Changelog¶

0.10.0 - April 27th 2020¶

Created all of the extract framework to support customising the output and analysis during a run.
Created a better Workspace class for holding accumulated data during extract
Completed most of the extractor tutorial
Added in WardInfo(s) to get metadata about wards, and to support searching for wards via their name, code, authority and region

Merged in latest changes from the C code. Now gives complete agreement, including via a custom iterator that repeats the lockdown model.
Support x/y and lat/lon coordinates and distances. Now works properly with the 2011UK model data
Added an example of a lockdown parameter set scan

Automated github actions for building a versioned website plus automating building the packages.
Switched default for UV parameter to 0.0, as this should not normally be 1.0
Added custom user variables both for scanning and to act as inputs that may be used by custom advance and iterate functions. Detailed tutorial now shows how these can be used to model a lockdown.
Improved speed of custom iterators

Lots of progress with the project website, including a detailed tutorial
Support fully customisable disease models, and can adjust any disease parameter using a more flexible input file format
Can record the date in a model run, plus set the starting day and date
Broken up the iterate function into metawards.iterators, and can now have the user create their own custom iterators. Tutorial on how to do this will appear soon.
Broken up the extract_data function into metawards.extractors, and will soon enable a user to create their own. Tutorial on how to do this will appear soon.
Added metawards-plot to create simple plots and animations. This is particularly useful when working through the tutorial.
General code cleaning, documentation improvements and nice-to-haves that make the code easier to use.

Wrote an initial draft of the complete project website
Fixed packaging problems that prevented installation of older packages on some systems

Support running multiple model runs in serial or in parallel
Support aggregation and writing of model multiple model run outputs to the same directory, including to a single shared CSV data file.
Support for parallel running via multiprocessing, mpi4py or scoop

Parallelisation of individual model runs using OpenMP
Parallel code scales to large numbers of cores and can complete individual runs in 10-15 seconds.

Adding in a simple profiler to support optimisation of the code
Replaced GSL random number generator with a more liberally licensed and easily bundled generator extracted from numpy.
Switched code to the https://github.com/metawards organisation
Optimised more using cython and raw C for file reading
Added automatic versioning of packages and files using versioneer
Cleaned up the repository and added status badges

Cythonizing the bottleneck code to bring the python code up to a comparable performance as the original C code.
Added in packaging information and general repository and file cleaning.

Fully working Python port of the original C code that completely reproduces the results of the C code when given the same random number seed. However, it is significantly slower! Python port has promise, so worth exploring different options for speeding the code up.

Imported code from https://github.com/ldanon/metawards and began thinking about what the code was and trying to understand it. Decided to write a port as I find that if I can translate something, then I can understand it.