Justin Lanfranchi

Make sure this is good reasonable and fast. Issues: - Don't want to re-read entire files if they have already been read in. - Filename and/or file path is not...

Test that source code hashing works

2

This is critical for transforms that get cached to disk, so we know that they're still valid.

disk caching very large items fails

1

Probably a limitation of sqlite3; possibly need to use filesystem. Workaround for now is to try/except to fail gracefully, and simply move on as if the item wasn't in the...

pickling very large maps fails

2

Not sure if this is fix-able. Pickle seems like a bad way to store really large maps (e.g. HDF5 would make more sense). But it might be a bug... Alternately,...

All objects should be able to be written to a human-readable ASCII format (e.g. JSON) and also to a more compact binary representation (e.g. binary pickle files, or maybe simply...

Undocumented dependencies in `stages.unfold.roounfold`

1

Fails at ```python from ROOT import RooUnfoldResponse, RooUnfoldBayes ``` Add info to `INSTALL.md` about optional dependencies (and which version(s) are required); add note in `roounfold.py` about the dependencies. Include any...

`root_numpy` dependency not documented, not easy to install

2

Also, apparently specific version(s) of ROOT and/or numpy might be required to get `root_numpy` to install. Please add info on which version(s) are "legal" to `INSTALL.md` and make some note...

Test only runs on CPU; if done so with fp64 it succeeds; if fp32, we get ``` $ export PISA_TARGET=cpu ; export PISA_FTYPE=fp32 ; python ~/src/pisa/pisa/stages/osc/prob3numba/numba_osc.py > Traceback (most recent...

Full build log is here for reference: [1_build (3.6).txt](https://github.com/IceCubeOpenSource/pisa/files/4351429/1_build.3.6.txt), which you can see directly [here](https://github.com/IceCubeOpenSource/pisa/pull/585/checks) General issues: - [x] `matplotlib.use(backend)` warnings everywhere (I'll look into this, might not be too...