Credit

credit is a fast tool for measuring Github repository contributions and the overall health of a project.

Use credit to find out:

• Who has the most Pull Requests merged to a project.
• Who engages in the most discussion in Issues and PRs.
• How long it takes maintainers to respond to and solve Issues.
• How long it takes to get PRs merged.
• If a library would be a safe long-term (i.e. maintained) dependency.

Installation

Arch Linux

With an AUR-compatible package manager like aura:

sudo aura -A credit-bin

Cargo

cargo install credit

Usage

To use credit, you'll need a Github Personal Access Token with public_repo permissions. See here for an additional example.

💡 Note: credit calls the GraphQL-based Github v4 API, which has a much higher rate limit than the REST-based v3 API. This allows credit to run quickly and work on projects with a long development history.

You can use credit limit to check your current API query allowance.

Markdown Output

By default, credit outputs text to stdout that can be piped into a .md file and displayed as you wish:

> credit repo --token=<token> rust-lang/rustfmt

# Project Report for rustfmt

## Issues

2462 issues found, 2189 of which are now closed (88.9%).

- 1899 (77.1%) of these received a response.
- 1553 (63.1%) have an official response from a repo Owner or organization Member.

Response Times (any):
- Median: 10 hours
- Average: 34 days

Response Times (official):
- Median: 13 hours
- Average: 39 days

## Pull Requests

1821 Pull Requests found, 1650 of which are now merged (90.6%).
168 have been closed without merging (9.2%).

- 1505 (82.6%) of these received a response.
- 1379 (75.7%) have an official response from a repo Owner or organization Member.

Response Times (any):
- Median: 8 hours
- Average: 2 days

Response Times (official):
- Median: 12 hours
- Average: 2 days

Time-to-Merge:
- Median: 17 hours
- Average: 3 days

## Contributors

Top 10 Commentors (Issues and PRs):
1. nrc: 2772
2. topecongiro: 1526
3. marcusklaas: 718
4. calebcartwright: 461
5. scampi: 331
6. kamalmarhubi: 120
7. rchaser53: 103
8. cassiersg: 100
9. gnzlbg: 79
10. otavio: 63

Top 10 Code Contributors (by merged PRs):
1. topecongiro: 513
2. marcusklaas: 125
3. calebcartwright: 74
4. nrc: 72
5. scampi: 64
6. rchaser53: 57
7. davidalber: 34
8. kamalmarhubi: 31
9. ayazhafiz: 28
10. sinkuu: 24

JSON Output

You can also output the raw results as --json, which could then be piped to tools like jq or manipulated as you wish:

> credit repo --token=<token> rust-lang/rustfmt --json

Large Projects

By default, credit queries for Issues and Pull Requests at the same time, which is fast and works well for most projects. For very large projects, however, this can make the Github API unhappy.

If you notice credit failing on projects ones with many thousands of Issues and Pull Requests, consider the --serial flag. This will pull Issues first, and then Pull Requests. --serial allows credit to even work on the Rust compiler itself!

> credit repo --token=<token> rust-lang/rust --serial

Caveats

Accuracy

The numbers given by credit are not perfect measures of developer productivity nor maintainer responsiveness. Please use its results in good faith.

Response Times: Particularly in the Open Source world, volunteer developers are under no obligation to respond in a time frame that is most convenient for us the users.

Merged PRs: Without human eyes to judge a code contribution, its importance can be difficult to measure. Some PRs are long, but do little. Some PRs are only a single commit, but save the company. credit takes the stance that, over time, with a large enough sample size, general trends of "who's doing the work" will emerge. Expect weird results for one-man projects or projects that otherwise have a long history of pushing directly to master without using PRs.

Why not use commit counts instead of PRs?

Per-user commit counts are already available on Github.

Median vs Mean

You may notice that sometimes the reported Median and Average results can be wildly different. Given the presence of outliers in a data set, it can sometimes be more accurate to consider the Median and not the Mean.

In the case of maintainer response times, consider a developer who usually responds to all new Issues within 10 minutes. Then he goes on vacation, and misses a few until his return 2 weeks later. His Average would be skewed in this case, but the Median would remain accurate.

credit doesn't attempt to remove outliers, but might in the future.