From 0d4f80593e5c466e77eeb6a73c09b991a1c68f72 Mon Sep 17 00:00:00 2001
From: Sachin Kumar <sachinkumar905846@gmail.com>
Date: Wed, 15 Apr 2026 10:25:59 +0530
Subject: [PATCH] feat(engine): add commit iteration support for pull request
 evaluation

---
 internal/engine/ingester/diff/diff.go | 74 +++++++++++++++++++++++++++
 internal/providers/github/common.go   | 43 ++++++++++++++++
 2 files changed, 117 insertions(+)

diff --git a/internal/engine/ingester/diff/diff.go b/internal/engine/ingester/diff/diff.go
index 4412e58a7f..f049d05fc3 100644
--- a/internal/engine/ingester/diff/diff.go
+++ b/internal/engine/ingester/diff/diff.go
@@ -18,6 +18,7 @@ import (
 
 	"github.com/go-git/go-billy/v5"
 	"github.com/go-git/go-billy/v5/helper/iofs"
+	"github.com/google/go-github/v63/github"
 	scalibr "github.com/google/osv-scalibr"
 	"github.com/google/osv-scalibr/extractor"
 	scalibr_fs "github.com/google/osv-scalibr/fs"
@@ -92,6 +93,13 @@ func (di *Diff) Ingest(
 	}
 	prNumber := int(pr.Number)
 
+	// Best-effort experiment: if the underlying provider supports listing PR commits,
+	// iterate over them here so we can observe commit-level data during evaluation. This
+	// is intentionally non-blocking for providers that do not implement the method.
+	if err := di.logPullRequestCommits(ctx, pr, prNumber); err != nil {
+		zerolog.Ctx(ctx).Debug().Err(err).Msg("failed to list pull request commits")
+	}
+
 	switch di.cfg.GetType() {
 	case "", pb.DiffTypeDep:
 		return di.getDepTypeDiff(ctx, prNumber, pr)
@@ -108,6 +116,72 @@ func (di *Diff) Ingest(
 	}
 }
 
+// commitLister is an internal, optional trait that providers can implement to
+// expose pull request commits for a given PR.
+type commitLister interface {
+	ListPullRequestCommits(
+		ctx context.Context,
+		owner string,
+		repo string,
+		prNumber int,
+		perPage int,
+		pageNumber int,
+	) ([]*github.RepositoryCommit, *github.Response, error)
+}
+
+// logPullRequestCommits iterates over all commits in the given pull request,
+// logging their SHAs and messages when the underlying provider implements the
+// commitLister trait. This is a no-op for providers that do not support it.
+func (di *Diff) logPullRequestCommits(
+	ctx context.Context,
+	pr *pbinternal.PullRequest,
+	prNumber int,
+) error {
+	cli, ok := di.cli.(commitLister)
+	if !ok {
+		return nil
+	}
+
+	logger := zerolog.Ctx(ctx)
+	page := 0
+	const prCommitsPerPage = 50
+
+	for {
+		commits, resp, err := cli.ListPullRequestCommits(
+			ctx,
+			pr.RepoOwner,
+			pr.RepoName,
+			prNumber,
+			prCommitsPerPage,
+			page,
+		)
+		if err != nil {
+			return fmt.Errorf("error listing pull request commits: %w", err)
+		}
+
+		for _, c := range commits {
+			sha := c.GetSHA()
+			msg := c.GetCommit().GetMessage()
+			firstLine := strings.Split(msg, "\n")[0]
+
+			logger.Debug().
+				Str("pr_repo", fmt.Sprintf("%s/%s", pr.RepoOwner, pr.RepoName)).
+				Int64("pr_number", pr.Number).
+				Str("commit_sha", sha).
+				Str("commit_msg", firstLine).
+				Msg("pull request commit")
+		}
+
+		if resp == nil || resp.NextPage == 0 {
+			break
+		}
+
+		page = resp.NextPage
+	}
+
+	return nil
+}
+
 func (di *Diff) getDepTypeDiff(ctx context.Context, prNumber int, pr *pbinternal.PullRequest) (*interfaces.Ingested, error) {
 	deps := pbinternal.PrDependencies{Pr: pr}
 	page := 0
diff --git a/internal/providers/github/common.go b/internal/providers/github/common.go
index 21a9008cb3..fe5778ec98 100644
--- a/internal/providers/github/common.go
+++ b/internal/providers/github/common.go
@@ -422,6 +422,49 @@ func (c *GitHub) ListFiles(
 	return resp.files, resp.resp, err
 }
 
+// ListPullRequestCommits is a wrapper for the GitHub API to list commits in a pull request.
+// It is currently intended for internal engine experimentation, such as iterating over all
+// commits in a PR during evaluation.
+func (c *GitHub) ListPullRequestCommits(
+	ctx context.Context,
+	owner string,
+	repo string,
+	prNumber int,
+	perPage int,
+	pageNumber int,
+) ([]*github.RepositoryCommit, *github.Response, error) {
+	type listCommitsRespWrapper struct {
+		commits []*github.RepositoryCommit
+		resp    *github.Response
+	}
+
+	op := func() (listCommitsRespWrapper, error) {
+		opt := &github.ListOptions{
+			Page:    pageNumber,
+			PerPage: perPage,
+		}
+		commits, resp, err := c.client.PullRequests.ListCommits(ctx, owner, repo, prNumber, opt)
+
+		wrapped := listCommitsRespWrapper{
+			commits: commits,
+			resp:    resp,
+		}
+
+		if isRateLimitError(err) {
+			waitErr := c.waitForRateLimitReset(ctx, err)
+			if waitErr == nil {
+				return wrapped, err
+			}
+			return wrapped, backoffv4.Permanent(err)
+		}
+
+		return wrapped, backoffv4.Permanent(err)
+	}
+
+	resp, err := performWithRetry(ctx, op)
+	return resp.commits, resp.resp, err
+}
+
 // CreateReview is a wrapper for the GitHub API to create a review
 func (c *GitHub) CreateReview(
 	ctx context.Context, owner, repo string, number int, reviewRequest *github.PullRequestReviewRequest,