[GraphQL API] How to find repositories by search phrase and sort the results by the first commit date in found repos

[morozover]

Select Topic Area

Question

Body

Hello there,

Could you please help me to create Github GraphQL API query?

I need implement this flow:

1. Find repositories by search phrase, for example, «LLM on CPU»
2. For every found repo get the date of first commit in this repo (I think I need commit author date)
3. Sort the order of repos by the the dates of first commit in this repos
4. Apply limit to search results (as far as I know technical limit is 1000)

Every element (repo) in search results must have the following fields (names of fields gathered from API reference docs):

• name
• createdAt
• url
• the date of first commit in this repo

I wrote this code, but it does not do everything I need:

query {
  search(query: "LLM on CPU", type: REPOSITORY, first: 4) {
    repositoryCount
    edges {
      node {
        ... on Repository {
          name
          createdAt
          url
        }
      }
    }
  }
}

Please help me to write right query

P.S.
Is there a method in REST API to get the date of first commit in particular repo by specifying the ID of the repo in form of {owner}/{repository} as it is specified in repo URL?

[morozover]

I haven't found required info in GraphQL API docs

[JoinEvgen]

@😂

[morozover]

Is there a way to gather required info?

[morozover]

Can anyone help me?

[morozover]

Maybe I haven't specified category for this question correctly?

[LangLangBart]

3. Sort the order of repos by the dates of first commit in these repos

The GraphQL API does not provide this option for searches.

4. Apply a limit to search results (the technical limit is 1000)

The GraphQL API limits are different from the REST API, see the docs¹.

2. For every found repo get the date of first commit in this repo

This is the challenging part.

One approach is to use pagination and traverse all commits. However, this method might lead to rate-limiting issues if the search results contain a large number of repositories and many commits per repository.

Additionally, it seems rather inefficient to query and discard a large amount of data just to obtain the last commit.

---

I recommend the following:

query ($searchPhrase: String = "LLM on CPU") {
  search(query: $searchPhrase, type: REPOSITORY, first: 100) {
    repositoryCount
    nodes {
      ... on Repository {
        url
        nameWithOwner
        createdAt
        defaultBranchRef {
          target {
            ... on Commit {
              history(first: 1) {
                totalCount
              }
            }
          }
        }
      }
    }
  }
}

1. Execute the above GraphQL API query. You may optionally include pagination² in the search section.

2. Iterate over the results using nameWithOwner and totalCount. Make a REST API³ call to
   obtain the first commit of each result. Ensure to use a token and consider adding a slight
   delay. Making too many requests in quick succession can trigger another type of rate limit.
   Experiment with it.

# Exemplary 'REST API' calls to get the 'first commit' using 'curl' and 'jq'
# [curl] https://curl.se/
# [jq] https://jqlang.github.io/jq/
curl -s 'https://api.github.com/repos/${nameWithOwner}/commits?per_page=1&page=${totalCount}' |
  jq '.[].commit.author.date'

3. Merge each response from the REST API query with the response from the GraphQL API query, and then sort the result.

Footnotes

1. Rate limits and node limits for the GraphQL API - GitHub Docs ↩

2. Using pagination in the GraphQL API - GitHub Docs ↩

3. REST API endpoints for commits - GitHub Docs ↩