A Princess theory for computing Parikh images of automata, and a tool to use it.
Warning: this is an active research project. It is not currently intended for use by anyone except me!
Note: this is built using nightly Princess.
$ sbt assembly
./bin/catra solve-satisfy --timeout 300000 experiments
Some UNIX-y path expansion works (e.g. ~/), but it is difficult to get right in Java and I suggest you use as non-fancy paths and glob patterns as possible.
There is also a built-in experiment runner that can be built and executed like so:
$ sbt benchmark/assembly
$ java -jar benchmark/target/scala-2.13/catra-benchmark-assembly-$VERSION.jar file1 file2
The evaluator will evaluate a number of configurations across all instances, in parallel. It is highly recommended to restart the JVM every once in a while. To do that, you can use xargs
:
$ find basket -type f | xargs -n 10 java -jar runner/target/scala-2.13/catra-benchmark-assembly-$VERSION.jar
This is mainly intended for artefact evaluation, and is not meant for end users, hence the sharp edges and nonexistent configuration options.
Of course, you can also run the experiments directly from sbt (quoting the arguments because SBT itself is not so much a few sharp edges as a shuriken):
$ sbt "benchmark/run basket" | tee results.log
Similar to the benchmark runner, there is also a validator that will take a list of (directories of) inputs and validate any results with nuXmv. You can access it using:
$ sbt validator/assembly
$ java -jar validator/target/scala-2.13/catra-validate-assembly-$VERSION.jar
Tests are available using the regular sbt test
. To measure coverage using
sbt-coverage, use sbt clean coverage test
followed by sbt coverageReport
. Your report is now in
target/scala-2.*/scoverage-report/index.html
.