feat(mcp): paths filter for get_task_diff, subset hint for preview_merge_set

get_task_diff gained an optional paths param (git pathspec, both stat and
full-diff modes) so a large multi-file task can be narrowed to the files
actually in question instead of shipping the whole diff.

preview_merge_set now also reports Subsets: pairs where one task's changed
files are a proper subset of another's in the same set, the strongest
available post-hoc signal that a task may be redundant with another.
This commit is contained in:
mika kuns
2026-08-10 13:39:55 +02:00
parent 6a2a19cc9e
commit c715c96af9
6 changed files with 174 additions and 30 deletions
+25 -9
View File
@@ -143,10 +143,12 @@ public sealed class GitService
/// (committed-on-branch changes + uncommitted work). Used for viewing a Claude
/// task's total impact relative to where the branch started.
/// </summary>
public async Task<string> GetBranchDiffAsync(string worktreePath, string baseRef, CancellationToken ct = default)
public async Task<string> GetBranchDiffAsync(
string worktreePath, string baseRef, IReadOnlyList<string>? paths = null, CancellationToken ct = default)
{
var (exitCode, stdout, _) = await RunGitAsync(worktreePath,
["diff", baseRef], ct);
var args = new List<string> { "diff", baseRef };
AppendPathFilter(args, paths);
var (exitCode, stdout, _) = await RunGitAsync(worktreePath, args, ct);
if (exitCode == 0 && !string.IsNullOrWhiteSpace(stdout))
return stdout;
// Fallback: whatever the worktree has vs HEAD (uncommitted only).
@@ -158,24 +160,38 @@ public sealed class GitService
/// task's changes after its worktree has been merged away (the commits survive on
/// the target branch even though the worktree directory and branch ref are gone).
/// </summary>
public async Task<string> GetCommitRangeDiffAsync(string repoDir, string baseCommit, string headCommit, CancellationToken ct = default)
public async Task<string> GetCommitRangeDiffAsync(
string repoDir, string baseCommit, string headCommit, IReadOnlyList<string>? paths = null, CancellationToken ct = default)
{
var (exitCode, stdout, stderr) = await RunGitAsync(repoDir,
["diff", $"{baseCommit}..{headCommit}"], ct);
var args = new List<string> { "diff", $"{baseCommit}..{headCommit}" };
AppendPathFilter(args, paths);
var (exitCode, stdout, stderr) = await RunGitAsync(repoDir, args, ct);
if (exitCode != 0)
throw new InvalidOperationException($"git diff {baseCommit}..{headCommit} failed (exit {exitCode}): {stderr}");
return stdout;
}
public async Task<string> DiffStatAsync(string worktreePath, string baseCommit, string headCommit, CancellationToken ct = default)
public async Task<string> DiffStatAsync(
string worktreePath, string baseCommit, string headCommit, IReadOnlyList<string>? paths = null, CancellationToken ct = default)
{
var (exitCode, stdout, stderr) = await RunGitAsync(worktreePath,
["diff", "--stat", $"{baseCommit}..{headCommit}"], ct);
var args = new List<string> { "diff", "--stat", $"{baseCommit}..{headCommit}" };
AppendPathFilter(args, paths);
var (exitCode, stdout, stderr) = await RunGitAsync(worktreePath, args, ct);
if (exitCode != 0)
throw new InvalidOperationException($"git diff --stat failed (exit {exitCode}): {stderr}");
return stdout.Trim();
}
// Appends a `-- <paths>` pathspec filter so git itself narrows the diff instead of the
// caller filtering the result after the fact (works identically for --stat and full diffs).
// No-op when paths is null/empty so existing callers see no behavior change.
private static void AppendPathFilter(List<string> args, IReadOnlyList<string>? paths)
{
if (paths is not { Count: > 0 }) return;
args.Add("--");
args.AddRange(paths);
}
public async Task<string> GetFileDiffAsync(string worktreePath, string? baseCommit, string relativePath, CancellationToken ct = default)
{
string[] args = string.IsNullOrEmpty(baseCommit)
@@ -162,9 +162,9 @@ public sealed partial class DiffViewerViewModel : ViewModelBase
try
{
raw = FromCommitRange && BaseRef is not null && HeadCommit is not null
? await _git.GetCommitRangeDiffAsync(WorktreePath, BaseRef, HeadCommit, ct)
? await _git.GetCommitRangeDiffAsync(WorktreePath, BaseRef, HeadCommit, ct: ct)
: BaseRef is not null
? await _git.GetBranchDiffAsync(WorktreePath, BaseRef, ct)
? await _git.GetBranchDiffAsync(WorktreePath, BaseRef, ct: ct)
: await _git.GetDiffAsync(WorktreePath, ct);
}
catch (Exception ex)
+55 -13
View File
@@ -100,8 +100,15 @@ public sealed record MergePreviewSetEntryDto(
public sealed record FileOverlapDto(string File, IReadOnlyList<string> TaskIds);
// TaskId's changed-file list is a non-empty PROPER subset of SupersetTaskId's -- e.g. one task
// deletes two files and another deletes just one of those same two. Stronger than an Overlaps
// entry (which only says some files are shared): a real subset is the best available machine
// signal that TaskId's work may already be entirely covered by SupersetTaskId.
public sealed record SubsetRelationDto(string TaskId, string SupersetTaskId);
public sealed record MergePreviewSetResultDto(
IReadOnlyList<MergePreviewSetEntryDto> Tasks, IReadOnlyList<FileOverlapDto> Overlaps);
IReadOnlyList<MergePreviewSetEntryDto> Tasks, IReadOnlyList<FileOverlapDto> Overlaps,
IReadOnlyList<SubsetRelationDto> Subsets);
public sealed record WorktreeListItemDto(
string? TaskId, string Path, string Branch,
@@ -620,7 +627,7 @@ public sealed class ExternalMcpService
{
if (!Directory.Exists(child.Worktree.Path)) return false;
var files = ParseDiffStatFileNames(
await _git.DiffStatAsync(child.Worktree.Path, child.Worktree.BaseCommit, "HEAD", ct));
await _git.DiffStatAsync(child.Worktree.Path, child.Worktree.BaseCommit, "HEAD", ct: ct));
return files.Count == 0;
}
if (child.HandlerBaseCommit is { Length: > 0 } handlerBase && child.HandlerHeadCommit is { Length: > 0 } handlerHead)
@@ -714,6 +721,10 @@ public sealed class ExternalMcpService
[Description("false (default): the full unified diff, capped at 200 KB. true: a --stat summary with " +
"per-file insertion/deletion counts — start here when the diff may be large.")]
bool stat = false,
[Description("Restrict the diff to these paths (relative to the repo root), e.g. after a --stat pass or a " +
"conflict report already narrowed down which files matter. Omit/empty for every changed file " +
"— the default and the only prior behavior. Works in both stat and full-diff mode.")]
IReadOnlyList<string>? paths = null,
CancellationToken cancellationToken = default)
{
var (repoPath, baseCommit, headCommit) = await LoadDiffRangeAsync(taskId, cancellationToken);
@@ -722,13 +733,13 @@ public sealed class ExternalMcpService
if (stat)
{
var diffStat = await _git.DiffStatAsync(repoPath, baseCommit, headCommit ?? "HEAD", cancellationToken);
var diffStat = await _git.DiffStatAsync(repoPath, baseCommit, headCommit ?? "HEAD", paths, cancellationToken);
return new TaskDiffDto(diffStat, ParseDiffStatFileNames(diffStat), false, diffStat.Length);
}
var diff = headCommit is null
? await _git.GetBranchDiffAsync(repoPath, baseCommit, cancellationToken)
: await _git.GetCommitRangeDiffAsync(repoPath, baseCommit, headCommit, cancellationToken);
? await _git.GetBranchDiffAsync(repoPath, baseCommit, paths, cancellationToken)
: await _git.GetCommitRangeDiffAsync(repoPath, baseCommit, headCommit, paths, cancellationToken);
var files = ParseDiffFileNames(diff);
if (diff.Length <= maxBytes)
@@ -949,11 +960,13 @@ public sealed class ExternalMcpService
[McpServerTool, Description(
"Plan a batch merge: preview_merge for several tasks against the same targetBranch, plus a file-overlap " +
"check between them. Per entry you get preview_merge's fields, or error instead when that task could not " +
"be previewed (it is then left out of the overlap computation). overlaps names, for each file touched by " +
"MORE THAN ONE of the given tasks, which tasks touch it — a single taskId always yields no overlaps. " +
"IMPORTANT: overlap is a HINT and its absence is not safety — two tasks touching entirely different files " +
"(one deleting a symbol, another still referencing it) can still collide unflagged, and as with " +
"preview_merge a clean result does not mean the merge builds.")]
"be previewed (it is then left out of the overlap/subset computation). overlaps names, for each file " +
"touched by MORE THAN ONE of the given tasks, which tasks touch it — a single taskId always yields no " +
"overlaps. subsets flags a stronger relation: TaskId's changed files are a PROPER subset of " +
"SupersetTaskId's — the strongest hint you get post-hoc that TaskId may be redundant with SupersetTaskId, " +
"worth checking before merging both. IMPORTANT: neither overlap nor subset is a safety guarantee — two " +
"tasks touching entirely different files (one deleting a symbol, another still referencing it) can still " +
"collide unflagged, and as with preview_merge a clean result does not mean the merge builds.")]
public async Task<MergePreviewSetResultDto> PreviewMergeSet(
IReadOnlyList<string> taskIds,
[Description("Branch to preview every task against; defaults to the repo's current branch.")]
@@ -990,7 +1003,36 @@ public sealed class ExternalMcpService
.OrderBy(o => o.File, StringComparer.OrdinalIgnoreCase)
.ToList();
return new MergePreviewSetResultDto(entries, overlaps);
var subsets = FindSubsetRelations(filesByTask);
return new MergePreviewSetResultDto(entries, overlaps, subsets);
}
// A task's changed-file set is flagged only against another task's STRICTLY LARGER set that
// contains every one of its files -- an empty set (isEmpty task) never qualifies, since that
// case is already surfaced via IsEmpty and would otherwise match every other task trivially.
private static List<SubsetRelationDto> FindSubsetRelations(Dictionary<string, IReadOnlyList<string>> filesByTask)
{
var fileSets = filesByTask.ToDictionary(
kv => kv.Key,
kv => new HashSet<string>(kv.Value, StringComparer.OrdinalIgnoreCase));
var subsets = new List<SubsetRelationDto>();
foreach (var (taskId, fileSet) in fileSets)
{
if (fileSet.Count == 0) continue;
foreach (var (otherTaskId, otherFileSet) in fileSets)
{
if (otherTaskId == taskId || otherFileSet.Count <= fileSet.Count) continue;
if (fileSet.IsSubsetOf(otherFileSet))
subsets.Add(new SubsetRelationDto(taskId, otherTaskId));
}
}
return subsets
.OrderBy(s => s.TaskId, StringComparer.OrdinalIgnoreCase)
.ThenBy(s => s.SupersetTaskId, StringComparer.OrdinalIgnoreCase)
.ToList();
}
// Shared core for PreviewMerge/PreviewMergeSet: throws a clear InvalidOperationException instead of
@@ -1029,7 +1071,7 @@ public sealed class ExternalMcpService
var behind = await GitRevListCountAsync(list.WorkingDir, $"{wt.BranchName}..{target}", ct);
var changedFiles = Directory.Exists(wt.Path)
? ParseDiffStatFileNames(await _git.DiffStatAsync(wt.Path, wt.BaseCommit, "HEAD", ct))
? ParseDiffStatFileNames(await _git.DiffStatAsync(wt.Path, wt.BaseCommit, "HEAD", ct: ct))
: Array.Empty<string>();
return (preview, behind, changedFiles, changedFiles.Count == 0);
@@ -1043,7 +1085,7 @@ public sealed class ExternalMcpService
var isEmpty = string.Equals(handlerBase, handlerHead, StringComparison.Ordinal);
var changedFiles = isEmpty
? Array.Empty<string>()
: ParseDiffStatFileNames(await _git.DiffStatAsync(list.WorkingDir, handlerBase, handlerHead, ct));
: ParseDiffStatFileNames(await _git.DiffStatAsync(list.WorkingDir, handlerBase, handlerHead, ct: ct));
var preview = new MergePreviewResult(TaskMergeService.PreviewClean, Array.Empty<string>(), changedFiles.Count);
return (preview, 0, changedFiles, isEmpty);
@@ -65,7 +65,7 @@ public sealed class PlanningAggregator
string unified;
try
{
unified = await _git.GetBranchDiffAsync(wt.Path, wt.BaseCommit, ct);
unified = await _git.GetBranchDiffAsync(wt.Path, wt.BaseCommit, ct: ct);
}
catch (Exception ex)
{
@@ -147,7 +147,7 @@ public sealed class WorktreeManager
await _git.CommitAsync(ctx.WorktreePath, message, ct);
var head = await _git.RevParseHeadAsync(ctx.WorktreePath, ct);
var diffStat = await _git.DiffStatAsync(ctx.WorktreePath, ctx.BaseCommit, head, ct);
var diffStat = await _git.DiffStatAsync(ctx.WorktreePath, ctx.BaseCommit, head, ct: ct);
using var context = _dbFactory.CreateDbContext();
var wtRepo = new WorktreeRepository(context);
@@ -769,7 +769,7 @@ public sealed class ExternalMcpServiceTests : IDisposable
var sut = BuildSut(CreateQueue());
await Assert.ThrowsAsync<InvalidOperationException>(
() => sut.GetTaskDiff(task.Id, false, CancellationToken.None));
() => sut.GetTaskDiff(task.Id, false, null, CancellationToken.None));
}
[Fact]
@@ -785,12 +785,52 @@ public sealed class ExternalMcpServiceTests : IDisposable
await mgr.CommitIfChangedAsync(wt, task, list, CancellationToken.None);
var sut = BuildSut(CreateQueue());
var diff = await sut.GetTaskDiff(task.Id, false, CancellationToken.None);
var diff = await sut.GetTaskDiff(task.Id, false, null, CancellationToken.None);
Assert.Contains("added.txt", diff.Files);
Assert.False(diff.Truncated);
}
[Fact]
public async Task GetTaskDiff_WithPathsFilter_ReturnsOnlyRequestedFile()
{
if (!GitAvailable) { Assert.True(true, "git not available -- skipping"); return; }
var (task, list, wt) = await SeedWorktreeAsync();
File.WriteAllText(Path.Combine(wt.WorktreePath, "added.txt"), "content");
File.WriteAllText(Path.Combine(wt.WorktreePath, "other.txt"), "other content");
var cfg = new WorkerConfig { WorktreeRootStrategy = "sibling" };
var mgr = new WorktreeManager(new GitService(), _db.CreateFactory(), cfg, NullLogger<WorktreeManager>.Instance);
await mgr.CommitIfChangedAsync(wt, task, list, CancellationToken.None);
var sut = BuildSut(CreateQueue());
var diff = await sut.GetTaskDiff(task.Id, false, ["added.txt"], CancellationToken.None);
Assert.Equal(new[] { "added.txt" }, diff.Files);
Assert.DoesNotContain("other.txt", diff.Content);
}
[Fact]
public async Task GetTaskDiff_StatMode_WithPathsFilter_ReturnsOnlyRequestedFile()
{
if (!GitAvailable) { Assert.True(true, "git not available -- skipping"); return; }
var (task, list, wt) = await SeedWorktreeAsync();
File.WriteAllText(Path.Combine(wt.WorktreePath, "added.txt"), "content");
File.WriteAllText(Path.Combine(wt.WorktreePath, "other.txt"), "other content");
var cfg = new WorkerConfig { WorktreeRootStrategy = "sibling" };
var mgr = new WorktreeManager(new GitService(), _db.CreateFactory(), cfg, NullLogger<WorktreeManager>.Instance);
await mgr.CommitIfChangedAsync(wt, task, list, CancellationToken.None);
var sut = BuildSut(CreateQueue());
var diff = await sut.GetTaskDiff(task.Id, true, ["other.txt"], CancellationToken.None);
Assert.Equal(new[] { "other.txt" }, diff.Files);
Assert.DoesNotContain("added.txt", diff.Content);
}
[Fact]
public async Task GetTaskDiff_WorktreeLessHandlerTask_UsesHandlerCommitRangeOverListWorkingDir()
{
@@ -819,7 +859,7 @@ public sealed class ExternalMcpServiceTests : IDisposable
}
var sut = BuildSut(CreateQueue());
var diff = await sut.GetTaskDiff(task.Id, false, CancellationToken.None);
var diff = await sut.GetTaskDiff(task.Id, false, null, CancellationToken.None);
Assert.Contains("handled.txt", diff.Files);
Assert.False(diff.Truncated);
@@ -850,7 +890,7 @@ public sealed class ExternalMcpServiceTests : IDisposable
}
var sut = BuildSut(CreateQueue());
var diff = await sut.GetTaskDiff(task.Id, true, CancellationToken.None);
var diff = await sut.GetTaskDiff(task.Id, true, null, CancellationToken.None);
Assert.Contains("handled.txt", diff.Content);
}
@@ -1637,6 +1677,52 @@ public sealed class ExternalMcpServiceTests : IDisposable
Assert.Equal(
new[] { taskA.Id, taskB.Id }.OrderBy(x => x),
overlap.TaskIds.OrderBy(x => x));
// Equal file sets overlap but neither is a PROPER subset of the other.
Assert.Empty(result.Subsets);
}
[Fact]
public async Task PreviewMergeSet_TaskFilesAreProperSubsetOfAnother_ReportsSubset()
{
if (!GitAvailable) { Assert.True(true, "git not available -- skipping"); return; }
var repo = new GitRepoFixture();
_repos.Add(repo);
GitRepoFixture.RunGit(repo.RepoDir, "branch", "-m", "main");
var listId = Guid.NewGuid().ToString();
await _lists.AddAsync(new ListEntity { Id = listId, Name = "L", WorkingDir = repo.RepoDir, CreatedAt = DateTime.UtcNow });
var subsetTask = await SeedTaskAsync(listId, "subset", TaskStatus.WaitingForReview);
var supersetTask = await SeedTaskAsync(listId, "superset", TaskStatus.WaitingForReview);
await CreateActiveWorktreeAsync(repo, subsetTask.Id, "shared.txt", "shared change\n");
var wtPath = Path.Combine(Path.GetTempPath(), $"wt_{Guid.NewGuid():N}");
_worktreeCleanups.Add((repo.RepoDir, wtPath));
var branch = $"claudedo/{supersetTask.Id[..8]}";
GitRepoFixture.RunGit(repo.RepoDir, "worktree", "add", "-b", branch, wtPath, repo.BaseCommit);
File.WriteAllText(Path.Combine(wtPath, "shared.txt"), "shared change too\n");
File.WriteAllText(Path.Combine(wtPath, "extra.txt"), "extra\n");
GitRepoFixture.RunGit(wtPath, "add", "-A");
GitRepoFixture.RunGit(wtPath, "commit", "-m", "edit shared+extra");
using (var ctx = _db.CreateContext())
{
ctx.Worktrees.Add(new WorktreeEntity
{
TaskId = supersetTask.Id, Path = wtPath, BranchName = branch,
BaseCommit = repo.BaseCommit,
HeadCommit = GitRepoFixture.RunGit(wtPath, "rev-parse", "HEAD").Trim(),
State = WorktreeState.Active, CreatedAt = DateTime.UtcNow,
});
await ctx.SaveChangesAsync();
}
var sut = BuildSut(CreateQueue());
var result = await sut.PreviewMergeSet([subsetTask.Id, supersetTask.Id], "main", CancellationToken.None);
var relation = Assert.Single(result.Subsets);
Assert.Equal(subsetTask.Id, relation.TaskId);
Assert.Equal(supersetTask.Id, relation.SupersetTaskId);
}
[Fact]