check-in on release steps

[paciorek]

@paul-vdb based on my partial reconstruction from my notes of our September call, I think you had the following on your plate to prepare for release:

• further quadrature grid testing of sparsity/pruning
• adding testing to test-nested-fit based on examples from your talk; are you still thinking of doing this? Alternatively if you have the example(s) set up, I am happy to take that an turn into tests.
• look through nimbleQuad repo issues and close if appropriate
• vignette (I think the thought was to use material from "quadrature pieces" and the Overleaf tech document
• report skew to users (issue report skew to users in nested approx #30 )

I am basically ready to start working through final release steps. But I can wait depending on your timeline and how critical we think the above items are.

[paul-vdb]

@paciorek This week is hectic, but I can make next week a priority for final steps. I'm keen to work on the vignette but if you have time to get that started that would be great and I can work off your template.

The skew comment is a good one still. I can resolve that today or tomorrow.

I did add the quadrature tests after that convo, but I haven't done those tests at the application level. I'll add pruning and sparsity to the test-nested-fit level when I add my salamander example.

Thanks for following up.

[paul-vdb]

I also have a note to read through normTooling.R which I can do this week but guessing how things are I assume still passing tests I'm confident this will be minor.

[paciorek]

Ok, thanks. I can continue to tee things up so that after you have some time to work on this next week, we can release quickly.

Regarding vignette, I already have the penicillin example and some overall description in the manual in the nimble repo. So I'm not sure what we want in the vignette. Did you have an idea of a different example (or multiple examples) that would go in there? And were you planning to draw from the ApproximatePosteriorsNimble.tex Overleaf doc to give more technical detail than I put in the manual? (Also I have some note about a "quadrature pieces" document, but not sure if that is the same as ApproximatePosteriorsNimble.tex.

No need to respond right away; just noting my uncertainty.

[paul-vdb]

@paciorek Just looping back through this. I had a look at the vignette. I think it's fine to just have the one example.

Things I will do by end of week:

1. Work through Issues and close those that are no longer relevant.
2. Read through NormTooling.R (said I would ages ago and making certain)
3. Start a local vignette that is similar to the one from the manual with a focus on customizing the quadrature grid etc. That will have piece by piece examples of control from both runNestedApprox and buldNestedApprox (we can release before this is finished so last thing I'll work on).
4. Tidy up testing environment.

Target release next week... fingers crossed :)

[paciorek]

@paul-vdb can you check off the tickboxes in the first item in this thread to make clear what has been done and what we are leaving for future work.

Also, I've merged everything onto devel. Can you run all tests under Windows locally on your machine (ideally on Monday if that's not a big lift). I added a bit of code so that if you set runFailingWindowsTests=TRUE in your R session, you should be able to source the test files (or run testthat::test_file (or perhaps testthat::test_package) and get all expectations/tests to execute.

[paul-vdb]

@paciorek The only thing I haven't completed that I think would be good is to add tests for the different quadrature grids into the marginals for actual data. This would be the first check box. I have the code and can make a test this morning if you would like but also don't think it's the end of the world if you want to keep devel locked down.

[paciorek]

Please go ahead and add it.

[paul-vdb]

@paciorek done and pushed to devel. Made a minor issue about one thing I noticed that I think is fine to ignore for release if we want to for now.

Changed the marginal param grid type for a 3d example and added a check with pruning for latent nodes.

Will run all tests in windows next. I'm assuming just the batch run in windows which has the skips for some things that we will ignore for now?

[paciorek]

Sorry, I can't quite parse "I'm assuming just the batch run in windows which has the skips for some things that we will ignore for now?". I'd like you to run all the tests, including the skipped ones. This might require running tests individually, so if it turns out to be too much of a hassle, let me know.

[paul-vdb]

@paciorek Roger that. I will get that running in the background and will let you know when complete.

[paul-vdb]

@paciorek I have not finished working through all the tests but did identify a problem that I've flagged as a new issue. Will continue to run tests tomorrow.

[paul-vdb]

@paciorek All tests passed on my local windows machine, with just the problem of setting up a model twice ending up crashing my Windows machine. Not sure if that is a nimbleQuad problem or a deeper problem with how NIMBLE is doing things.