Improve GitHub branch listing pagination

src/VCS/Adapter.php

@@ -221,13 +221,23 @@ abstract public function getEvent(string $event, string $payload): array;
     abstract public function getRepositoryName(string $repositoryId): string;
 
     /**
-     * Lists branches for a given repository
+     * Lists branches for a given repository.
+     *
+     * Search is prefix-based: 'feat' matches 'feature-branch' but not 'my-feature'.
+     * GitHub uses true server-side cursor pagination via GraphQL; pass the returned
+     * nextCursor as $page on subsequent calls to advance the page.
+     * Other providers (GitLab, Gitea, Gogs, Forgejo) fetch all matching branches
+     * client-side and slice; for them nextCursor is always null, but hasNext
+     * correctly reflects whether more items exist beyond the current slice.
      *
      * @param string $owner Owner name of the repository
      * @param string $repositoryName Name of the repository
-     * @return array<string> List of branch names as array
+     * @param int $perPage Number of results per page, clamped to [1, 100]
+     * @param int|string|null $page For GitHub: pass 1 for the first page; for subsequent pages always pass the opaque cursor string from nextCursor — GitHub has no concept of integer offset pages and any integer other than 1 is treated as page 1. For other providers: 1-based integer page number; string cursors are ignored and treated as page 1.
+     * @param string $search Prefix filter for branch names; empty string returns all branches
+     * @return array{items: array<string>, hasNext: bool, nextCursor: string|null}
      */
-    abstract public function listBranches(string $owner, string $repositoryName): array;
+    abstract public function listBranches(string $owner, string $repositoryName, int $perPage = 100, int|string|null $page = 1, string $search = ''): array;
 
     /**
      * Updates status check of each commit

src/VCS/Adapter/Git/GitHub.php

@@ -742,32 +742,126 @@ public function getPullRequestFromBranch(string $owner, string $repositoryName,
     }
 
     /**
-     * Lists branches for a given repository
+     * Lists branches using GitHub GraphQL repository.refs with prefix search and cursor pagination.
      *
-     * @param  string  $owner Owner name of the repository
-     * @param  string  $repositoryName Name of the GitHub repository
-     * @param  int  $perPage Number of branches to fetch per page
-     * @param  int  $page Page number to start fetching from
-     * @return array<string> List of branch names as array
+     * Search matches branch names by prefix only ('feat' → 'feature-x', not 'my-feature').
+     * Pass an integer $page to walk forward page-by-page (each step costs one extra GraphQL call
+     * to resolve the cursor chain); pass a cursor string from a previous nextCursor to jump
+     * directly. perPage is clamped to [1, 100].
+     *
+     * @param  string  $owner
+     * @param  string  $repositoryName
+     * @param  int  $perPage Clamped to [1, 100]
+     * @param  int|string|null  $page Pass 1 (or null) for the first page. For subsequent pages
+     *   always pass the opaque cursor string from the previous nextCursor — GitHub uses
+     *   cursor-based GraphQL pagination and has no concept of integer page offsets.
+     *   Any integer value other than 1 is treated as page 1.
+     * @param  string  $search Prefix filter; empty returns all branches
+     * @return array{items: array<string>, hasNext: bool, nextCursor: string|null}
      */
-    public function listBranches(string $owner, string $repositoryName, int $perPage = 100, int $page = 1): array
+    public function listBranches(string $owner, string $repositoryName, int $perPage = 100, int|string|null $page = 1, string $search = ''): array
     {
-        $url = "/repos/$owner/$repositoryName/branches";
         $perPage = min(max($perPage, 1), 100);
+        $cursor = is_string($page) ? $page : null;
+
+        $gql = <<<'GRAPHQL'
+query ListBranches($owner: String!, $name: String!, $first: Int!, $after: String, $query: String) {
+  repository(owner: $owner, name: $name) {
+    refs(refPrefix: "refs/heads/", first: $first, after: $after, orderBy: {field: ALPHABETICAL, direction: ASC}, query: $query) {
+      edges {
+        cursor
+        node {
+          name
+        }
+      }
+      pageInfo {
+        hasNextPage
+        endCursor
+      }
+    }
+  }
+}
+GRAPHQL;
+
+        // We use GraphQL instead of REST for two reasons that the REST API cannot satisfy:
+        //  1. Server-side search narrowing: REST GET /repos/{owner}/{repo}/branches has no
+        //     search or filter parameter at all; GraphQL refs() accepts a `query` variable.
+        //  2. Per-edge cursors: REST only supports integer ?page=N offsets; GraphQL edges
+        //     carry individual cursors so we can resume from an exact item across calls.
+        //
+        // GraphQL `query` does substring matching, so we additionally enforce prefix
+        // semantics client-side with str_starts_with. We collect up to $perPage + 1
+        // matching edges across as many GraphQL pages as needed:
+        //  - If we find the +1 probe item, hasNext=true and nextCursor points to the
+        //    cursor of the last returned item, so the next call resumes exactly where
+        //    we stopped.
+        //  - If GitHub is exhausted before the probe, hasNext=false.
+        // This ensures items is never empty while hasNext is true.
+        /** @var array<array{name: string, cursor: string}> $collected */
+        $collected = [];
+        $currentCursor = $cursor;
+        $hasNextPage = false;
+
+        do {
+            $response = $this->call(self::METHOD_POST, '/graphql', ['Authorization' => "Bearer $this->accessToken"], [
+                'query' => $gql,
+                'variables' => [
+                    'owner' => $owner,
+                    'name' => $repositoryName,
+                    'first' => $perPage,
+                    'after' => $currentCursor,
+                    'query' => $search !== '' ? $search : null,
+                ],
+            ]);
 
-        $response = $this->call(self::METHOD_GET, $url, ['Authorization' => "Bearer $this->accessToken"], [
-            'page' => $page,
-            'per_page' => $perPage,
-        ]);
+            $statusCode = $response['headers']['status-code'] ?? 0;
+            $responseBody = $response['body'] ?? [];
 
-        $statusCode = $response['headers']['status-code'] ?? 0;
-        $responseBody = $response['body'] ?? [];
+            if ($statusCode < 200 || $statusCode >= 300 || !is_array($responseBody) || array_key_exists('errors', $responseBody)) {
+                return ['items' => [], 'hasNext' => false, 'nextCursor' => null];
+            }
 
-        if ($statusCode < 200 || $statusCode >= 300 || !is_array($responseBody)) {
-            return [];
+            $refs = $responseBody['data']['repository']['refs'] ?? null;
+
+            if (!is_array($refs)) {
+                return ['items' => [], 'hasNext' => false, 'nextCursor' => null];
+            }
+
+            $pageInfo = $refs['pageInfo'] ?? [];
+            $hasNextPage = (bool) ($pageInfo['hasNextPage'] ?? false);
+            $currentCursor = $pageInfo['endCursor'] ?? null;
+
+            $probeFound = false;
+            foreach ($refs['edges'] ?? [] as $edge) {
+                $name = $edge['node']['name'] ?? '';
+                if ($search === '' || str_starts_with($name, $search)) {
+                    $collected[] = ['name' => $name, 'cursor' => $edge['cursor'] ?? ''];
+                    if (count($collected) > $perPage) {
+                        $probeFound = true;
+                        break;
+                    }
+                }
+            }
+
+            if ($probeFound) {
+                break;
+            }
+        } while ($hasNextPage);
+
+        if (count($collected) > $perPage) {
+            $toReturn = array_slice($collected, 0, $perPage);
+            return [
+                'items' => array_column($toReturn, 'name'),
+                'hasNext' => true,
+                'nextCursor' => $toReturn[$perPage - 1]['cursor'],
+            ];
         }
 
-        return array_values(array_map(fn ($branch) => $branch['name'] ?? '', $responseBody));
+        return [
+            'items' => array_column($collected, 'name'),
+            'hasNext' => false,
+            'nextCursor' => null,
+        ];
     }
 
     /**
@@ -831,15 +925,15 @@ public function getLatestCommit(string $owner, string $repositoryName, string $b
         $responseBody = $response['body'] ?? [];
         $responseBodyCommit = $responseBody['commit'] ?? [];
         $responseBodyCommitAuthor = $responseBodyCommit['author'] ?? [];
-        $responseBodyAuthor = $responseBody['author'] ?? [];
+        // GitHub sets author to null for commits from App installations whose email
+        // does not match any GitHub user — treat it as an empty array to allow fallbacks.
+        $responseBodyAuthor = is_array($responseBody['author'] ?? null) ? $responseBody['author'] : [];
 
         if (
             !array_key_exists('name', $responseBodyCommitAuthor) ||
             !array_key_exists('message', $responseBodyCommit) ||
             !array_key_exists('sha', $responseBody) ||
-            !array_key_exists('html_url', $responseBody) ||
-            !array_key_exists('avatar_url', $responseBodyAuthor) ||
-            !array_key_exists('html_url', $responseBodyAuthor)
+            !array_key_exists('html_url', $responseBody)
         ) {
             throw new Exception("Latest commit response is missing required information.");
         }

src/VCS/Adapter/Git/GitLab.php

@@ -700,32 +700,47 @@ public function getPullRequestFromBranch(string $owner, string $repositoryName,
         ];
     }
 
-    public function listBranches(string $owner, string $repositoryName): array
+    public function listBranches(string $owner, string $repositoryName, int $perPage = 100, int|string|null $page = 1, string $search = ''): array
     {
         $ownerPath = $this->getOwnerPath($owner);
         $projectPath = urlencode("{$ownerPath}/{$repositoryName}");
+        $perPage = min(max($perPage, 1), 100);
+        $requestedPage = is_int($page) ? max($page, 1) : 1;
+
+        $query = "per_page={$perPage}&page={$requestedPage}";
+        if ($search !== '') {
+            // GitLab's search param accepts ^term to match branches that begin with
+            // term, giving us server-side prefix semantics without fetching all branches.
+            $query .= '&search=' . urlencode('^' . $search);
+        }
+
+        $url = "/projects/{$projectPath}/repository/branches?{$query}";
+        $response = $this->call(self::METHOD_GET, $url, ['PRIVATE-TOKEN' => $this->accessToken]);
+        $responseHeaders = $response['headers'] ?? [];
+        $responseHeadersStatusCode = $responseHeaders['status-code'] ?? 0;
+
+        if ($responseHeadersStatusCode >= 400) {
+            return ['items' => [], 'hasNext' => false, 'nextCursor' => null];
+        }
+
+        $responseBody = $response['body'] ?? [];
+        if (!is_array($responseBody)) {
+            return ['items' => [], 'hasNext' => false, 'nextCursor' => null];
+        }
 
         $branches = [];
-        $page = 1;
-        do {
-            $pagedUrl = "/projects/{$projectPath}/repository/branches?per_page=100&page={$page}";
-            $response = $this->call(self::METHOD_GET, $pagedUrl, ['PRIVATE-TOKEN' => $this->accessToken]);
-            $responseHeaders = $response['headers'] ?? [];
-            $responseHeadersStatusCode = $responseHeaders['status-code'] ?? 0;
-            if ($responseHeadersStatusCode >= 400) {
-                return [];
-            }
-            $responseBody = $response['body'] ?? [];
-            if (!is_array($responseBody) || empty($responseBody)) {
-                break;
-            }
-            foreach ($responseBody as $branch) {
-                $branches[] = $branch['name'] ?? '';
-            }
-            $page++;
-        } while (count($responseBody) === 100);
+        foreach ($responseBody as $branch) {
+            $branches[] = $branch['name'] ?? '';
+        }
 
-        return $branches;
+        // X-Next-Page is an empty string when there is no further page
+        $hasNext = !empty($responseHeaders['x-next-page']);
+
+        return [
+            'items' => $branches,
+            'hasNext' => $hasNext,
+            'nextCursor' => null,
+        ];
     }
 
     public function getCommit(string $owner, string $repositoryName, string $commitHash): array

src/VCS/Adapter/Git/Gitea.php

@@ -729,24 +729,29 @@ public function getPullRequestFromBranch(string $owner, string $repositoryName,
      *
      * @param string $owner Owner of the repository
      * @param string $repositoryName Name of the repository
-     * @return array<string> Array of branch names
+     * @return array{items: array<string>, hasNext: bool, nextCursor: string|null}
      */
-    public function listBranches(string $owner, string $repositoryName): array
+    public function listBranches(string $owner, string $repositoryName, int $perPage = 100, int|string|null $page = 1, string $search = ''): array
     {
+        // Gitea's branches endpoint only supports page/limit pagination — it has no
+        // server-side search or filter parameter (a q param is silently ignored).
+        // We must fetch all pages and apply str_starts_with client-side.
         $allBranches = [];
-        $perPage = 50;
+        $requestedPerPage = min(max($perPage, 1), 100);
+        $requestedPage = is_int($page) ? max($page, 1) : 1;
+        $apiPerPage = 50;
         $maxPages = 100;
 
         for ($currentPage = 1; $currentPage <= $maxPages; $currentPage++) {
-            $url = "/repos/{$owner}/{$repositoryName}/branches?page={$currentPage}&limit={$perPage}";
+            $url = "/repos/{$owner}/{$repositoryName}/branches?page={$currentPage}&limit={$apiPerPage}";
 
             $response = $this->call(self::METHOD_GET, $url, ['Authorization' => "token $this->accessToken"], decode: false);
 
             $responseHeaders = $response['headers'] ?? [];
             $responseHeadersStatusCode = $responseHeaders['status-code'] ?? 0;
 
             if ($responseHeadersStatusCode === 404) {
-                return [];
+                return ['items' => [], 'hasNext' => false, 'nextCursor' => null];
             }
 
             if ($responseHeadersStatusCode >= 400) {
@@ -770,12 +775,22 @@ public function listBranches(string $owner, string $repositoryName): array
                 }
             }
 
-            if ($pageCount < $perPage) {
+            if ($pageCount < $apiPerPage) {
                 break;
             }
         }
 
-        return $allBranches;
+        if ($search !== '') {
+            $allBranches = array_values(array_filter($allBranches, fn ($branch) => str_starts_with($branch, $search)));
+        }
+
+        $offset = ($requestedPage - 1) * $requestedPerPage;
+
+        return [
+            'items' => array_values(array_slice($allBranches, $offset, $requestedPerPage)),
+            'hasNext' => ($offset + $requestedPerPage) < count($allBranches),
+            'nextCursor' => null,
+        ];
     }
 
     /**

src/VCS/Adapter/Git/Gogs.php

@@ -260,9 +260,18 @@ public function getCommit(string $owner, string $repositoryName, string $commitH
      */
     public function getLatestCommit(string $owner, string $repositoryName, string $branch): array
     {
-        // Gogs ignores sha param — verify branch exists first
-        $branches = $this->listBranches($owner, $repositoryName);
-        if (!in_array($branch, $branches, true)) {
+        // Gogs ignores sha param — verify branch exists by scanning all pages
+        $page = 1;
+        $found = false;
+        do {
+            $result = $this->listBranches($owner, $repositoryName, 100, $page);
+            if (in_array($branch, $result['items'], true)) {
+                $found = true;
+                break;
+            }
+            $page++;
+        } while ($result['hasNext']);
+        if (!$found) {
             throw new Exception("Branch '{$branch}' not found");
         }
 
@@ -501,12 +510,16 @@ public function getCommitStatuses(string $owner, string $repositoryName, string
     /**
      * List branches
      *
-     * Gogs supports listing branches but without pagination parameters.
+     * Gogs API returns all branches in a single request (no pagination support).
      *
-     * @return array<string>
+     * @return array{items: array<string>, hasNext: bool, nextCursor: string|null}
      */
-    public function listBranches(string $owner, string $repositoryName): array
+    public function listBranches(string $owner, string $repositoryName, int $perPage = 100, int|string|null $page = 1, string $search = ''): array
     {
+        // Gogs returns all branches in one response with no server-side pagination or
+        // search support, so we fetch everything and slice/filter client-side.
+        $perPage = min(max($perPage, 1), 100);
+        $requestedPage = is_int($page) ? max($page, 1) : 1;
         $url = "/repos/{$owner}/{$repositoryName}/branches";
 
         $response = $this->call(self::METHOD_GET, $url, ['Authorization' => "token $this->accessToken"]);
@@ -515,7 +528,7 @@ public function listBranches(string $owner, string $repositoryName): array
         $responseHeadersStatusCode = $responseHeaders['status-code'] ?? 0;
 
         if ($responseHeadersStatusCode === 404) {
-            return [];
+            return ['items' => [], 'hasNext' => false, 'nextCursor' => null];
         }
 
         if ($responseHeadersStatusCode >= 400) {
@@ -525,7 +538,7 @@ public function listBranches(string $owner, string $repositoryName): array
         $responseBody = $response['body'] ?? [];
 
         if (!is_array($responseBody)) {
-            return [];
+            return ['items' => [], 'hasNext' => false, 'nextCursor' => null];
         }
 
         $branches = [];
@@ -535,6 +548,16 @@ public function listBranches(string $owner, string $repositoryName): array
             }
         }
 
-        return $branches;
+        if ($search !== '') {
+            $branches = array_values(array_filter($branches, fn ($branch) => str_starts_with($branch, $search)));
+        }
+
+        $offset = ($requestedPage - 1) * $perPage;
+
+        return [
+            'items' => array_values(array_slice($branches, $offset, $perPage)),
+            'hasNext' => ($offset + $perPage) < count($branches),
+            'nextCursor' => null,
+        ];
     }
 }