How To Zen Your Python -- experimentation and detection code

Repository structure

Source selection

• The original sources used by Alexandru et. al are found in sources_original_repaired.txt. These were used in Experiment 1 (the replication experiment) Some of the repos are removed from this list as they were removed, either due to a DMCA takedown or by the author themselves. These are listed below.

• Our "new" sources for Experiment 2, which we found by re-running src/main/bash/get-projects.bash, can be found in sources_new.txt. You can gather your own sources by running /src/main/bash/get-projects.bash <ouputfilename>.

• The list of repositories and commit hashes we used for Experiment 3 can be found in the sources_commit_20xx_xx.txt files

Scripts

• The original scripts Alexandru et al. wrote are also included in this repositories
• The new scripts we wrote are:
  • src/main/python/aggregate.py -- this was slightly overhauled to fix LaTeX support, and also generate tables from multiple sources and automatically add columns in one go. Instructions in Aggregating results section.
  • src/main/python/random-selection-projects.py -- This computes the overlap between the original sources and the "new" sources we collected, and displays 10 of them randomly. Can simply be run using python3 src/main/python/random-selection-projects.py, outputs to standard output.
  • src/main/python/get-commits.py -- This script uses the GitHub API to find, given a list of Git repository sources and chosen time periods, commit hashes for each time period for each of the repositories. This was used for Experiment 3. The list of repositories was taken from random-selection-projects.py and you can change the list on line 6 of get-commits.py to change this list. You can also change the ranges of dates to use on line 17. After that, running python3 src/main/python/get-commits.py will create the .txt files needed with the sources and commit hashes.
  • src/main/python/repair-sources.py -- The list of repositories the original authors gave did not correspond to the repositories of the resulting data in ./data-original. We tried to reconstruct the list just using the folder names using this script. Run using python3 src/main/python/repair-sources.py and it will save the repaired sources list to sources_original_repaired.txt.
  • ./aggregate-timeperiods.ipynb was used to generate graphs using the data discovered in Experiment 3 in data-commits/. This can be run using Jupyter notebook.

The remaining script, i.e project-stats.bash were from the authors of the orignal paper and is not used in this project.

Detectors

• The original detector is in src/main/scala/CrawlerOrig.scala. This was unmodified from the replication package.
• The new detector, which makes use of the enhanced idiom detector analyses, is in src/main/scala/CrawlerNew.scala.
• The new detector with the new Git Agent, which can additionally check out commit hashes after cloning repos, can be found in src/main/scala/CrawlerNewCommit.scala.
• The new analyses and Git Agent are in src/main/scala/slim.

NOTE: These scripts may take several hours to run if using all ~1,000 sources. You can terminate early at any time with CTRL+C because stats are saved after detection on each project is finished, not at the end. And if restarting the script, it will pick up where you left off.

If you want to test on a small selection of sources, sources_sample.txt contains 10 repositories, so modifying src/main/scala/CrawlerNew.scala to use this text file would terminate after a few minutes.

Alternatively, you can add a new .txt file with a subset of the sources and modify the Source.fromFile() function call in src/main/scala/CrawlerNew.scala to use your new file.

Results

• The results from Alexandru et. al are found in ./data-original
• The results from the re-run using the same sources from "today" are in ./data-rerun-repaired
• The results from the "new" sources are in ./data-new
• The results from the commit experiment are in ./data-commits

Memory requirements

Adjust the settings in .sbtopts according to how much memory you can spare

Running analyses

Changing the source file (list of repositories to use as sources) can be done by modifying the projects variable in the appropriate Crawler file in src/main/scala/

To run the analyses:

sbt 'runMain <crawler to use> ./<destination folder to store result CSVs>'

A description of each crawler is given above in the Detectors section.

Example:

sbt 'runMain CrawlerNew ./data-new'

As the script runs, you will see in the output the progress of the analyses: cloning the git repository, analyzing the files, and then it will move on to the next project.

Aggregating results

This will aggregate the results from the CSV files in ./data-new and output to console

./src/main/python/aggregate.py -d ./data-new --stdout 

You can also generate a LaTeX table:

./src/main/python/aggregate.py -d ./data-new --tex 

Or combine multiple sources of data into one table:

./src/main/python/aggregate.py -d ./data-original -d ./data-rerun-repaired -d ./data-new --tex 

Reproducing each of the tables in paper

Note that calculations for percentage difference and grouping were done by hand, so these results will not be included in the outputs.

• Table 2: ./src/main/python/aggregate.py -d ./data-original -d ./data-rerun-repaired --tex
• Table 3: ./src/main/python/aggregate.py -d ./data-rerun-repaired -d ./data-new --tex

Removed sources

The below repositories were removed from lists due to errors such as:

• The author deleted the repository/changed the name or username, so the link no longer works.
• Ran out of memory while trying to run the detector

Repaired sources list

git://github.com/AKSHAYUBHAT/DeepVideoAnalytics.git git://github.com/Russell91/pythonpy.git git://github.com/SirCmpwn/evilpass.git git://github.com/alex/django-ta.git git://github.com/dalocean/netbox.git git://github.com/instabot-py/instabot.py.git git://github.com/kamyu104/LeetCode.git git://github.com/kemayo/sublime-text.git git://github.com/kennethreitz/l.git git://github.com/livid/v2ex.git git://github.com/xiyouMc/WebHubBot.git

Incorrect "original" sources list

git://github.com/chrismsimpson/Metropolis.git git://github.com/dmulholland/ivy.git git://github.com/kamyu104/LeetCode.git git://github.com/livid/v2ex.git git://github.com/m4ll0k/WAScan.gitgit://github.com/Azure/azure-sdk-for-python.git git://github.com/mwhite/resume.git git://github.com/threerocks/studyFiles.git

"new" sources

git://github.com/owid/covid-19-data.git git://github.com/Azure/azure-sdk-for-python.git