Tools for Reproducibly running the mrc-ide covid sim
The world has ended, everything is terrible, and some aspect of it is related to a bit of code. Which code? This code: https://github.com/mrc-ide/covid-sim - henceforth "mics" for short.
For various reasons I'm interested in the operation of mics and honestly as a programmer and not an epidemiologist I find a lot of its construction and documentation confusing. At the same time I realised a few weeks ago that I miss Pascal in a way that's uncomfortable. Slowly the two ideas gelled together in my head - I could use Pascal as a tool to help investigate the behaviour of mics. It's genius because Pascal should disuade anyone from using my code, right?
Disclaimer - this is the first large Pascal project I've done in 20 years. The code is probably bad.
I'm stuck indoors, obtuse and grumpy and this is what happens.
Well coRvid is my set of tools for launching the mics code reproducibly with input files in the form of good old traditional .ini files. I'll probaby also write some tools for comparing outputs of the simulations because these are very interestingly tab-separated files with a .xls extension (for reasons) and the current regression testing regime checksums said files and therefore tells you nothing about the degree of error.
What the name tells you is I really wanted to name it after a crow.
This code requires FreePascal 3 (ish), and is definitely intentionally and definitely not due to incompetence/lack of caring only going to work on a Unix-like system due to path separators.
Check out the repo and then:
cd src
make
make install DESTDIR=/some/install/path
The binaries are static so you do not need the FreePascal compiler in your environment to use it.
(yeah - .exe, deal with it)
This is main executable which launches simulations. To use it, you need to write a .ini file - you can see examples in input/. For reasons I don't understand, mics appears to have to operational modes - one where it creates a subset of input files and runs a simulation and one which re-uses the input files from the earlier run. In the Python wrapper it first does this setup run with a "no intervention" set of constraints and then does a second run with a set of constraints. These correspond to input/uk-sim-initial.ini and input/uk-sim.ini respectively.
You will need to copy the data directory from mics somewhere. On UCL resources this along with the code is provided in a module file so we'll assume this configuration, and assume coRvid.exe is in your $PATH.
For versions before 0.9.0, there are separate binaries for UK and US simulations so .ini files for those have also been provided.
On a UCL system, in a directory with your .ini files in it:
# Setup modules
module purge
module load rcps-core compilers/intel/2020 python3/3.7 covid-19-spatial-sim/0.8.0
# Get a copy of the data, and un-gzip the gzipped files
cp -R $SPATIALSIM_DATA .
cd data/populations
for a in *.gz
do
gzip -d $a
done
cd ../..
# Run the zero intervention sim
coRvid.exe uk-sim-initial.ini
# Create a copy of the data for the next sim
cp -R output output-1
# Run the sim with the test intervention
coRvid.exe uk-sim.iniThis program uses the FreePascal random function to set the four random seeds in a simulation .ini file between 0 and maxint and write the results out to a new file:
coRvidReseed.exe <input file> <output file>
This application compares two TSV format files, such as the .xls files created by mics. It writes the comparison file to either CompareTSV.out or the third argument it is passed:
CompareTSV.exe <file1> <file2> <output file>
The output file is also in TSV format, with cells that contain:
| Case | Value |
|---|---|
| numerical | cell 2 - cell 1 |
| == strings | contents of cell |
| ¬= strings | cell 1 << >> cell 2 |
In addition, it returns 0 if the files match, 1 if they don't to make it scriptable.
There is a very alpha level GUI for editing (and directly launching) .ini files in src/gui. This can be built with Lazarus, and is not built as part of the rest of the tools above.
