The PlateWash analysis repo.
Matlab TeX Makefile Shell M Awk
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
.lyx
NewtonianSimulation
autocollimator
bin
bootstrap
calibration
conclusion
extracted
glib
globalConfig
headers
ifo
metrology
mlib
runAnalysis
runConfig
simulatedData
systematics
thesis
tmp
.gitignore
LICENSE
Makefile
Makefile.inc
README.md
SIBugs.txt
bugs
bugs.fods
debianPackagesThesisBuild.txt

README.md

PlateWash

This repository contains the code needed to generate the PlateWash data analysis and Charlie Hagedorn's thesis.

To use this code: clone the repository. Get the data (<= 46 GB) and store it in ~/PWData/ (or alter the repos's root Makefile to point at it). Our group has not yet authorized the release of our data into the wild; if we make it publicly accessible, this file will be updated. Interested parties should contact Charlie Hagedorn ( cah49@uw.edu ) or Jens Gundlach ( gundlach@npl.washington.edu ).

Once installation is complete, executing 'make' in the project directory will do exactly that. The thesis will appear in thesis/thesis.pdf . Execution time for the unblinded code (on an Intel(R) Core(TM) i5-3330 CPU, 8GB RAM) is less than nine days. Bootstrapped simulated annealing is slow. The rest of the code executes in a few hours.

The unblinding was performed with the commit tagged "UnblindedWithThisCommit". Several important updates to prose were made before the unblinding, tagged with "LastCommitBeforePublicUnblinding".

This should execute on any linux box running GNU octave (3.8.2) with octave-forge, GNU make (4.0), LyX (2.1.2), GNU awk (4.1.1), GNU sed (4.2.2), gnuplot (4.6 patch 6), dia (0.97), ImageMagick (6.8.9-9), git (2.1.4), and dpkg (1.17.25). The octave-forge installation may need to be a Debian one (see path curiosities in initializeOctave.). The reference system is Debian Jessie, which went stable the day after unblinding began. The forthcoming thesis will contain a complete accounting of every package on the unblinding computer.

Because paths are delicately configured, the only correct way to execute code is through make, not with octave/gnuplot. Using make .INTERACT will yield a persistent Octave shell (see Makefile.inc) for debugging. Make is called recursively throughout, so it cannot compute the full dependency graph. The only reliable way to be certain that the result is built as specified is to build it from a clean repository. We welcome your suggestions for a more-usable and bulletproof build arrangment.

This code is not beautiful, but it works. Constructive criticism is welcome, especially if you have a patch/pull request or implementation suggestion.

Please consult the LICENSE file for details regarding code reuse.