Hi there! This repository contains the data and code for Team Discussions and Dynamics During DevOps Tool Adoptionsin OSS Projects (ASE 2020).

If you find this work useful in your research, please consider citing:

@inproceedings{yin2020tool,
Author={Likang Yin and Vladimir Filkov},
Booktitle={2020 35th IEEE/ACM International Conference on Automated Software Engineering (ASE)}, 
Title={Team Discussions and Dynamics During DevOps Tool Adoptions in OSS Projects}, 
Year={2020},
Pages={697-708}
}

• The adoption data (i.e., GitHub Badges) is stored in adoption_data.csv file (under the main folder), the json version adoption data (which is much tighter) is stored in data/tool_adoption_dict.json.

• Comments data (with sentiment) is stored in data/comments_with_sentiment.json file.

• Exposure data is stored in data/author_knowledge_dict.json.

• Commits are stored in the relative_sentiment_devs/merged_dict.json.

• Github monthly events data from year 2011 to year 2018 is stored in folder data/events.

• Sentiment data per project is stored under folder relative_sentiment_devs/final_combinations.

We use Senti4SD as for sentimental prediction.

You need to install Git LFS extension to install Senti4SD locally. Once installed and initialized Git LFS, simply run:

$ git lfs clone https://github.com/collab-uniba/Senti4SD.git

To reproduce the results in the paper, you can use code and data in the following specific folder.

• To obtain the curves of distribution of the adopted datas of tools and trends of various number of developers:

  • Run the jupyter file developers_trends/plot_curves.ipynb cell by cell. The jupyter file is dependent on the data in developers_trends/data folder.

• To get the relative negativity of developers:

  • run python relative_sentiment_devs/negativity_new_and_senior.py, which gives the negativity of new and senior developers per month.
  • Then run python relative_sentiment_devs/negativity_un_exposed.py to obtain the negativity of developers with exposure v.s. developers without exposure.
  • Lastly, run Rscript relative_sentiment_devs/plot_negativity.R to generate the curve plot.

• To reproduce the relative negativity of comparable tools

  • Run python negativity_on_categories.py
  • Then run Rscript plot.R to generate the negativity plots in each tool category.

• To build our glmer models for adoption success and discussion length.

  • Run python glmer_model/final_table.py to generate the aggregated final table for regressions,
  • Then run Rscript glmer_model/glmer_regression_models.R to obtain the two models with R^2 scores.