crop-testing

This is the testing repository for https://github.com/jonasbovyn/comunica-crop (the CROP implementation is found in the feat/crop branch). To generate the WebAssembly binary, check the https://github.com/jonasbovyn/crop-cpp repository.

Setting up the testing environment

Preparing the datasets

1. Under the datasets/ folder, 3 folders are located which must be filled with data generated from https://github.com/comunica/watdiv-docker.
   1. Under datasets/dataset100k use -s 1 -q 5 as command line arguments which will result in around 100k triples.
   2. Under datasets/dataset1M use -s 10 -q 5 which will result in around 1M triples.
   3. Under datasets/dataset100k use -s 10 -q 5 which will result in around 10M triples.
2. Each folder datasets/X/ should now have a subfolder queries/ and a file dataset.nt.
3. In addition, use the https://github.com/rdfhdt/hdt-cpp tool to generate dataset.hdt files from the dataset.nt files.
   1. The command is ./rdf2hdt dataset.nt dataset.hdt

Installing the testing environment

1. Use yarn install in the project directory
2. Create a soft link in node-modules/ pointing to the custom Comunica repository, named @comunica-crop/.
   1. In linux/maxOS, the command for this is ln -s [source] [destination]
   2. Make sure Comunica is built and checked out in the right branch. Use the mem-benchmark branch when testing optimization memory usage, otherwise use the feat/crop branch.

Running the tests

If tests require a dataset to be deployed using Server.js, first run node runServer.js. The other test files (benchmark<XXX>.js) each have their own purpose and configuration, and if they require additional command line arguments, this is specified in the file. (we used node v16.13.0)

Additionally, the data can be analyzed in results/, which has a python notebook (python v3.10 was used)