VisionEval Next Generation Beta Release 0.5: Visualize your Visions

[jrawbits]

I've just updated the VisionEval "NextGen" beta release (new installers, new code) based on the beta-release-0.4 tag.

Key features here include some small fixes to the build system, bug fixes for locate employment and for the multi-modal module, and a fully functional implementation of category-based scenarios (like the old VEScenario package) that works all the way through to popping up (or saving as a directory) the R visualizer. The visualizer is dyanamic: you have complete control of the categories (and input files) that build the scenarios, as well as complete control over what metrics are included.

The VEModel test script (which is what I use to determine if the build is "working") will run in the installer runtime. Just launch your runtime VisionEval, then source("tools/tests/VEModel/test.R"). It will give you a list of tests, and you can look at the script source code to see what they do (yes, it remains a bit cryptic). For fun, do this two-step process:

test_install("VERSPM","scenarios-cat")
test_visual()

If all goes well, it will generate 8 scenarios from a couple of category combinations (you can uncomment more of those in the model's "scenarios/visioneval.cnf" and eventually do the full 600+ scenario set), run all those scenarios (in a single thread, see below), run queries that duplicate the original VERSPM metrics from VEScenario, and then pop up a live visualizer in your default web browser.

To check out multiprocessing, you could do the visualization manually:

model <- installModel("VERSPM","scenarios-cat") # this will create a new copy of the model.
model$plan() # enable multiprocessing using all available cores; or model$plan(workers=3) to set a specific number
model$run() # the base stages will run one at a time, then all the defined scenario alternatives will run in parallel
query <- model$query( model$query() ) # Load the first (by name) query defined for the model. The sample models have one query.
query$run()   # Run the query on all the model result scenarios
query$extract() # Optional; See the query results as a data.frame
query$visual() # Pop up a visualizer using the query's results.

Have fun. In help deciding if your problem or issue is with the VisionEval code or the (lack of) documentation, you can email [email protected]. Reproducible issues can be added to this repository. We will be cleaning up the issues over the next month or so - some the ones open now have been fixed.

Have fun!

[jrawbits]

Beta has been updated to version 0.5, fixing a significant bug preventing the binary installer from working. The current installers for beta 0.5 work. The open pull request for the new test-architecture will be merged shortly.