Using the GitHub API in Pharo 7

In my mining activities with Pharo, I found it useful to query GitHub to select projects to mine.

Posted by on March 22, 2019 · 6 mins read

In my mining activities with Pharo, I found it useful to query GitHub to select projects to mine according to their popularity (e.g., forks and stars).

Which API (REST or GraphQL)?

GitHub offers different ways to query its infrastructure. The traditional REST query is the v3 of its API, and a more powerful (which requires an authentication token) GraphQL v4 of its API.


GitHub’s v3 API for REST is OK for getting info about projects, such as the ones that have lots of stars or forks.

Here’s a snippet of code to show how to use it in Pharo:

znClient := ZnClient new.
jsonString := znClient

Note the specifics in the query, e.g. language:java and sort=stars. You can read more about these options on GitHub’s documentation for API v3.

Note this query returns at most 5 results (per_page=5). Without this option, GitHub gives 30 results per query. It’s possible to get up to 100 results per query by setting per_page=100. To get multiple “pages”, you specify the page number (page=2, page=3, etc.) in queries.

The jsonString is a JSON string response, which can be accessed nicely using NeoJSON (which must be loaded in your Pharo image).

NeoJSON to get to the values

Assuming you loaded NeoJSON, here’s how you could get to the values from the jsonString response above:

queryDic := NeoJSONReader fromString: jsonString.
projectsColl := (queryDic at: 'items')
    collect: [ :item | 
        {(item at: 'full_name').
        (item at: 'git_url').
        (item at: 'watchers_count').
        (item at: 'forks_count')} ].

projectsColl asString
"'#(#(''iluwatar/java-design-patterns'' ''git://''
      45632 14784)
    #(''elastic/elasticsearch'' ''git://''
      39349 13141)
    #(''ReactiveX/RxJava'' ''git://''
      38107 6443)
    #(''spring-projects/spring-boot'' ''git://''
      35550 23522)
    #(''kdn251/interviews'' ''git://''
      33272 6610))'"

It’s not bad, but I was hoping to get data about how many issues (total) existed for projects. This proved troublesome with version 3 of the REST API. It’s possible, but requires multiple calls, which amounts to separate requests through each repo’s issues page. Googling this problem, I found the mention of GitHub encouraging people to move to GraphQL.

GraphQL (v4 API)

GraphQL is more complex, but in theory more powerful. You have to do POST requests, and you need an authorization token (which means GitHub tracks what you are doing). Since the responses are still in JSON, NeoJSON can pick them apart for you as before.

So, here’s what I did according to the instructions from GitHub:

  • I first created a token with scope repo (and all repo sub-scopes). I described it as a GitHub repo mining token. For obvious reasons, I don’t reveal the token here in my examples.
  • Before making the first query call, I make a call to login with the token.
  • I specify the token on subsequent calls as before.

Here’s a Playground script that could be converted easily to a class with methods:

| graphQLLoginString graphQLGetIssueCount doGraphQLQuery myAPIToken response queryDic issueCount |
graphQLLoginString :=
    'query { viewer { login }}'.
graphQLGetIssueCount :=
    'query { repository(owner:"isaacs", name:"github") {issues(states:OPEN) {totalCount}}}'.
"Get a token for repo scope
"block closure to be called like a method"
doGraphQLQuery := [ :authToken :graphQLquery | 
    | escaper |
    "escape any double-quotes in the GraphQL query"
    escaper := [ :stringToEscape | 
        stringToEscape copyWithRegex: '\"' matchesReplacedWith: '\"' ].
    ZnClient new
        url: '';
        headerAt: 'Authorization' put: 'bearer ' , authToken;
                with: '{"query": "' , (escaper value: graphQLquery) , '"}'
                type: ZnMimeType applicationJson);
        post ].
"Log in first, should check for errors in the response"
response := doGraphQLQuery
    value: myAPIToken value: graphQLLoginString.
"Do the query to get the number of issues"
response := doGraphQLQuery
    value: myAPIToken
    value: graphQLGetIssueCount.
"Pull the number out of the JSON"
queryDic := NeoJSONReader fromString: response.
issueCount := ((((queryDic at: 'data') at: 'repository') at: 'issues')
    at: 'totalCount') asInteger

Photo credit