Open emoszkowski opened 7 years ago
To clarify, this is on master before opening the gensys-type
branch.
Do the tests pass on the gensys-type branch?
Also, I agree with your comment on moving test data out of the source files. Because these are just numerical arrays, I can't decide if it would be best to add a heavy dependency like HDF5/JLD, or if we could get away with just reading in a csv/JSON/YAML version of each of the matrices.
I think JSON.jl is a much lighter dependency than HDF5/JLD
The tests don't pass on gensys-type
either, but they fail the same way as they do on master. I'm happy to convert the data to JSON or csv format there or on a separate branch, but I'm worried about writing tests to pass rather than doing a rigorous test. I might try with the DSGE.jl test data directly and see how that works.
That's a really good point -- we don't want to just get tests that pass.
Do you have good tests for the discrete time functionality in DSGE.jl?
Yup, we do. We got it by making sure we get the same results as Chris Sims's MATLAB code.
Hi,
Thanks for the great work.
I have also noticed that tests fail on julia v0.6:
Test Failed
Expression: sumabs2(G1 .- G1_solution) <= 1.0e-5
Evaluated: 0.0034548960384160276 <= 1.0e-5
ERROR: LoadError: There was an error during testing
while loading /home/julien/.julia/v0.6/Gensys/test/runtests.jl, in expression starting on line 497
Same question as above: is that the expected behavior ?
The test for
gensysct
is failing. Is that expected behavior or is something going wrong on my end?Also, it might be better practice to either test on a smaller system or to move the giant arrays into an HDF5/JLD file, rather than to have large, unwieldy arrays hard-coded into the tests.