to blog, text file, HTML, etc.
• Undo/Redo support

The first three are what I’d like to definitely get done for 0.6.0, but before I can do that I need to sort out the naming for “links” and “automatic links”. Obviously these are rubbish names, especially considering one can currently add a “note link”. I had considered “attachments”, but that’s not much better. Does anyone have any ideas?

Motörhead themselves were absolutely amazing, playing a variety of stuff: new, old and classic. Mikkey did a fairly hefty drum solo (easily the loudest part of a concert from which my right ear still hasn’t stopped ringing), and they managed to fit an acoustic number in, with Lemmy playing the harmonica. I’m told by a friend (whose uncle works for the venue) that Motörhead reached 118dB before the encore, which is not bad at all.

It’s already had three releases (this is the fourth), but I didn’t want to make the 0.1.x releases too public, as they were mainly for testing the program at school. The 0.2.0 release was a brown paper bag release, due to a bad GTK+ version dependency, and some files necessary for the Windows zip package not being included.

It’s available on the MCUS project page, with 0.2.1 being the latest release. A Windows installer is available too, built with NSIS; as well as a simple zip package of the installed directory tree on Windows, since the program’s relocatable.

My school is already using it, and I’m releasing it in the hope of other schools being able to make use of it. As always, it’s open source, released under the GPLv3+.

Future plans include a graph to visualise the ADC waveform, and the addition of more simulated input and output hardware.

I got back from Chile three days ago, and now I’m sufficiently caught-up with the goings-on of the past four weeks, it’s time to summarise what I did on this World Challenge expedition.

The expedition was split into four different phases: acclimatisation, project, main trek, and rest and relaxation. After flying into Santiago (via Paris), we did our acclimatisation in La Campana National Park, climbing the eponymous mountain and camping for three nights. Climbing La Campana was tough, and in hindsight, probably tougher than the main trek up Guane Guane.

Our project phase was in Iquique, where we re-painted an old people’s home, then part of the school in which we stayed. It was quite fulfilling, but somewhat longer than it could’ve been; we spent several days wandering around Iquique relaxing, when we could’ve instead had a longer main trek.

The main trek was in Lauca National Park at high altitude; we had to spend several days acclimatising to 3500masl in Putre before we could ascend to our base camp at Parinacota (4500masl). At Parinacota we had to spend more time acclimatising, before we attempted (and succeeded) a climb up Guane Guane. The views up on the altiplano were fantastic, and I’ve got (perhaps too?) many photos of llamas, alpacas and vicuñas.

Our rest and relaxation phase was a little short, but we managed to pack in most of the things San Pedro de Atacama offers, including sandboarding, a trip round the Valley of the Moon and a trip to see the El Tatio geysers.

I would definitely recommend this to anyone who has the time, money and driving force. Some of us are already thinking of going back to have an attempt on Mount Parinacota!

I’m slowly putting photos up on Picasa before I go off on holiday again for a week on Monday.

I won’t be around to see them get into the archives, but Stefan Ebner has been working on packaging both Diary (now called “Almanah” due to Fedora issues) and Hitori for Ubuntu. My thanks to him for his hard work on that, and patience in helping me fix all the nasty little autofoo problems.

I’ll try to make an Almanah (Diary) 0.4 release before I go, but I want to leave as long as possible for translations to catch up with the several string changes.
Almanah 0.4.0 is out! The main change for this release is to rename it and fix some problems preventing it being packaged, and so it should be compatible with old databases. The build requirements haven’t changed.

As a short overview, the new version introduces:

• Dynamic board sizing (allowing for boards from 5×5 to 10×10)
• Rendering improvements, a Tangoisation, and a conversion to use GtkStyle
• Relicense to GPLv3+
• A new user manual!
• Improvements to the algorithms, halving memory usage in the best case
• A new icon by Jakub Szypulka
• A desktop file

Hitori now depends on GTK+ >= 2.13, so it may not be easily compilable on some distributions, which is an unfortunate side-effect of using the new gtk_show_uri function.

As before, please contact me to report bugs.

I don’t really have any plans to extend the program any further, but if anybody has any ideas, I’m open to them.

There are no dependency or config changes with this release, so upgrading should be seamless. Any bugs should be reported to the Diary product on GNOME Bugzilla.

Update: I’ve just released 0.3.1, which is a semi-brown-paper-bag release to fix build failure when compiling without encryption support, as well as fix the desktop file translation.