I just uploaded GillespieSSA 0.5-4 to CRAN. It should be just a matter of days before it has propagated itself across all mirrors.
This release consists of minor revisions with no (intended) changes in functionality. The main change (and it is also minor) is a fix for warning messages in recent versions of R. I owe a thanks to Thomas Petzoldt who took upon him to do the actual leg work on this one. My only contribution was to delay the fix for so long that CRAN dropped the package and moved it to the morgue. Reports of the death of GillespieSSA were, however, grossly exaggerated and now that the package passess all the checks again it will hopefully appear on CRAN in due time. Note that these days the development of the package (albeit slow) is taking place on GitHub, where the most recent version is also available.
Interestingly the very same day CRAN relegated the package to their formerly available packages list I started to receive emails from distressed users from all parts of the world. I was quite taken back that so many people were interested in the package (which of course is not the same as people actually using it), and once Thomas (out of the blue) fixed the warning issue (which was the cause of the relegation in the first place) I truly started feeling guilty.
Just to address a some of the questions I received:
- Has the package been removed from CRAN because there are problems with it?
The reason the package was removed from CRAN was because warning messages arose during checks under recent versions of R. These messages were related to the documentation. Currently there are no reported issues/bugs (which likely reflects poor usage rather than actual lack of bugs).
- Has the licence been changed?
Nope, the package is still under GPL 3.
- What is the best way to run a large number of realizations with GillespieSSA? Is there something more elegant than calling ‘ssa’ a thousand times?
Nope, that’s it.
- It seems that GillespieSSA is not part of the CRAN distribution anymore. Does it mean that it will not be developed further?
Currently the package is implemented entirely in R (that’s right) and as such there are no major further developments planed in terms of adding functionality or other types of stochastic simulation algorithms. In my opinion the main functionality currently missing is the ability to run multiple realizations and in particular summarizing the results. Due to lack of time it is unlikely that I will actually get around adding this. Volunteer contributions/involvement are encouraged.
The main direction the I am planing to develop the package in is to replace the actual SSA R code with a C++ implementation. The plan is to integrate the package with StochKit, a state of the art stochastic simulation framework developed in C++. Again, due to time constraints, this will likely take some time. I encourage contributions/involvement by anyone willing and able. Once this is done, however, the package is likely to blow your socks off. Preliminary benchmarking suggest that one might see up to 3 orders of magnitude improvement of in speed, i.e. what now takes hrs to run would be a matter of seconds.
- Would it be possible to add the
SSA method to the package?
In principle yes, if someone volunteers to do the actual coding (I might be able to help incorporating it into the code base). Now that the code repository for the package is available on GitHub anyone can download it and create their own branch which can be merged back into the original repository. If you are not familiar with amazingly cool concept of distributed SCM, check out this great talk by the inventor of Git, Linus himself.
Feedback, bug reports, ideas, code contributions are most welcome and will move things forward more expeditiously.
This is from the “Mario’s Entangled Bank” blog (http://pineda-krch.com) of Mario Pineda-Krch, a theoretical biologist at the University of Alberta.
Filed under: computer simulations, GillespieSSA, Git, Linus Torvalds, R