Thesis

Originally, Florian Sihler created flowR as a part of his master's thesis, to be found at: http://dx.doi.org/10.18725/OPARU-50107.

This page explains how to reproduce (and replicate) the results from the thesis. Follow it with v1.0.0, the submission state, which is the release the stats command below belongs to.

• How to Reproduce the Statistics From the Master’s Thesis

Each step assumes, that you start in the root directory of this repository. You need a working installation of R and npm.

This mainly describes how to extract the statistics from the CRAN package sources, however, starting from step 3, the steps are basically the same and only differ in the paths that have to be supplied (the social science sources are attached alongside the master's thesis release).

1. If you want to update the set of packages, use the scripts/top-downloads.R script and potentially modify the package limit by setting N to a different value. The (sorted) results should be versioned and can be found in scripts/top-r-downloads.txt.

   cd scripts/ && Rscript top-downloads.R

2. If you haven't done so already, or updated the package list in the previous step, download the package sources. For this, you can use the scripts/download-top-pkg-sources.R script. But make sure, that you set the to variable to the output path you want.

    cd scripts/ && Rscript download-top-pkg-sources.R

   Downloading and extracting the sources can take a while.

3. Make sure you have the latest version of the flowr package installed.

   npm ci

4. Now you can run the statistics program on the downloaded sources. You can do this in two ways (run npm run stats -- --help from the cli directory for more information). In any case, the extraction may take a long time, so be prepared for that! Furthermore, you may want to store the output of the tool as it provides additional information.

   • On the complete folder

     First, make sure you are currently in the cli directory to allow running command-line interface commands.

     npm run stats -- --input <location-of-source folders> --output-dir <output-dir>

     If you left the to variable in the previous step at its default value, you may want something like this:

     npm run stats -- --input "${HOME}/r-pkg-sources/" --output-dir "./statistics-out/cran-500"

   • On a folder subset
     You may very well have downloaded all or more package sources than you want to analyze.

   Theoretically, you should be able to stop the extraction at any time and still get usable information with the next step, of course limited to only those files that have been processed so far.

5. Afterward, your output folder should contain several folders with the recorded stats of all extracted features. To make sense of them, you need to use the command-line interface's post-processor, which prints the summarized information to the command-line:

   npm run stats -- --post-process "./statistics-out/cran-500" --no-ansi > "./statistics-out/cran-500/cran-500-summary.log"

   Additionally, the post-processor will create .dat files for several (sub-)features which contains histogram information. Depending on the size of the sources, you may want to

   • increase the heap-size of node (export NODE_OPTIONS=--max_old_space_size=8192).
   • limit the features to be analyzed by using the --features option.

   By default, the post-processing will limit the histograms to the top 50 values, because who needs more histograms?!