From 1aa5cad4c8d6455eeb1f10893549e18bcca11996 Mon Sep 17 00:00:00 2001 From: Runxi Yu Date: Wed, 25 Mar 2026 19:33:32 +0000 Subject: object/fetch: Rename from object/resolve --- object/fetch/doc.go | 5 ++ object/fetch/exact_blob.go | 24 +++++++ object/fetch/exact_blob_reader.go | 14 ++++ object/fetch/exact_commit.go | 24 +++++++ object/fetch/exact_commit_reader.go | 16 +++++ object/fetch/exact_object.go | 18 +++++ object/fetch/exact_reader.go | 26 ++++++++ object/fetch/exact_tag.go | 24 +++++++ object/fetch/exact_tag_reader.go | 16 +++++ object/fetch/exact_tree.go | 24 +++++++ object/fetch/exact_tree_reader.go | 16 +++++ object/fetch/fetcher.go | 17 +++++ object/fetch/object_parse.go | 28 ++++++++ object/fetch/path.go | 103 ++++++++++++++++++++++++++++ object/fetch/peel_to_blob.go | 29 ++++++++ object/fetch/peel_to_blob_id.go | 39 +++++++++++ object/fetch/peel_to_blob_reader.go | 18 +++++ object/fetch/peel_to_commit.go | 29 ++++++++ object/fetch/peel_to_commit_id.go | 39 +++++++++++ object/fetch/peel_to_commit_reader.go | 20 ++++++ object/fetch/peel_to_tag.go | 12 ++++ object/fetch/peel_to_tag_id.go | 8 +++ object/fetch/peel_to_tag_reader.go | 20 ++++++ object/fetch/peel_to_tree.go | 33 +++++++++ object/fetch/peel_to_tree_id.go | 46 +++++++++++++ object/fetch/peel_to_tree_reader.go | 20 ++++++ object/fetch/treefs.go | 30 +++++++++ object/fetch/treefs_entry.go | 90 +++++++++++++++++++++++++ object/fetch/treefs_info.go | 75 +++++++++++++++++++++ object/fetch/treefs_new.go | 17 +++++ object/fetch/treefs_op.go | 28 ++++++++ object/fetch/treefs_open.go | 122 ++++++++++++++++++++++++++++++++++ object/fetch/treefs_path.go | 29 ++++++++ object/fetch/treefs_readdir.go | 20 ++++++ object/fetch/treefs_readfile.go | 40 +++++++++++ object/fetch/treefs_stat.go | 22 ++++++ object/fetch/treefs_sub.go | 22 ++++++ object/fetch/treefs_test.go | 111 +++++++++++++++++++++++++++++++ 38 files changed, 1274 insertions(+) create mode 100644 object/fetch/doc.go create mode 100644 object/fetch/exact_blob.go create mode 100644 object/fetch/exact_blob_reader.go create mode 100644 object/fetch/exact_commit.go create mode 100644 object/fetch/exact_commit_reader.go create mode 100644 object/fetch/exact_object.go create mode 100644 object/fetch/exact_reader.go create mode 100644 object/fetch/exact_tag.go create mode 100644 object/fetch/exact_tag_reader.go create mode 100644 object/fetch/exact_tree.go create mode 100644 object/fetch/exact_tree_reader.go create mode 100644 object/fetch/fetcher.go create mode 100644 object/fetch/object_parse.go create mode 100644 object/fetch/path.go create mode 100644 object/fetch/peel_to_blob.go create mode 100644 object/fetch/peel_to_blob_id.go create mode 100644 object/fetch/peel_to_blob_reader.go create mode 100644 object/fetch/peel_to_commit.go create mode 100644 object/fetch/peel_to_commit_id.go create mode 100644 object/fetch/peel_to_commit_reader.go create mode 100644 object/fetch/peel_to_tag.go create mode 100644 object/fetch/peel_to_tag_id.go create mode 100644 object/fetch/peel_to_tag_reader.go create mode 100644 object/fetch/peel_to_tree.go create mode 100644 object/fetch/peel_to_tree_id.go create mode 100644 object/fetch/peel_to_tree_reader.go create mode 100644 object/fetch/treefs.go create mode 100644 object/fetch/treefs_entry.go create mode 100644 object/fetch/treefs_info.go create mode 100644 object/fetch/treefs_new.go create mode 100644 object/fetch/treefs_op.go create mode 100644 object/fetch/treefs_open.go create mode 100644 object/fetch/treefs_path.go create mode 100644 object/fetch/treefs_readdir.go create mode 100644 object/fetch/treefs_readfile.go create mode 100644 object/fetch/treefs_stat.go create mode 100644 object/fetch/treefs_sub.go create mode 100644 object/fetch/treefs_test.go (limited to 'object/fetch') diff --git a/object/fetch/doc.go b/object/fetch/doc.go new file mode 100644 index 00000000..e5ca33a9 --- /dev/null +++ b/object/fetch/doc.go @@ -0,0 +1,5 @@ +// Package resolve resolves stored Git objects by exact type, by peeling +// tree-ish or commit-ish references, and by path within trees. +// +// A Fetcher does not take ownership of the underlying object store. +package fetch diff --git a/object/fetch/exact_blob.go b/object/fetch/exact_blob.go new file mode 100644 index 00000000..96ededdc --- /dev/null +++ b/object/fetch/exact_blob.go @@ -0,0 +1,24 @@ +package fetch + +import ( + "fmt" + + "codeberg.org/lindenii/furgit/object/blob" + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" +) + +// ExactBlob reads, parses, and wraps the blob at id. +func (r *Fetcher) ExactBlob(id objectid.ObjectID) (*stored.Stored[*blob.Blob], error) { + parsed, err := r.parseObject(id) + if err != nil { + return nil, err + } + + blob, ok := parsed.(*blob.Blob) + if !ok { + return nil, fmt.Errorf("object/fetch: expected blob object %s, got %v", id, parsed.ObjectType()) + } + + return stored.New(id, blob), nil +} diff --git a/object/fetch/exact_blob_reader.go b/object/fetch/exact_blob_reader.go new file mode 100644 index 00000000..95f5e1aa --- /dev/null +++ b/object/fetch/exact_blob_reader.go @@ -0,0 +1,14 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// ExactBlobReader returns a reader for the content of the blob at id, +// together with its content size in bytes. +func (r *Fetcher) ExactBlobReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + return r.exactReader(id, objecttype.TypeBlob, "blob") +} diff --git a/object/fetch/exact_commit.go b/object/fetch/exact_commit.go new file mode 100644 index 00000000..d3f9dee1 --- /dev/null +++ b/object/fetch/exact_commit.go @@ -0,0 +1,24 @@ +package fetch + +import ( + "fmt" + + "codeberg.org/lindenii/furgit/object/commit" + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" +) + +// ExactCommit reads, parses, and wraps the commit at id. +func (r *Fetcher) ExactCommit(id objectid.ObjectID) (*stored.Stored[*commit.Commit], error) { + parsed, err := r.parseObject(id) + if err != nil { + return nil, err + } + + commit, ok := parsed.(*commit.Commit) + if !ok { + return nil, fmt.Errorf("object/fetch: expected commit object %s, got %v", id, parsed.ObjectType()) + } + + return stored.New(id, commit), nil +} diff --git a/object/fetch/exact_commit_reader.go b/object/fetch/exact_commit_reader.go new file mode 100644 index 00000000..5ccae4b1 --- /dev/null +++ b/object/fetch/exact_commit_reader.go @@ -0,0 +1,16 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// ExactCommitReader returns a reader for the content of the commit at id, +// together with its content size in bytes. +// +// Usage of this method is unusual. +func (r *Fetcher) ExactCommitReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + return r.exactReader(id, objecttype.TypeCommit, "commit") +} diff --git a/object/fetch/exact_object.go b/object/fetch/exact_object.go new file mode 100644 index 00000000..869d811f --- /dev/null +++ b/object/fetch/exact_object.go @@ -0,0 +1,18 @@ +package fetch + +import ( + "codeberg.org/lindenii/furgit/object" + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" +) + +// ExactObject reads, parses, and wraps the object at id without constraining +// its concrete object kind. +func (r *Fetcher) ExactObject(id objectid.ObjectID) (*stored.Stored[object.Object], error) { + parsed, err := r.parseObject(id) + if err != nil { + return nil, err + } + + return stored.New(id, parsed), nil +} diff --git a/object/fetch/exact_reader.go b/object/fetch/exact_reader.go new file mode 100644 index 00000000..54cf246c --- /dev/null +++ b/object/fetch/exact_reader.go @@ -0,0 +1,26 @@ +package fetch + +import ( + "fmt" + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// exactReader reads one object's content stream and verifies that its header +// type matches wantType. +func (r *Fetcher) exactReader(id objectid.ObjectID, wantType objecttype.Type, wantName string) (io.ReadCloser, int64, error) { + gotType, size, rc, err := r.store.ReadReaderContent(id) + if err != nil { + return nil, 0, err + } + + if gotType != wantType { + _ = rc.Close() + + return nil, 0, fmt.Errorf("object/fetch: expected %s object %s, got %v", wantName, id, gotType) + } + + return rc, size, nil +} diff --git a/object/fetch/exact_tag.go b/object/fetch/exact_tag.go new file mode 100644 index 00000000..03c66019 --- /dev/null +++ b/object/fetch/exact_tag.go @@ -0,0 +1,24 @@ +package fetch + +import ( + "fmt" + + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" + "codeberg.org/lindenii/furgit/object/tag" +) + +// ExactTag reads, parses, and wraps the tag at id. +func (r *Fetcher) ExactTag(id objectid.ObjectID) (*stored.Stored[*tag.Tag], error) { + parsed, err := r.parseObject(id) + if err != nil { + return nil, err + } + + tag, ok := parsed.(*tag.Tag) + if !ok { + return nil, fmt.Errorf("object/fetch: expected tag object %s, got %v", id, parsed.ObjectType()) + } + + return stored.New(id, tag), nil +} diff --git a/object/fetch/exact_tag_reader.go b/object/fetch/exact_tag_reader.go new file mode 100644 index 00000000..79a31a4c --- /dev/null +++ b/object/fetch/exact_tag_reader.go @@ -0,0 +1,16 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// ExactTagReader returns a reader for the content of the tag at id, +// together with its content size in bytes. +// +// Usage of this method is unusual. +func (r *Fetcher) ExactTagReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + return r.exactReader(id, objecttype.TypeTag, "tag") +} diff --git a/object/fetch/exact_tree.go b/object/fetch/exact_tree.go new file mode 100644 index 00000000..cab10ba2 --- /dev/null +++ b/object/fetch/exact_tree.go @@ -0,0 +1,24 @@ +package fetch + +import ( + "fmt" + + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" + "codeberg.org/lindenii/furgit/object/tree" +) + +// ExactTree reads, parses, and wraps the tree at id. +func (r *Fetcher) ExactTree(id objectid.ObjectID) (*stored.Stored[*tree.Tree], error) { + parsed, err := r.parseObject(id) + if err != nil { + return nil, err + } + + tree, ok := parsed.(*tree.Tree) + if !ok { + return nil, fmt.Errorf("object/fetch: expected tree object %s, got %v", id, parsed.ObjectType()) + } + + return stored.New(id, tree), nil +} diff --git a/object/fetch/exact_tree_reader.go b/object/fetch/exact_tree_reader.go new file mode 100644 index 00000000..4660bd34 --- /dev/null +++ b/object/fetch/exact_tree_reader.go @@ -0,0 +1,16 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// ExactTreeReader returns a reader for the content of the tree at id, +// together with its content size in bytes. +// +// Usage of this method is unusual. +func (r *Fetcher) ExactTreeReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + return r.exactReader(id, objecttype.TypeTree, "tree") +} diff --git a/object/fetch/fetcher.go b/object/fetch/fetcher.go new file mode 100644 index 00000000..ed4e74a1 --- /dev/null +++ b/object/fetch/fetcher.go @@ -0,0 +1,17 @@ +package fetch + +import objectstorer "codeberg.org/lindenii/furgit/object/storer" + +// Fetcher resolves parsed and streamed objects from an object store. +// +// A Fetcher does not take ownership of the store and does not close it. +type Fetcher struct { + store objectstorer.Store +} + +// New returns a Fetcher that reads objects from store. +// +// The returned Fetcher does not take ownership of store. +func New(store objectstorer.Store) *Fetcher { + return &Fetcher{store: store} +} diff --git a/object/fetch/object_parse.go b/object/fetch/object_parse.go new file mode 100644 index 00000000..a29e17c0 --- /dev/null +++ b/object/fetch/object_parse.go @@ -0,0 +1,28 @@ +package fetch + +import ( + "fmt" + + "codeberg.org/lindenii/furgit/object" + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +func (r *Fetcher) parseObject(id objectid.ObjectID) (object.Object, error) { + ty, content, err := r.store.ReadBytesContent(id) + if err != nil { + return nil, err + } + + parsed, err := object.ParseObjectWithoutHeader(ty, content, id.Algorithm()) + if err != nil { + tyName, ok := objecttype.Name(ty) + if !ok { + tyName = fmt.Sprintf("type %d", ty) + } + + return nil, fmt.Errorf("object/fetch: parse object %s (%s): %w", id, tyName, err) + } + + return parsed, nil +} diff --git a/object/fetch/path.go b/object/fetch/path.go new file mode 100644 index 00000000..f26379c3 --- /dev/null +++ b/object/fetch/path.go @@ -0,0 +1,103 @@ +package fetch + +import ( + "fmt" + + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/tree" +) + +// PathEmptyError indicates that Path received no segments. +type PathEmptyError struct{} + +func (err *PathEmptyError) Error() string { + return "object/fetch: empty tree path" +} + +// PathSegmentEmptyError indicates that one path segment is empty. +type PathSegmentEmptyError struct { + Index int +} + +func (err *PathSegmentEmptyError) Error() string { + return fmt.Sprintf("object/fetch: empty tree path segment at index %d", err.Index) +} + +// PathNotFoundError indicates that one tree path segment was not found. +type PathNotFoundError struct { + Index int + Name []byte +} + +func (err *PathNotFoundError) Error() string { + return fmt.Sprintf("object/fetch: tree entry %q not found at index %d", err.Name, err.Index) +} + +// PathNotTreeError indicates that one intermediate path segment was not a tree. +type PathNotTreeError struct { + Index int + Name []byte +} + +func (err *PathNotTreeError) Error() string { + return fmt.Sprintf("object/fetch: path segment %q at index %d is not a tree", err.Name, err.Index) +} + +// Path resolves parts within the tree identified by root and returns the final +// tree entry. +// +// The root object may be any tree-ish object accepted by PeelToTree. +// +// parts must contain at least one path segment. Intermediate path segments +// must resolve to tree entries. The final entry is returned without loading +// its object. Path segments may not contain \x00. +// +// The path cannot be accurately represented as a string or a single []byte +// because Git tree entry names may include slashes. While []string is +// technically possible (since Go strings are not necessarily UTF-8), they +// do often imply UTF-8 in practice, which would be undesirable. +// +// If your entry names are valid UTF-8 and uses / solely as segment separators, +// it may be convenient to use TreeFS for an io/fs.FS-like interface. +func (r *Fetcher) Path(root objectid.ObjectID, parts [][]byte) (tree.TreeEntry, error) { + if len(parts) == 0 { + return tree.TreeEntry{}, &PathEmptyError{} + } + + current, err := r.PeelToTree(root) + if err != nil { + return tree.TreeEntry{}, err + } + + for i, part := range parts { + if len(part) == 0 { + return tree.TreeEntry{}, &PathSegmentEmptyError{Index: i} + } + + entry := current.Object().Entry(part) + if entry == nil { + return tree.TreeEntry{}, &PathNotFoundError{ + Index: i, + Name: append([]byte(nil), part...), + } + } + + if i == len(parts)-1 { + return *entry, nil + } + + if entry.Mode != tree.FileModeDir { + return tree.TreeEntry{}, &PathNotTreeError{ + Index: i, + Name: append([]byte(nil), part...), + } + } + + current, err = r.ExactTree(entry.ID) + if err != nil { + return tree.TreeEntry{}, err + } + } + + return tree.TreeEntry{}, &PathNotFoundError{Index: len(parts) - 1} +} diff --git a/object/fetch/peel_to_blob.go b/object/fetch/peel_to_blob.go new file mode 100644 index 00000000..b0a1ad1c --- /dev/null +++ b/object/fetch/peel_to_blob.go @@ -0,0 +1,29 @@ +package fetch + +import ( + "fmt" + + "codeberg.org/lindenii/furgit/object/blob" + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" + "codeberg.org/lindenii/furgit/object/tag" +) + +// PeelToBlob peels tags until it reaches a blob. +func (r *Fetcher) PeelToBlob(id objectid.ObjectID) (*stored.Stored[*blob.Blob], error) { + for { + obj, err := r.ExactObject(id) + if err != nil { + return nil, err + } + + switch parsed := obj.Object().(type) { + case *blob.Blob: + return stored.New(id, parsed), nil + case *tag.Tag: + id = parsed.Target + default: + return nil, fmt.Errorf("object/fetch: expected blob-ish object %s, got %v", id, parsed.ObjectType()) + } + } +} diff --git a/object/fetch/peel_to_blob_id.go b/object/fetch/peel_to_blob_id.go new file mode 100644 index 00000000..9d13f640 --- /dev/null +++ b/object/fetch/peel_to_blob_id.go @@ -0,0 +1,39 @@ +package fetch + +import ( + "fmt" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// PeelToBlobID peels tags until it reaches a blob object ID. +func (r *Fetcher) PeelToBlobID(id objectid.ObjectID) (objectid.ObjectID, error) { + for { + ty, _, err := r.store.ReadHeader(id) + if err != nil { + return objectid.ObjectID{}, err + } + + switch ty { + case objecttype.TypeBlob: + return id, nil + case objecttype.TypeTag: + tag, err := r.ExactTag(id) + if err != nil { + return objectid.ObjectID{}, err + } + + id = tag.Object().Target + case objecttype.TypeInvalid, + objecttype.TypeCommit, + objecttype.TypeTree, + objecttype.TypeFuture, + objecttype.TypeOfsDelta, + objecttype.TypeRefDelta: + return objectid.ObjectID{}, fmt.Errorf("object/fetch: expected blob-ish object %s, got %v", id, ty) + default: + return objectid.ObjectID{}, fmt.Errorf("object/fetch: expected blob-ish object %s, got %v", id, ty) + } + } +} diff --git a/object/fetch/peel_to_blob_reader.go b/object/fetch/peel_to_blob_reader.go new file mode 100644 index 00000000..d7072531 --- /dev/null +++ b/object/fetch/peel_to_blob_reader.go @@ -0,0 +1,18 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" +) + +// PeelToBlobReader returns a reader for the content of the peeled blob at id, +// together with its content size in bytes. +func (r *Fetcher) PeelToBlobReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + blobID, err := r.PeelToBlobID(id) + if err != nil { + return nil, 0, err + } + + return r.ExactBlobReader(blobID) +} diff --git a/object/fetch/peel_to_commit.go b/object/fetch/peel_to_commit.go new file mode 100644 index 00000000..5ec66411 --- /dev/null +++ b/object/fetch/peel_to_commit.go @@ -0,0 +1,29 @@ +package fetch + +import ( + "fmt" + + "codeberg.org/lindenii/furgit/object/commit" + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" + "codeberg.org/lindenii/furgit/object/tag" +) + +// PeelToCommit peels tags until it reaches a commit. +func (r *Fetcher) PeelToCommit(id objectid.ObjectID) (*stored.Stored[*commit.Commit], error) { + for { + obj, err := r.ExactObject(id) + if err != nil { + return nil, err + } + + switch parsed := obj.Object().(type) { + case *commit.Commit: + return stored.New(id, parsed), nil + case *tag.Tag: + id = parsed.Target + default: + return nil, fmt.Errorf("object/fetch: expected commit-ish object %s, got %v", id, parsed.ObjectType()) + } + } +} diff --git a/object/fetch/peel_to_commit_id.go b/object/fetch/peel_to_commit_id.go new file mode 100644 index 00000000..b6b257cf --- /dev/null +++ b/object/fetch/peel_to_commit_id.go @@ -0,0 +1,39 @@ +package fetch + +import ( + "fmt" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// PeelToCommitID peels tags until it reaches a commit object ID. +func (r *Fetcher) PeelToCommitID(id objectid.ObjectID) (objectid.ObjectID, error) { + for { + ty, _, err := r.store.ReadHeader(id) + if err != nil { + return objectid.ObjectID{}, err + } + + switch ty { + case objecttype.TypeCommit: + return id, nil + case objecttype.TypeTag: + tag, err := r.ExactTag(id) + if err != nil { + return objectid.ObjectID{}, err + } + + id = tag.Object().Target + case objecttype.TypeInvalid, + objecttype.TypeTree, + objecttype.TypeBlob, + objecttype.TypeFuture, + objecttype.TypeOfsDelta, + objecttype.TypeRefDelta: + return objectid.ObjectID{}, fmt.Errorf("object/fetch: expected commit-ish object %s, got %v", id, ty) + default: + return objectid.ObjectID{}, fmt.Errorf("object/fetch: expected commit-ish object %s, got %v", id, ty) + } + } +} diff --git a/object/fetch/peel_to_commit_reader.go b/object/fetch/peel_to_commit_reader.go new file mode 100644 index 00000000..02bf9f6f --- /dev/null +++ b/object/fetch/peel_to_commit_reader.go @@ -0,0 +1,20 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" +) + +// PeelToCommitReader returns a reader for the content of the peeled commit at +// id, together with its content size in bytes. +// +// Usage of this method is unusual. +func (r *Fetcher) PeelToCommitReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + commitID, err := r.PeelToCommitID(id) + if err != nil { + return nil, 0, err + } + + return r.ExactCommitReader(commitID) +} diff --git a/object/fetch/peel_to_tag.go b/object/fetch/peel_to_tag.go new file mode 100644 index 00000000..60693069 --- /dev/null +++ b/object/fetch/peel_to_tag.go @@ -0,0 +1,12 @@ +package fetch + +import ( + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" + "codeberg.org/lindenii/furgit/object/tag" +) + +// PeelToTag returns the tag at id without further peeling. +func (r *Fetcher) PeelToTag(id objectid.ObjectID) (*stored.Stored[*tag.Tag], error) { + return r.ExactTag(id) +} diff --git a/object/fetch/peel_to_tag_id.go b/object/fetch/peel_to_tag_id.go new file mode 100644 index 00000000..dc02c60a --- /dev/null +++ b/object/fetch/peel_to_tag_id.go @@ -0,0 +1,8 @@ +package fetch + +import objectid "codeberg.org/lindenii/furgit/object/id" + +// PeelToTagID returns id unchanged. +func (r *Fetcher) PeelToTagID(id objectid.ObjectID) (objectid.ObjectID, error) { + return id, nil +} diff --git a/object/fetch/peel_to_tag_reader.go b/object/fetch/peel_to_tag_reader.go new file mode 100644 index 00000000..562ef9e8 --- /dev/null +++ b/object/fetch/peel_to_tag_reader.go @@ -0,0 +1,20 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" +) + +// PeelToTagReader returns a reader for the content of the tag at id, +// together with its content size in bytes. +// +// Usage of this method is unusual. +func (r *Fetcher) PeelToTagReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + tagID, err := r.PeelToTagID(id) + if err != nil { + return nil, 0, err + } + + return r.ExactTagReader(tagID) +} diff --git a/object/fetch/peel_to_tree.go b/object/fetch/peel_to_tree.go new file mode 100644 index 00000000..119e76ac --- /dev/null +++ b/object/fetch/peel_to_tree.go @@ -0,0 +1,33 @@ +package fetch + +import ( + "fmt" + + "codeberg.org/lindenii/furgit/object/commit" + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/stored" + "codeberg.org/lindenii/furgit/object/tag" + "codeberg.org/lindenii/furgit/object/tree" +) + +// PeelToTree peels tags until it reaches a tree or commit. If it reaches a +// commit, it returns the commit's root tree. +func (r *Fetcher) PeelToTree(id objectid.ObjectID) (*stored.Stored[*tree.Tree], error) { + for { + obj, err := r.ExactObject(id) + if err != nil { + return nil, err + } + + switch parsed := obj.Object().(type) { + case *tree.Tree: + return stored.New(id, parsed), nil + case *commit.Commit: + return r.ExactTree(parsed.Tree) + case *tag.Tag: + id = parsed.Target + default: + return nil, fmt.Errorf("object/fetch: expected tree-ish object %s, got %v", id, parsed.ObjectType()) + } + } +} diff --git a/object/fetch/peel_to_tree_id.go b/object/fetch/peel_to_tree_id.go new file mode 100644 index 00000000..21dcfde4 --- /dev/null +++ b/object/fetch/peel_to_tree_id.go @@ -0,0 +1,46 @@ +package fetch + +import ( + "fmt" + + objectid "codeberg.org/lindenii/furgit/object/id" + objecttype "codeberg.org/lindenii/furgit/object/type" +) + +// PeelToTreeID peels tags until it reaches a tree object ID, or a commit whose +// root tree object ID is then returned. +func (r *Fetcher) PeelToTreeID(id objectid.ObjectID) (objectid.ObjectID, error) { + for { + ty, _, err := r.store.ReadHeader(id) + if err != nil { + return objectid.ObjectID{}, err + } + + switch ty { + case objecttype.TypeTree: + return id, nil + case objecttype.TypeCommit: + commit, err := r.ExactCommit(id) + if err != nil { + return objectid.ObjectID{}, err + } + + return commit.Object().Tree, nil + case objecttype.TypeTag: + tag, err := r.ExactTag(id) + if err != nil { + return objectid.ObjectID{}, err + } + + id = tag.Object().Target + case objecttype.TypeInvalid, + objecttype.TypeBlob, + objecttype.TypeFuture, + objecttype.TypeOfsDelta, + objecttype.TypeRefDelta: + return objectid.ObjectID{}, fmt.Errorf("object/fetch: expected tree-ish object %s, got %v", id, ty) + default: + return objectid.ObjectID{}, fmt.Errorf("object/fetch: expected tree-ish object %s, got %v", id, ty) + } + } +} diff --git a/object/fetch/peel_to_tree_reader.go b/object/fetch/peel_to_tree_reader.go new file mode 100644 index 00000000..fe1c5b93 --- /dev/null +++ b/object/fetch/peel_to_tree_reader.go @@ -0,0 +1,20 @@ +package fetch + +import ( + "io" + + objectid "codeberg.org/lindenii/furgit/object/id" +) + +// PeelToTreeReader returns a reader for the content of the peeled tree at id, +// together with its content size in bytes. +// +// Usage of this method is unusual. +func (r *Fetcher) PeelToTreeReader(id objectid.ObjectID) (io.ReadCloser, int64, error) { + treeID, err := r.PeelToTreeID(id) + if err != nil { + return nil, 0, err + } + + return r.ExactTreeReader(treeID) +} diff --git a/object/fetch/treefs.go b/object/fetch/treefs.go new file mode 100644 index 00000000..06b6ad25 --- /dev/null +++ b/object/fetch/treefs.go @@ -0,0 +1,30 @@ +package fetch + +import ( + "io/fs" + + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/tree" +) + +// TreeFS exposes one Git tree as an fs.FS. +// +// TreeFS interprets names using io/fs path rules. Those rules do not match raw +// Git tree entry naming exactly: names are UTF-8, slash-separated, and must be +// valid fs.FS paths. Tree entries that cannot be represented under those rules +// are not addressable through this API. +// +// TreeFS does not take ownership of its Fetcher. +type TreeFS struct { + resolver *Fetcher + rootTree objectid.ObjectID + rootEntry *tree.TreeEntry +} + +var ( + _ fs.FS = (*TreeFS)(nil) + _ fs.ReadFileFS = (*TreeFS)(nil) + _ fs.ReadDirFS = (*TreeFS)(nil) + _ fs.StatFS = (*TreeFS)(nil) + _ fs.SubFS = (*TreeFS)(nil) +) diff --git a/object/fetch/treefs_entry.go b/object/fetch/treefs_entry.go new file mode 100644 index 00000000..4aa7730a --- /dev/null +++ b/object/fetch/treefs_entry.go @@ -0,0 +1,90 @@ +package fetch + +import ( + "errors" + "fmt" + "io/fs" + + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/tree" +) + +func (treeFS *TreeFS) resolvePath(op treeFSOp, name string) (treeEntryValue, error) { + if !treeFSValidPath(name) { + return treeEntryValue{}, treeFSPathError(op, name, fs.ErrInvalid) + } + + if name == "." { + return treeEntryValue{ + name: ".", + mode: tree.FileModeDir, + treeID: treeFS.rootTree, + treeEntry: treeFS.rootEntry, + }, nil + } + + entry, err := treeFS.resolver.Path(treeFS.rootTree, treeFSSplitPath(name)) + if err != nil { + return treeEntryValue{}, treeFS.pathResolveError(op, name, err) + } + + return treeEntryValue{ + name: string(entry.Name), + mode: entry.Mode, + objectID: entry.ID, + treeEntry: &entry, + }, nil +} + +func (treeFS *TreeFS) pathResolveError(op treeFSOp, name string, err error) error { + if _, ok := errors.AsType[*PathNotFoundError](err); ok { + return treeFSPathError(op, name, fs.ErrNotExist) + } + + if _, ok := errors.AsType[*PathNotTreeError](err); ok { + return treeFSPathError(op, name, fs.ErrInvalid) + } + + if _, ok := errors.AsType[*PathEmptyError](err); ok { + return treeFSPathError(op, name, fs.ErrInvalid) + } + + if _, ok := errors.AsType[*PathSegmentEmptyError](err); ok { + return treeFSPathError(op, name, fs.ErrInvalid) + } + + return treeFSPathError(op, name, err) +} + +type treeEntryValue struct { + name string + mode tree.FileMode + objectID objectid.ObjectID + treeID objectid.ObjectID + treeEntry *tree.TreeEntry +} + +func (entry treeEntryValue) isDir() bool { + return entry.mode == tree.FileModeDir +} + +func (entry treeEntryValue) blobSize(resolve *Fetcher) (int64, error) { + _, size, err := resolve.store.ReadHeader(entry.objectID) + if err != nil { + return 0, err + } + + return size, nil +} + +func (entry treeEntryValue) subtreeID() (objectid.ObjectID, error) { + if entry.name == "." { + return entry.treeID, nil + } + + if entry.mode != tree.FileModeDir { + return objectid.ObjectID{}, fmt.Errorf("object/fetch: path %q is not a tree", entry.name) + } + + return entry.objectID, nil +} diff --git a/object/fetch/treefs_info.go b/object/fetch/treefs_info.go new file mode 100644 index 00000000..eecf183d --- /dev/null +++ b/object/fetch/treefs_info.go @@ -0,0 +1,75 @@ +package fetch + +import ( + "io/fs" + "time" + + "codeberg.org/lindenii/furgit/object/tree" +) + +type treeFSInfo struct { + name string + mode fs.FileMode + size int64 + sys any + isDir bool +} + +var ( + _ fs.FileInfo = (*treeFSInfo)(nil) + _ fs.DirEntry = (*treeFSInfo)(nil) +) + +func (info *treeFSInfo) Name() string { return info.name } +func (info *treeFSInfo) Size() int64 { return info.size } +func (info *treeFSInfo) Mode() fs.FileMode { return info.mode } +func (info *treeFSInfo) Type() fs.FileMode { return info.mode.Type() } +func (info *treeFSInfo) IsDir() bool { return info.isDir } +func (info *treeFSInfo) ModTime() time.Time { return time.Time{} } +func (info *treeFSInfo) Sys() any { return info.sys } +func (info *treeFSInfo) Info() (fs.FileInfo, error) { + return info, nil +} + +func treeFSEntryMode(mode tree.FileMode) fs.FileMode { + switch mode { + case tree.FileModeDir: + return fs.ModeDir | 0o555 + case tree.FileModeRegular: + return 0o444 + case tree.FileModeExecutable: + return 0o555 + case tree.FileModeSymlink: + return fs.ModeSymlink | 0o444 + case tree.FileModeGitlink: + return fs.ModeIrregular + default: + return fs.ModeIrregular + } +} + +func (treeFS *TreeFS) statEntry(entry treeEntryValue) (*treeFSInfo, error) { + size := int64(0) + + if entry.mode == tree.FileModeRegular || entry.mode == tree.FileModeExecutable || entry.mode == tree.FileModeSymlink { + var err error + + size, err = entry.blobSize(treeFS.resolver) + if err != nil { + return nil, err + } + } + + var sys any + if entry.treeEntry != nil { + sys = *entry.treeEntry + } + + return &treeFSInfo{ + name: entry.name, + mode: treeFSEntryMode(entry.mode), + size: size, + sys: sys, + isDir: entry.isDir(), + }, nil +} diff --git a/object/fetch/treefs_new.go b/object/fetch/treefs_new.go new file mode 100644 index 00000000..d2224fcf --- /dev/null +++ b/object/fetch/treefs_new.go @@ -0,0 +1,17 @@ +package fetch + +import objectid "codeberg.org/lindenii/furgit/object/id" + +// TreeFS returns a new filesystem view rooted at root, which may be any +// tree-ish object accepted by PeelToTreeID. +func (r *Fetcher) TreeFS(root objectid.ObjectID) (*TreeFS, error) { + rootTree, err := r.PeelToTreeID(root) + if err != nil { + return nil, err + } + + return &TreeFS{ + resolver: r, + rootTree: rootTree, + }, nil +} diff --git a/object/fetch/treefs_op.go b/object/fetch/treefs_op.go new file mode 100644 index 00000000..f0472923 --- /dev/null +++ b/object/fetch/treefs_op.go @@ -0,0 +1,28 @@ +package fetch + +type treeFSOp uint8 + +const ( + treeFSOpOpen treeFSOp = iota + treeFSOpReadFile + treeFSOpReadDir + treeFSOpStat + treeFSOpSub +) + +func (op treeFSOp) pathErrorOp() string { + switch op { + case treeFSOpOpen: + return "open" + case treeFSOpReadFile: + return "readfile" + case treeFSOpReadDir: + return "readdir" + case treeFSOpStat: + return "stat" + case treeFSOpSub: + return "sub" + default: + return "treefs" + } +} diff --git a/object/fetch/treefs_open.go b/object/fetch/treefs_open.go new file mode 100644 index 00000000..59c6ec5d --- /dev/null +++ b/object/fetch/treefs_open.go @@ -0,0 +1,122 @@ +package fetch + +import ( + "fmt" + "io" + "io/fs" + + "codeberg.org/lindenii/furgit/object/tree" +) + +// Open opens name for reading. +// +// Directories are returned as fs.ReadDirFile values. Gitlink entries are not +// readable through TreeFS. +func (treeFS *TreeFS) Open(name string) (fs.File, error) { + entry, err := treeFS.resolvePath(treeFSOpOpen, name) + if err != nil { + return nil, err + } + + info, err := treeFS.statEntry(entry) + if err != nil { + return nil, treeFSPathError(treeFSOpOpen, name, err) + } + + if entry.isDir() { + treeID, err := entry.subtreeID() + if err != nil { + return nil, treeFSPathError(treeFSOpOpen, name, err) + } + + tree, err := treeFS.resolver.ExactTree(treeID) + if err != nil { + return nil, treeFSPathError(treeFSOpOpen, name, err) + } + + entries := make([]fs.DirEntry, 0, len(tree.Object().Entries)) + for _, child := range tree.Object().Entries { + childEntry := treeEntryValue{ + name: string(child.Name), + mode: child.Mode, + objectID: child.ID, + treeEntry: &child, + } + + childInfo, err := treeFS.statEntry(childEntry) + if err != nil { + return nil, treeFSPathError(treeFSOpOpen, name, err) + } + + entries = append(entries, childInfo) + } + + return &treeFSDir{ + info: info, + entries: entries, + }, nil + } + + if entry.mode == tree.FileModeGitlink { + return nil, treeFSPathError(treeFSOpOpen, name, fmt.Errorf("object/fetch: gitlink entries are not readable as files")) + } + + reader, _, err := treeFS.resolver.ExactBlobReader(entry.objectID) + if err != nil { + return nil, treeFSPathError(treeFSOpOpen, name, err) + } + + return &treeFSBlob{ + info: info, + reader: reader, + }, nil +} + +type treeFSBlob struct { + info *treeFSInfo + reader io.ReadCloser +} + +var _ fs.File = (*treeFSBlob)(nil) + +func (file *treeFSBlob) Stat() (fs.FileInfo, error) { return file.info, nil } +func (file *treeFSBlob) Read(p []byte) (int, error) { return file.reader.Read(p) } +func (file *treeFSBlob) Close() error { return file.reader.Close() } + +type treeFSDir struct { + info *treeFSInfo + entries []fs.DirEntry + offset int +} + +var ( + _ fs.File = (*treeFSDir)(nil) + _ fs.ReadDirFile = (*treeFSDir)(nil) +) + +func (dir *treeFSDir) Stat() (fs.FileInfo, error) { return dir.info, nil } +func (dir *treeFSDir) Close() error { return nil } + +func (dir *treeFSDir) Read(_ []byte) (int, error) { + return 0, fs.ErrInvalid +} + +func (dir *treeFSDir) ReadDir(n int) ([]fs.DirEntry, error) { + if dir.offset >= len(dir.entries) && n > 0 { + return nil, io.EOF + } + + if n <= 0 { + out := append([]fs.DirEntry(nil), dir.entries[dir.offset:]...) + dir.offset = len(dir.entries) + + return out, nil + } + + end := min(dir.offset+n, len(dir.entries)) + + out := append([]fs.DirEntry(nil), dir.entries[dir.offset:end]...) + dir.offset = end + + return out, nil +} diff --git a/object/fetch/treefs_path.go b/object/fetch/treefs_path.go new file mode 100644 index 00000000..7aa31c7f --- /dev/null +++ b/object/fetch/treefs_path.go @@ -0,0 +1,29 @@ +package fetch + +import ( + "io/fs" + "strings" +) + +func treeFSValidPath(name string) bool { + return name == "." || fs.ValidPath(name) +} + +func treeFSSplitPath(name string) [][]byte { + if name == "." { + return nil + } + + parts := strings.Split(name, "/") + + out := make([][]byte, len(parts)) + for i, part := range parts { + out[i] = []byte(part) + } + + return out +} + +func treeFSPathError(op treeFSOp, path string, err error) error { + return &fs.PathError{Op: op.pathErrorOp(), Path: path, Err: err} +} diff --git a/object/fetch/treefs_readdir.go b/object/fetch/treefs_readdir.go new file mode 100644 index 00000000..7518c607 --- /dev/null +++ b/object/fetch/treefs_readdir.go @@ -0,0 +1,20 @@ +package fetch + +import "io/fs" + +// ReadDir reads and returns all directory entries for name. +func (treeFS *TreeFS) ReadDir(name string) ([]fs.DirEntry, error) { + file, err := treeFS.Open(name) + if err != nil { + return nil, err + } + + defer func() { _ = file.Close() }() + + readDirFile, ok := file.(fs.ReadDirFile) + if !ok { + return nil, treeFSPathError(treeFSOpReadDir, name, fs.ErrInvalid) + } + + return readDirFile.ReadDir(-1) +} diff --git a/object/fetch/treefs_readfile.go b/object/fetch/treefs_readfile.go new file mode 100644 index 00000000..d3671f07 --- /dev/null +++ b/object/fetch/treefs_readfile.go @@ -0,0 +1,40 @@ +package fetch + +import ( + "fmt" + "io" + + "codeberg.org/lindenii/furgit/object/tree" +) + +// ReadFile reads the blob contents at name. +// +// Directories and gitlink entries are not readable through TreeFS. +func (treeFS *TreeFS) ReadFile(name string) ([]byte, error) { + entry, err := treeFS.resolvePath(treeFSOpReadFile, name) + if err != nil { + return nil, err + } + + if entry.isDir() { + return nil, treeFSPathError(treeFSOpReadFile, name, fmt.Errorf("is a directory")) + } + + if entry.mode == tree.FileModeGitlink { + return nil, treeFSPathError(treeFSOpReadFile, name, fmt.Errorf("object/fetch: gitlink entries are not readable as files")) + } + + reader, _, err := treeFS.resolver.ExactBlobReader(entry.objectID) + if err != nil { + return nil, treeFSPathError(treeFSOpReadFile, name, err) + } + + defer func() { _ = reader.Close() }() + + data, err := io.ReadAll(reader) + if err != nil { + return nil, treeFSPathError(treeFSOpReadFile, name, err) + } + + return data, nil +} diff --git a/object/fetch/treefs_stat.go b/object/fetch/treefs_stat.go new file mode 100644 index 00000000..7d7a6418 --- /dev/null +++ b/object/fetch/treefs_stat.go @@ -0,0 +1,22 @@ +package fetch + +import "io/fs" + +// Stat returns synthetic file metadata for name. +// +// TreeFS metadata reflects Git tree entry mode and blob size where applicable. +// It does not represent filesystem stat metadata: ModTime is zero, ownership is +// unavailable, and Sys returns the underlying tree.TreeEntry when one exists. +func (treeFS *TreeFS) Stat(name string) (fs.FileInfo, error) { + entry, err := treeFS.resolvePath(treeFSOpStat, name) + if err != nil { + return nil, err + } + + info, err := treeFS.statEntry(entry) + if err != nil { + return nil, treeFSPathError(treeFSOpStat, name, err) + } + + return info, nil +} diff --git a/object/fetch/treefs_sub.go b/object/fetch/treefs_sub.go new file mode 100644 index 00000000..faf47092 --- /dev/null +++ b/object/fetch/treefs_sub.go @@ -0,0 +1,22 @@ +package fetch + +import "io/fs" + +// Sub returns a new TreeFS rooted at dir. +func (treeFS *TreeFS) Sub(dir string) (fs.FS, error) { + entry, err := treeFS.resolvePath(treeFSOpSub, dir) + if err != nil { + return nil, err + } + + treeID, err := entry.subtreeID() + if err != nil { + return nil, treeFSPathError(treeFSOpSub, dir, fs.ErrInvalid) + } + + return &TreeFS{ + resolver: treeFS.resolver, + rootTree: treeID, + rootEntry: entry.treeEntry, + }, nil +} diff --git a/object/fetch/treefs_test.go b/object/fetch/treefs_test.go new file mode 100644 index 00000000..067283f7 --- /dev/null +++ b/object/fetch/treefs_test.go @@ -0,0 +1,111 @@ +package fetch_test + +import ( + "errors" + "io/fs" + "testing" + + "codeberg.org/lindenii/furgit/internal/testgit" + objectid "codeberg.org/lindenii/furgit/object/id" + "codeberg.org/lindenii/furgit/object/fetch" + "codeberg.org/lindenii/furgit/object/tree" + "codeberg.org/lindenii/furgit/repository" +) + +func TestTreeFS(t *testing.T) { + t.Parallel() + + testgit.ForEachAlgorithm(t, func(t *testing.T, algo objectid.Algorithm) { //nolint:thelper + t.Parallel() + + repoData := testgit.NewRepo(t, testgit.RepoOptions{ObjectFormat: algo}) + repoData.WriteFile(t, "plain.txt", []byte("plain\n"), 0o644) + repoData.WriteFileAll(t, "dir/exec.sh", []byte("#!/bin/sh\nexit 0\n"), 0o755, 0o755) + repoData.SymbolicRef(t, "HEAD", "refs/heads/main") + _ = repoData.Run(t, "add", ".") + treeHex := repoData.Run(t, "write-tree") + + treeID, err := objectid.ParseHex(algo, treeHex) + if err != nil { + t.Fatalf("ParseHex(write-tree): %v", err) + } + + commitID := repoData.CommitTree(t, treeID, "treefs") + + root := repoData.OpenGitRoot(t) + + repo, err := repository.Open(root) + if err != nil { + t.Fatalf("repository.Open: %v", err) + } + + defer func() { _ = repo.Close() }() + + resolver := fetch.New(repo.Objects()) + + treeFS, err := resolver.TreeFS(commitID) + if err != nil { + t.Fatalf("resolver.TreeFS: %v", err) + } + + content, err := treeFS.ReadFile("plain.txt") + if err != nil { + t.Fatalf("ReadFile(plain.txt): %v", err) + } + + if string(content) != "plain\n" { + t.Fatalf("ReadFile(plain.txt) = %q, want %q", string(content), "plain\n") + } + + entries, err := treeFS.ReadDir(".") + if err != nil { + t.Fatalf("ReadDir(.): %v", err) + } + + if len(entries) != 2 { + t.Fatalf("len(ReadDir(.)) = %d, want 2", len(entries)) + } + + info, err := treeFS.Stat("plain.txt") + if err != nil { + t.Fatalf("Stat(plain.txt): %v", err) + } + + entry, ok := info.Sys().(tree.TreeEntry) + if !ok { + t.Fatalf("Stat(plain.txt).Sys() type = %T, want tree.TreeEntry", info.Sys()) + } + + if entry.Mode != tree.FileModeRegular { + t.Fatalf("Stat(plain.txt).Sys().Mode = %o, want %o", entry.Mode, tree.FileModeRegular) + } + + subFS, err := treeFS.Sub("dir") + if err != nil { + t.Fatalf("Sub(dir): %v", err) + } + + subReadFileFS, ok := subFS.(fs.ReadFileFS) + if !ok { + t.Fatalf("Sub(dir) type does not implement fs.ReadFileFS") + } + + subContent, err := subReadFileFS.ReadFile("exec.sh") + if err != nil { + t.Fatalf("Sub(dir).ReadFile(exec.sh): %v", err) + } + + if string(subContent) != "#!/bin/sh\nexit 0\n" { + t.Fatalf("Sub(dir).ReadFile(exec.sh) = %q", string(subContent)) + } + + _, err = treeFS.ReadFile("dir") + if err == nil { + t.Fatal("ReadFile(dir) unexpectedly succeeded") + } + + if _, ok := errors.AsType[*fs.PathError](err); !ok { + t.Fatalf("ReadFile(dir) err type = %T, want *fs.PathError", err) + } + }) +} -- cgit v1.3.1-10-gc9f91