Will you come back to contribute? Investigating the inactivity of OSS developers in GitHub

Setup

Use the productivity branch for the latest updates.

Add to the root a folder named Resources/ with the following files:

• repositories.txt containing the list of projects (one per line) to be analyzed, in the following format org/repo_name (e.g., `atom/atom);
• tokens.txt (optional) containing the list of GH tokens to be used;

Sampling of developers

Core Developers Selection

Refer to this README.md file.

Truck-Factor Developer Selection

Refer to this README.md file.

---

CommitExtractor.py

Params

Uses the tokens defined in Resources/tokens.txt and the list of repository urls in Resources/repositories.txt, as defined in the Settings.py file.

• None.

Requirements

• Set files and folders names in the Settings.py file

Execution

python CommitExtractor.py

Output

• logs/Commit_Extraction_organization.log: log file
• Organizations/<organization>/[<repo1>...<repoN>]/: Results folders
• For each repo folder:
  • commit_list.csv: List of the commits in the format: <SHA; author_id; date>
  • commit_history_table.csv: Matrix of autors and dates. The cells contain the number of the commits of a developer in one day
  • pauses_duration_list.csv: List of pauses durations in days for each developer in the format: <dev; listOfDurations>
  • pauses_dates_list.csv: List of pauses dates for each developer in the format: <dev; listOfPauseDates>
• The same files are given after merging the commits of every organization's repo in the Organizations/<organization>/ folder.

if you came here from point 2 of core selection you can now perform step 3 following (CoreSelection | Step 3)

---

ActivitiesExtractor.py

Params

• None

Requirements

• Set files and folders names in the Settings.py file

Execution

python ActivitiesExtractor.py

Output

• logs/Commit_Extraction_organization.log: log file
• Organizations/<organization>/[<repo1>...<repoN>]/Other_Activities/: Results folders
• For each repo folder:
  • issues_comments_repo.csv: List of the issue comments in the format: <id; date; creator_login>
  • issues_events_repo.csv: List of the issue events in the format: <id; date; creator_login>
  • issues_prs_repo.csv: List of the issue and pull request creations in the format: <id; date; creator_login>
  • pulls_comments_repo.csv: List of the pull request comments in the format: <id; date; creator_login>

PullRequestExtractor.py

NonMergedCommitsExtractor.py

MissingStuffCollector.py

CodingTableBuilder.py

---

BreaksIdentification.py

Params

• mode: enter one of following modes ['tf', 'a80', 'a80mod', 'a80api']

Requirements

• Set files and folders names in the Settings.py file
• Insert the list of the TF/core developers (<TF_developers_file>) in the right folder. Formatted as a list of <name;login>. The path to save the file is set in the Settings.py file.
• Set the window size and the shift size in the Settings.py file

Execution

python BreaksIdentification.py tf | a80 | a80mod | a80api

Output

• logs/Breaks_Identification.log: log file
• Organizations/<organization>/Dev_Breaks/: Results folders
• For each developer in the TF file:
  • <devLogin>_breaks.csv: List of the breaks in the format: <len; dates; Tfov_used>

Algorithm

Let D be a developer to analyze and let life(D) be the number of days between its first and last commits. For each sliding window W in life(D) which slides of shift days. The values of variables window (default 90 days) and shift (default 7 days) are set in the Settings.py file).

The goal is to select all the breaks (pauses that are larger than usual) associated with the Tfov (Far-out-value threshold) of the first window where they have been found:

1. PAUSES SELECTION STEP

• In the list win_pauses, put all the pauses within W (only these pauses define the rythm of D in W).
• In the list partially_included, put all the pauses partially within W (i.e., pauses that start in W and end in the next window).

2. Tfov DEFINITION STEP

• If win_pauses contains >=4 pauses then the W is valid, then use win_pauses to calculate Tfov. If Tfov is valid (i.e., IQR>1), then proceed to the breaks identification step (go to STEP 3).

• Else, when win_pauses < 4 (i.e., Tfov cannot be calculated) or if Tfov is invalid (i.e., IQR<=1) for W, then:

  • If a previous Tfov exists, then consider it as the current Tfov and proceed to the next step for breaks identification (go to STEP 3).
  • Otherwise, save into the list clear_breaks all the pauses from partially_included that are larger than the window size and have not been considered yet, ignore the other pauses in win_pauses; move forward W by shift days and RESTART (go back to STEP 1).

  (Note: The pauses that are larger than shift days will be considered in the next W and so on, whereas the smaller ones are not breaks and can be safely ignored).

3. BREAKS IDENTIFICATION STEP

• Select as break each couple <p, t> from the lists win_pauses and partially_included where t is Tfov and p is a pause > Tfov.
  • Move forward W by shift days and RESTART (go back to STEP 1).

4. FINAL STEP (When there are no more W)

• Compute Avg_Tfov as the average of all the valid Tfovs found.
• Save the pauses in the list clear_breaks as breaks (<p, t> where t is Avg_Tfov, and p is a pause > Avg_Tfov as for list definition).

---

BreaksLabeling.py

Params

• mode: choose one of following modes ['tf', 'a80', 'a80mod', 'a80api']

Requirements

• Make sure to have already executed the BreaksIdentification.py script to get the <devLogin>_breaks.csv files (one for each developer).

Execution

python BreaksLabeling.py tf | a80 | a80mod | a80api

Output

• logs/Breaks_Labeling.log: events log file
• Organizations/<organization>/Dev_Breaks/: Results folders
• For each developer in the TF file:
  • <devLogin>_labeled_breaks.csv: List of the breaks in the format: <len; dates; Tfov_used; label; previously>

Algorithm

1. Get a break from the Breaks list.

2. If there is not any other activity performed by the developer during the break, then label it INACTIVE if < 365 days; GONE otherwise.

3. If there are other activities in the period:

• Define sub_breaks_list as the list of the intervals between such activities (sub_break).
• Identify each sub_break > Tfov from the sub_breaks_list and label it based on the defined state diagram (∆t_inactive = ∆t_non-coding = Tfov).