June | 2013 | Computational Semigroup Theory

2013 GAP-SEMIGROUPS-SGPDEC (rainy) Summer Code Camp

June 15, 2013 egri-nagy 1 comment

Hatfield again, same people. Still BIOMICS funded. Many things happened at high speed. Both on the software and on the mathware side. Also, antisocial geek behaviour reached new heights, for instance sitting down at the dinner table, opening the laptop and telling the others “You may talk to me, but anything not code and math-related will be ignored…”

So, what is the progress?

We did beta testing of James’ development branch of GAP. The novelty here is that calculating with transformations now done in the kernel, which means SPEED!!! increase in every functions using transformations. Of course, having stuff in the kernel implies chasing nasty segmentation faults… but there was a happy end. So the new code is on the way to be included in the upcoming GAP 4.7.

Hand in hand with the new GAP release the semigroup packages are also updated. Citrus got renamed (again), the package is called Semigroups now, less fancy name, but more telltale. Saving future disappointments of orange farmers from Florida. Smallsemi was also hit by the wind of change, duplicate definitions had to be removed.

Following all these upstream changes, SgpDec also got its fair share of commits. Good things: the 0.7 development branch now has all the features of the old 0.6 branch. Everything is back, including holonomy decomposition. Bad thing: except some comments in the testfiles the package is completely devoid of any documentation. Something that will soon be rectified.

Talking of holonomy. We still believe in a short and elementary description of the proof. Working backwards from the working code, this will be part of the documentation.

We also agreed on how to make the one-line notation for finite transformations more compact. It is now called the compact notation.

Regarding the extreme programming part, the 8.5 hours long coding session in a cafe in Hatfield’s Galleria was… hm.. long. At the end of the day we cleaned up the datatypes in SgpDec. So it was very successful, discounting those lost three hours. The trouble came from the combination of these two mistakes:

1. I kept the test files separately, so James could not use them.
2. James continued where he finished in February, but “change as many things as possible at the same time” works fine in the beginning of the development, when everything is broken, but not when the codebase already has some consistency.

Reconciliation was successful. There are now proper test files for SgpDec and we also found those two swapped arguments that caused havoc everywhere.

Some photos:

Vertical screen space is important. We like WQHD monitors.