bruceauyeung / github-contrib-stats Goto Github PK

for example, a github project may have thousands of close PRs. retrieving them over and over again will result in performance penalty. but cosidering closed PRs can be reopened, i have no idea how we can cache them.

http://stackalytics.com also do statistics analysis about Commits in Kubernetes. but stackalytics count commits incorrectly(most of time more than actual value).
we should add a config stackalyticsDeviation for each user , then we can render a table view of stackalytics version(incorrect version)

currently only overall statistics is implemented.we shoud add statistics per week /per month

error: GET https://api.github.com/search/issues?page=1&per_page=100&q=beeb65b4eafc2bf78349aaf1035401be24b87f5f+repo%3Akubernetes%2Fkubernetes.github.io+type%3Ap
r+author%3Axiangpengzhao: 403 API rate limit exceeded for bruceauyeung.; rate reset in -21.3340042s
error: GET https://api.github.com/search/issues?page=1&per_page=100&q=beeb65b4eafc2bf78349aaf1035401be24b87f5f+repo%3Akubernetes%2Fkubernetes.github.io+type%3Ap
r+author%3Axiangpengzhao: 403 API rate limit exceeded for bruceauyeung.; rate reset in -21.6160042s
error: GET https://api.github.com/search/issues?page=1&per_page=100&q=facd9bcce7aa046b3dd6baa326ba292fedf9b7b3+repo%3Akubernetes%2Fkubernetes.github.io+type%3Ap
r+author%3AYuPengZTE: 403 API rate limit exceeded for bruceauyeung.; rate reset in -23.5020042s

there is a bug waiting for fixed

if a man import some codes from one repository to another, like this pr:
kubernetes/community /pull/118 ( i don't want make a link points to that pr , it will notice the participants in that pr )
the pr imported some docs form repo kubernetes/kubernetes to repo kubernetes/community with commit logs
so a person who commit in repo kubernetes/kubernetes now will have a copy in kubernetes/community too(but different SHA, i.e. you have tow commits are totally same but different SHA in two repos). in this case , the author and the committer of these two commits are same, so here comes the problem.
the scheme that we find a pr form a commit is that :

1. seaching a pr from commit's SHA with repos and author name specified , if not found , it may be the case that the committer and the author are not the same, go step 2
2. seaching a pr from commit's SHA, but now change author name to committer .
   but according this scheme can not pr in the situation i mention above, so i think we should add a step 3 that do not specified the author arg

i also doubt whether it is reduntant that specified author arg when searching pr. may be specified repos arg is enough. if returns many, we could filter them by author's name.

@bruceauyeung

currently only numbers of merged / lgtmed / nonlgtmed PRs of individual users are presented in the final statistics table.

currently we use stackalyticsDeviation mechanism. but this is inherent incorrect. we should list commits on a repository ( https://developer.github.com/v3/repos/commits/#list-commits-on-a-repository ) and filter them by user, date period, etc.

@xialonglee

merged column should look like this:

       merged
--------------------
PRs    |   Commits
--------------------
x      |      y
--------------------

panic: Get https://api.github.com/repos/kubernetes/test-infra/commits?author=StudyNick&per_page=100: net/http: TLS handshake timeout

goroutine 1 [running]:
panic(0x672940, 0xc047f83260)
D:/Go/src/runtime/panic.go:500 +0x1af
/d/GOPATH/INTERNAL/src/github.com/bruceauyeung/github-contrib-stats/githubstat.getStackalyticsCommits(0xc0420e0ea0, 0xc0422d0100, 0xa, 0xc0422d00d0, 0xa, 0xc0
42039f60, 0x9, 0x0, 0x0, 0x0)
d:/GOPATH/INTERNAL/src/github.com/bruceauyeung/github-contrib-stats/githubstat/metrics_commit_request.go:129 +0x598
/d/GOPATH/INTERNAL/src/github.com/bruceauyeung/github-contrib-stats/githubstat.(*PullRequestMetricsRequest).FetchMetrics(0xc042064058, 0xc0420e8780, 0x0)
d:/GOPATH/INTERNAL/src/github.com/bruceauyeung/github-contrib-stats/githubstat/metrics_pull_request.go:486 +0x5f3
main.main()
d:/GOPATH/INTERNAL/src/github.com/bruceauyeung/github-contrib-stats/main.go:50 +0x418
exit status 2

listing pull requests takes lots of time. we should consider adding progress bar while doing statistics.
considering that we have a config statBeginTime to avoid listing all pull requests, we can not design a progress bar at pull request page level. but we can design it at pull request level.
a rough thought:

( ( number of  finished open pull requests ) + ( number of  finished closed pull requests ) ) / ( ( number of  total open pull requests ) + ( number of  total closed pull requests ) )

( ( number of total open pull requests ) + ( number of total closed pull requests ) ) actually is number of repositories * 2

for example, a PR needs rebase, but operator rebased it and merged it in github website, this will result in a extra commit in that PR. this kind of commit usually look like "Merge branch 'master' into ***".
we should filter out this kind of PR

displaying github-username ( real name ) in USERNAME column looks good enough.