GitHub.jl provides a Julia interface to the GitHub API v3. Using GitHub.jl, you can do things like:
- query for basic repository, organization, and user information
- programmatically take user-level actions (e.g. starring a repository, commenting on an issue, etc.)
- set up listeners that can detect and respond to repository events
- create and retrieve commit statuses (i.e. report CI pending/failure/success statuses to GitHub)
Here's a table of contents for this rather lengthy README:
GitHub's JSON responses are parsed and returned to the caller as types of the form G<:GitHub.GitHubType
. Here's some useful information about these types:
-
All fields are
Nullable
. -
Field names generally match the corresponding field in GitHub's JSON representation (the exception is
"type"
, which has the corresponding field nametyp
to avoid the obvious language conflict). -
GitHubType
s can be passed as arguments to API methods in place of (and in combination with) regular identifying properties. For example,create_status(repo, commit)
could be called as:create_status(::GitHub.Repo, ::GitHub.Commit)
create_status(::GitHub.Repo, ::AbstractString)
where the second argument is the SHAcreate_status(::AbstractString, ::GitHub.Commit)
where the first argument is the full qualified repo namecreate_status(::AbstractString, ::AbstractString)
where the first argument is the repo name, and the second is the SHA
Here's a table that matches up the provided GitHubType
s with their corresponding API documentation, as well as alternative identifying values:
type | alternative identifying property | link(s) to documentation |
---|---|---|
Owner |
login, e.g. "octocat" |
organizations, users |
Repo |
full_name, e.g. "JuliaWeb/GitHub.jl" |
repositories |
Commit |
sha, e.g. "d069993b320c57b2ba27336406f6ec3a9ae39375" |
repository commits |
Branch |
name, e.g. master |
repository branches |
Content |
path, e.g. "src/owners/owners.jl" |
repository contents |
Comment |
id, e.g. 162224613 |
commit comments, issue comments, PR review comments |
Status |
id, e.g. 366961773 |
commit statuses |
PullRequest |
number, e.g. 44 |
pull requests |
Issue |
number, e.g. 31 |
issues |
Team |
id, e.g. 1 |
teams |
Gist |
id, e.g. 0bace7cc774df4b3a4b0ee9aaa271ef6 |
gists |
Review |
id, e.g. 1 |
reviews |
You can inspect which fields are available for a type G<:GitHubType
by calling fieldnames(G)
.
GitHub.jl implements a bunch of methods that make REST requests to GitHub's API. The below sections list these methods (note that a return type of Tuple{Vector{T}, Dict}
means the result is paginated).
method | return type | documentation |
---|---|---|
owner(owner[, isorg = false]) |
Owner |
get owner as a user or organization |
orgs(owner) |
Tuple{Vector{Owner}, Dict} |
get the owner 's organizations |
followers(owner) |
Tuple{Vector{Owner}, Dict} |
get the owner 's followers |
following(owner) |
Tuple{Vector{Owner}, Dict} |
get the users followed by owner |
repos(owner[, isorg = false]) |
Tuple{Vector{Repo}, Dict} |
get the owner 's repositories/get an organization's repositories |
teams(owner) |
Tuple{Vector{Team}, Dict} |
get the organizations 's teams repositories |
method | return type | documentation |
---|---|---|
members(team) |
Tuple{Vector{Owner}, Dict} |
get team members as users |
method | return type | documentation |
---|---|---|
repo(repo) |
Repo |
get repo |
create_fork(repo) |
Repo |
create a fork of repo |
forks(repo) |
Tuple{Vector{Repo}, Dict} |
get repo 's forks |
contributors(repo) |
Dict |
get repo 's contributors |
collaborators(repo) |
Tuple{Vector{Owner}, Dict} |
get repo 's collaborators |
iscollaborator(repo, user) |
Bool |
check if user is a collaborator on repo |
add_collaborator(repo, user) |
HttpCommon.Response |
add user as a collaborator to repo |
remove_collaborator(repo, user) |
HttpCommon.Response |
remove user as a collaborator from repo |
stats(repo, stat[, attempts = 3]) |
HttpCommon.Response |
get information on stat (e.g. "contributors", "code_frequency", "commit_activity", etc.) |
commit(repo, sha) |
Commit |
get the commit specified by sha |
commits(repo) |
Tuple{Vector{Commit}, Dict} |
get repo 's commits |
branch(repo, branch) |
Branch |
get the branch specified by branch |
branches(repo) |
Tuple{Vector{Branch}, Dict} |
get repo 's branches |
file(repo, path) |
Content |
get the file specified by path |
directory(repo, path) |
Tuple{Vector{Content}, Dict} |
get the contents of the directory specified by path |
create_file(repo, path) |
Dict |
create a file at path in repo |
update_file(repo, path) |
Dict |
update a file at path in repo |
delete_file(repo, path) |
Dict |
delete a file at path in repo |
permalink(content::Content, commit) |
HttpCommon.URI |
get a permalink for content at the SHA specified by commit |
readme(repo) |
Content |
get repo 's README |
create_status(repo, sha) |
Status |
create a status for the commit specified by sha |
statuses(repo, ref) |
Tuple{Vector{Status}, Dict} |
get the statuses posted to ref |
status(repo, ref) |
Status |
get the combined status for ref |
method | return type | documentation |
---|---|---|
pull_request(repo, pr) |
PullRequest |
get the pull request specified by pr |
pull_requests(repo) |
Tuple{Vector{PullRequest}, Dict} |
get repo 's pull requests |
issue(repo, issue) |
Issue |
get the issue specified by issue |
issues(repo) |
Tuple{Vector{Issue}, Dict} |
get repo 's issues |
create_issue(repo) |
Issue |
create an issue in repo |
edit_issue(repo, issue) |
Issue |
edit issue in repo |
reviews(repo, pr) |
Tuple{Vector{PullRequest}, Dict} |
get a pr 's reviews |
dismiss_review(repo, review) |
HttpCommon.Response |
dismiss review in repo |
method | return type | documentation |
---|---|---|
comment(repo, comment, :issue) |
Comment |
get an issue comment from repo |
comment(repo, comment, :pr) |
Comment |
get a PR comment from repo |
comment(repo, comment, :review) |
Comment |
get an review comment from repo |
comment(repo, comment, :commit) |
Comment |
get a commit comment from repo |
comments(repo, issue, :issue) |
Tuple{Vector{Comment}, Dict} |
get the comments on issue in repo |
comments(repo, pr, :pr) |
Tuple{Vector{Comment}, Dict} |
get the comments on pr in repo |
comments(repo, pr, :review) |
Tuple{Vector{Comment}, Dict} |
get the review comments on pr in repo |
comments(repo, commit, :commit) |
Tuple{Vector{Comment}, Dict} |
get the comments on commit in repo |
create_comment(repo, issue, :issue) |
Comment |
create a comment on issue in repo |
create_comment(repo, pr, :pr) |
Comment |
create a comment on pr in repo |
create_comment(repo, pr, :review) |
Comment |
create a review comment on pr in repo |
create_comment(repo, commit, :commit) |
Comment |
create a comment on commit in repo |
edit_comment(repo, comment, :issue) |
Comment |
edit the issue comment in repo |
edit_comment(repo, comment, :pr) |
Comment |
edit the PR comment in repo |
edit_comment(repo, comment, :review) |
Comment |
edit the review comment in repo |
edit_comment(repo, comment, :commit) |
Comment |
edit the commit comment in repo |
delete_comment(repo, comment, :issue) |
HttpCommon.Response |
delete the issue comment from repo |
delete_comment(repo, comment, :pr) |
HttpCommon.Response |
delete the PR comment from repo |
delete_comment(repo, comment, :review) |
HttpCommon.Response |
delete the review comment from repo |
delete_comment(repo, comment, :commit) |
HttpCommon.Response |
delete the commitcomment from repo |
delete_comment(repo, comment, :commit) |
HttpCommon.Response |
delete the commitcomment from repo |
reply_to(repo, review, comment, body) |
HttpCommon.Response |
reply to the comment (of review in repo ) creating a new comment with the specified body |
method | return type | documentation |
---|---|---|
star(repo) |
HttpCommon.Response |
star repo |
unstar(repo) |
HttpCommon.Response |
unstar repo |
stargazers(repo) |
Tuple{Vector{Owner}, Dict} |
get repo 's stargazers |
starred(user) |
Tuple{Vector{Repo}, Dict} |
get repositories starred by user |
watchers(repo) |
Tuple{Vector{Owner}, Dict} |
get repo 's watchers |
watched(user) |
Tuple{Vector{Repo}, Dict} |
get repositories watched by user |
watch(repo) |
HttpCommon.Response |
watch repo |
unwatch(repo) |
HttpCommon.Response |
unwatch repo |
method | return type | documentation |
---|---|---|
gist(id) |
Gist |
get the gist specified by id |
gist(id, revision) |
Gist |
get the gist specified by id and revision |
gists() |
Tuple{Vector{Gist}, Dict} |
get all public gists |
gists(owner) |
Tuple{Vector{Gist}, Dict} |
get all gists for owner |
create_gist() |
Gist |
create a gist |
edit_gist(gist) |
Gist |
edit a gist |
delete_gist(gist) |
HttpCommon.Response |
delete a gist |
create_gist_fork(gist) |
Gist |
fork a gist |
gist_forks(gist) |
Tuple{Vector{Gist}, Dict} |
list the forks of a gist |
star_gist(gist) |
HttpCommon.Response |
star gist |
starred_gists() |
Tuple{Vector{Gist}, Dict} |
get the starred gist s |
unstar_gist(gist) |
HttpCommon.Response |
unstar gist |
method | return type | documentation |
---|---|---|
app(id) |
App |
get the GitHub app with the specified id |
app(slug) |
App |
get the GitHub app with the specified slug |
app(;auth=auth) |
App |
get the GitHub app authenticated by the corresponding auth |
method | return type | documentation |
---|---|---|
rate_limit() |
Dict |
get your rate limit status |
authenticate(token) |
OAuth2 |
validate token and return an authentication object |
All REST methods accept the following keyword arguments:
keyword | type | default value | description |
---|---|---|---|
auth |
GitHub.Authorization |
GitHub.AnonymousAuth() |
The request's authorization |
params |
Dict |
Dict() |
The request's query parameters |
headers |
Dict |
Dict() |
The request's headers. Note that these headers will be mutated by GitHub.jl request methods. |
handle_error |
Bool |
true |
If true , a Julia error will be thrown in the event that GitHub's response reports an error. |
page_limit |
Real |
Inf |
The number of pages to return (only applies to paginated results, obviously) |
To authenticate your requests to GitHub, you'll need to generate an appropriate access token. Then, you can do stuff like the following (this example assumes that you set an environmental variable GITHUB_AUTH
containing the access token):
import GitHub
myauth = GitHub.authenticate(ENV["GITHUB_AUTH"]) # don't hardcode your access tokens!
GitHub.star("JuliaWeb/GitHub.jl"; auth = myauth) # star the GitHub.jl repo as the user identified by myauth
As you can see, you can propagate the identity/permissions of the myauth
token to GitHub.jl's methods by passing auth = myauth
as a keyword argument.
Note that if authentication is not provided, they'll be subject to the restrictions GitHub imposes on unauthenticated requests (such as stricter rate limiting)
GitHub apps (formerly called integrations) have their own authentication format
based on JSON Web Tokens. When creating a GitHub app, you will be
prompted to download your app's private key. You can use this private key to
authenticate as a Github App using the JWTAuth
type:
appauth = JWTAuth(1234, "privkey.pem") # Replace with your app id/privkey file
The following shows a complete example that opens an issue on every repository on which your application gets installed:
listener = GitHub.EventListener() do event
# On installation, open an issue on every repository we got installed in
if event.kind == "installation"
# Authenticate as the application
appauth = GitHub.JWTAuth(1234, "privkey.pem")
# Now, get permissions for this particular installation
installation = Installation(event.payload["installation"])
auth = create_access_token(installation, appauth)
for repo in event.payload["repositories"]
create_issue(GitHub.Repo(repo), auth=auth,
params = Dict(
:title => "Hello World",
:body => "Thank you for installing me - I needed that"
))
end
end
return HttpCommon.Response(200)
end
GitHub.run(listener, host=IPv4(0,0,0,0), port=8888)
GitHub will often paginate results for requests that return multiple items. On the GitHub.jl side of things, it's pretty easy to see which methods return paginated results by referring to the REST Methods documentation; if a method returns a Tuple{Vector{T}, Dict}
, that means its results are paginated.
Paginated methods return both the response values, and some pagination metadata. You can use the per_page
/page
query parameters and the page_limit
keyword argument to configure result pagination.
For example, let's request a couple pages of GitHub.jl's PRs, and configure our result pagination to see how it works:
# show all PRs (both open and closed), and give me 3 items per page starting at page 2
julia> myparams = Dict("state" => "all", "per_page" => 3, "page" => 2);
julia> prs, page_data = pull_requests("JuliaWeb/GitHub.jl"; params = myparams, page_limit = 2);
julia> prs # 3 items per page * 2 page limit == 6 items, as expected
6-element Array{GitHub.PullRequest,1}:
GitHub.PullRequest(44)
GitHub.PullRequest(43)
GitHub.PullRequest(42)
GitHub.PullRequest(41)
GitHub.PullRequest(39)
GitHub.PullRequest(38)
julia> page_data
Dict{String,String} with 4 entries:
"prev" => "https://api.github.com/repositories/16635105/pulls?page=2&per_page=3&state=all"
"next" => "https://api.github.com/repositories/16635105/pulls?page=4&per_page=3&state=all"
"first" => "https://api.github.com/repositories/16635105/pulls?page=1&per_page=3&state=all"
"last" => "https://api.github.com/repositories/16635105/pulls?page=7&per_page=3&state=all"
In the above, prs
contains the results from page 2 and 3. We know this because we specified page 2 as our starting page ("page" => 2
), and limited the response to 2 pages max (page_limit = 2
). In addition, we know that exactly 2 pages were actually retrieved, since there are 6 items and we said each page should only contain 3 items ("per_page" => 3
).
The values provided by page_data
are the same values that are included in the Link header of the last requested item. You can continue paginating by starting a new paginated request at one of these links using the start_page
keyword argument:
# Continue paging, starting with `page_data["next"]`.
# Note that the `params` kwarg can't be used here because
# the link passed to `start_page` has its own parameters
julia> prs2, page_data2 = pull_requests("JuliaWeb/GitHub.jl"; page_limit = 2, start_page = page_data["next"]);
julia> prs2
6-element Array{GitHub.PullRequest,1}:
GitHub.PullRequest(37)
GitHub.PullRequest(34)
GitHub.PullRequest(32)
GitHub.PullRequest(30)
GitHub.PullRequest(24)
GitHub.PullRequest(22)
julia> page_data2
Dict{String,String} with 4 entries:
"prev" => "https://api.github.com/repositories/16635105/pulls?page=4&per_page=3&state=all"
"next" => "https://api.github.com/repositories/16635105/pulls?page=6&per_page=3&state=all"
"first" => "https://api.github.com/repositories/16635105/pulls?page=1&per_page=3&state=all"
"last" => "https://api.github.com/repositories/16635105/pulls?page=7&per_page=3&state=all"
GitHub.jl comes with configurable EventListener
and CommentListener
types that can be used as basic servers for parsing and responding to events delivered by GitHub's repository Webhooks.
When an EventListener
receives an event, it performs some basic validation and wraps the event payload (and some other data) in a WebhookEvent
type. This WebhookEvent
instance, along with the provided Authorization
, is then fed to the server's handler function, which the user defines to determine the server's response behavior. The handler function is expected to return an HttpCommon.Response
that is then sent back to GitHub.
The EventListener
constructor takes the following keyword arguments:
auth
: GitHub authorization (usually with repo-level permissions).secret
: A string used to verify the event source. If the event is from a GitHub Webhook, it's the Webhook's secret. If a secret is not provided, the server won't validate the secret signature of incoming requests.repos
: A vector ofRepo
s (or fully qualified repository names) listing all acceptable repositories. All repositories are whitelisted by default.events
: A vector of event names listing all acceptable events (e.g. ["commit_comment", "pull_request"]). All events are whitelisted by default.forwards
: A vector ofHttpCommon.URI
s (or URI strings) to which any incoming requests should be forwarded (after being validated by the listener)
Here's an example that demonstrates how to construct and run an EventListener
that does benchmarking on every commit and PR:
import GitHub
# EventListener settings
myauth = GitHub.authenticate(ENV["GITHUB_AUTH"])
mysecret = ENV["MY_SECRET"]
myevents = ["pull_request", "push"]
myrepos = [GitHub.Repo("owner1/repo1"), "owner2/repo2"] # can be Repos or repo names
myforwards = [HttpCommon.URI("http://myforward1.com"), "http://myforward2.com"] # can be HttpCommon.URIs or URI strings
# Set up Status parameters
pending_params = Dict(
"state" => "pending",
"context" => "Benchmarker",
"description" => "Running benchmarks..."
)
success_params = Dict(
"state" => "success",
"context" => "Benchmarker",
"description" => "Benchmarks complete!"
)
error_params(err) = Dict(
"state" => "error",
"context" => "Benchmarker",
"description" => "Error: $err"
)
# We can use Julia's `do` notation to set up the listener's handler function
listener = GitHub.EventListener(auth = myauth,
secret = mysecret,
repos = myrepos,
events = myevents,
forwards = myforwards) do event
kind, payload, repo = event.kind, event.payload, event.repository
if kind == "pull_request" && payload["action"] == "closed"
return HttpCommon.Response(200)
end
if event.kind == "push"
sha = event.payload["after"]
elseif event.kind == "pull_request"
sha = event.payload["pull_request"]["head"]["sha"]
end
GitHub.create_status(repo, sha; auth = myauth, params = pending_params)
try
# run_and_log_benchmarks isn't actually a defined function, but you get the point
run_and_log_benchmarks(event, "\$(sha)-benchmarks.csv")
catch err
GitHub.create_status(repo, sha; auth = myauth, params = error_params(err))
return HttpCommon.Response(500)
end
GitHub.create_status(repo, sha; auth = myauth, params = success_params)
return HttpCommon.Response(200)
end
# Start the listener on localhost at port 8000
GitHub.run(listener, host=IPv4(127,0,0,1), port=8000)
A CommentListener
is a special kind of EventListener
that allows users to pass data to the listener's handler function via commenting. This is useful for triggering events on repositories that require configuration settings.
A CommentListener
automatically filters out all non-comment events, and then checks the body of each comment event against a trigger Regex
supplied by the user. If a match is found in the comment, then the CommentListener
calls its handler function, passing it the event and the corresponding RegexMatch
.
The CommentListener
constructor takes the following keyword arguments:
auth
: same asEventListener
secret
: same asEventListener
repos
: same asEventListener
forwards
: same asEventListener
check_collab
: Iftrue
, only acknowledge comments made by repository collaborators. Note that, ifcheck_collab
istrue
,auth
must have the appropriate permissions to query the comment's repository for the collaborator status of the commenter.check_collab
istrue
by default.
For example, let's set up a silly CommentListener
that responds to the commenter with a greeting. To give a demonstration of the desired behavior, if a collaborator makes a comment like:
Man, I really would like to be greeted today.
`sayhello("Bob", "outgoing")`
We want the CommentLister
to reply:
Hello, Bob, you look very outgoing today!
Here's the code that will make this happen:
import GitHub
# CommentListener settings
trigger = r"`sayhello\(.*?\)`"
myauth = GitHub.authenticate(ENV["GITHUB_AUTH"])
mysecret = ENV["MY_SECRET"]
# We can use Julia's `do` notation to set up the listener's handler function.
# Note that, in our example case, `phrase` will be "`sayhello(\"Bob\", \"outgoing\")`"
listener = GitHub.CommentListener(trigger; auth = myauth, secret = mysecret) do event, phrase
# In our example case, this code sets name to "Bob" and adjective to "outgoing"
name, adjective = matchall(r"\".*?\"", phrase)
comment_params = Dict("body" => "Hello, $name, you look very $adjective today!")
# Parse the original comment event for all the necessary reply info
comment = GitHub.Comment(event.payload["comment"])
if event.kind == "issue_comment"
comment_kind = :issue
reply_to = event.payload["issue"]["number"]
elseif event.kind == "commit_comment"
comment_kind = :commit
reply_to = get(comment.commit_id)
elseif event.kind == "pull_request_review_comment"
comment_kind = :review
reply_to = event.payload["pull_request"]["number"]
# load required query params for review comment creation
comment_params["commit_id"] = get(comment.commit_id)
comment_params["path"] = get(comment.path)
comment_params["position"] = get(comment.position)
end
# send the comment creation request to GitHub
GitHub.create_comment(event.repository, reply_to, comment_kind; auth = myauth, params = comment_params)
return HttpCommon.Response(200)
end
# Start the listener on localhost at port 8000
GitHub.run(listener, host=IPv4(127,0,0,1), port=8000)