fetch: update API usage

Author: andrewrk

lib/std/Io/Reader.zig

@@ -74,14 +74,18 @@ pub const VTable = struct {
     ///
     /// `data` may not contain an alias to `Reader.buffer`.
     ///
+    /// `data` is mutable because the implementation may to temporarily modify
+    /// the fields in order to handle partial reads. Implementations must
+    /// restore the original value before returning.
+    ///
     /// Implementations may ignore `data`, writing directly to `Reader.buffer`,
     /// modifying `seek` and `end` accordingly, and returning 0 from this
     /// function. Implementations are encouraged to take advantage of this if
     /// it simplifies the logic.
     ///
     /// The default implementation calls `stream` with either `data[0]` or
     /// `Reader.buffer`, whichever is bigger.
-    readVec: *const fn (r: *Reader, data: []const []u8) Error!usize = defaultReadVec,
+    readVec: *const fn (r: *Reader, data: [][]u8) Error!usize = defaultReadVec,
 
     /// Ensures `capacity` more data can be buffered without rebasing.
     ///
@@ -446,8 +450,8 @@ pub fn bufferedLen(r: *const Reader) usize {
     return r.end - r.seek;
 }
 
-pub fn hashed(r: *Reader, hasher: anytype) Hashed(@TypeOf(hasher)) {
-    return .{ .in = r, .hasher = hasher };
+pub fn hashed(r: *Reader, hasher: anytype, buffer: []u8) Hashed(@TypeOf(hasher)) {
+    return .init(r, hasher, buffer);
 }
 
 pub fn readVecAll(r: *Reader, data: [][]u8) Error!void {
@@ -1764,15 +1768,16 @@ pub fn Hashed(comptime Hasher: type) type {
     return struct {
         in: *Reader,
         hasher: Hasher,
-        interface: Reader,
+        reader: Reader,
 
         pub fn init(in: *Reader, hasher: Hasher, buffer: []u8) @This() {
             return .{
                 .in = in,
                 .hasher = hasher,
-                .interface = .{
+                .reader = .{
                     .vtable = &.{
-                        .read = @This().read,
+                        .stream = @This().stream,
+                        .readVec = @This().readVec,
                         .discard = @This().discard,
                     },
                     .buffer = buffer,
@@ -1782,33 +1787,39 @@ pub fn Hashed(comptime Hasher: type) type {
             };
         }
 
-        fn read(r: *Reader, w: *Writer, limit: Limit) StreamError!usize {
-            const this: *@This() = @alignCast(@fieldParentPtr("interface", r));
-            const data = w.writableVector(limit);
+        fn stream(r: *Reader, w: *Writer, limit: Limit) StreamError!usize {
+            const this: *@This() = @alignCast(@fieldParentPtr("reader", r));
+            const data = limit.slice(try w.writableSliceGreedy(1));
+            var vec: [1][]u8 = .{data};
+            const n = try this.in.readVec(&vec);
+            this.hasher.update(data[0..n]);
+            w.advance(n);
+            return n;
+        }
+
+        fn readVec(r: *Reader, data: [][]u8) Error!usize {
+            const this: *@This() = @alignCast(@fieldParentPtr("reader", r));
             const n = try this.in.readVec(data);
-            const result = w.advanceVector(n);
             var remaining: usize = n;
             for (data) |slice| {
                 if (remaining < slice.len) {
                     this.hasher.update(slice[0..remaining]);
-                    return result;
+                    return n;
                 } else {
                     remaining -= slice.len;
                     this.hasher.update(slice);
                 }
             }
             assert(remaining == 0);
-            return result;
+            return n;
         }
 
         fn discard(r: *Reader, limit: Limit) Error!usize {
-            const this: *@This() = @alignCast(@fieldParentPtr("interface", r));
-            var w = this.hasher.writer(&.{});
-            const n = this.in.stream(&w, limit) catch |err| switch (err) {
-                error.WriteFailed => unreachable,
-                else => |e| return e,
-            };
-            return n;
+            const this: *@This() = @alignCast(@fieldParentPtr("reader", r));
+            const peeked = limit.slice(try this.in.peekGreedy(1));
+            this.hasher.update(peeked);
+            this.in.toss(peeked.len);
+            return peeked.len;
         }
     };
 }

lib/std/Io/Writer.zig

@@ -2296,6 +2296,8 @@ pub fn fixedDrain(w: *Writer, data: []const []const u8, splat: usize) Error!usiz
 /// generic. A better solution will involve creating a writer for each hash
 /// function, where the splat buffer can be tailored to the hash implementation
 /// details.
+///
+/// Contrast with `Hashing` which terminates the stream pipeline.
 pub fn Hashed(comptime Hasher: type) type {
     return struct {
         out: *Writer,
@@ -2368,6 +2370,52 @@ pub fn Hashed(comptime Hasher: type) type {
     };
 }
 
+/// Provides a `Writer` implementation based on calling `Hasher.update`,
+/// discarding all data.
+///
+/// This implementation makes suboptimal buffering decisions due to being
+/// generic. A better solution will involve creating a writer for each hash
+/// function, where the splat buffer can be tailored to the hash implementation
+/// details.
+///
+/// The total number of bytes written is stored in `hasher`.
+///
+/// Contrast with `Hashed` which also passes the data to an underlying stream.
+pub fn Hashing(comptime Hasher: type) type {
+    return struct {
+        hasher: Hasher,
+        writer: Writer,
+
+        pub fn init(buffer: []u8) @This() {
+            return .initHasher(.init(.{}), buffer);
+        }
+
+        pub fn initHasher(hasher: Hasher, buffer: []u8) @This() {
+            return .{
+                .hasher = hasher,
+                .writer = .{
+                    .buffer = buffer,
+                    .vtable = &.{ .drain = @This().drain },
+                },
+            };
+        }
+
+        fn drain(w: *Writer, data: []const []const u8, splat: usize) Error!usize {
+            const this: *@This() = @alignCast(@fieldParentPtr("writer", w));
+            const hasher = &this.hasher;
+            hasher.update(w.buffered());
+            w.end = 0;
+            var n: usize = 0;
+            for (data[0 .. data.len - 1]) |slice| {
+                hasher.update(slice);
+                n += slice.len;
+            }
+            for (0..splat) |_| hasher.update(data[data.len - 1]);
+            return n + splat;
+        }
+    };
+}
+
 /// Maintains `Writer` state such that it writes to the unused capacity of an
 /// array list, filling it up completely before making a call through the
 /// vtable, causing a resize. Consequently, the same, optimized, non-generic

lib/std/net.zig

@@ -1932,7 +1932,8 @@ pub const Stream = struct {
 
             fn stream(io_r: *Io.Reader, io_w: *Io.Writer, limit: Io.Limit) Io.Reader.StreamError!usize {
                 const dest = limit.slice(try io_w.writableSliceGreedy(1));
-                const n = try readVec(io_r, &.{dest});
+                var bufs: [1][]u8 = .{dest};
+                const n = try readVec(io_r, &bufs);
                 io_w.advance(n);
                 return n;
             }

src/Package/Fetch.zig

@@ -1394,18 +1394,20 @@ fn unpackGitPack(f: *Fetch, out_dir: fs.Dir, resource: *Resource.Git) anyerror!U
 
         var index_file = try pack_dir.createFile("pkg.idx", .{ .read = true });
         defer index_file.close();
+        var index_file_buffer: [2000]u8 = undefined;
+        var index_file_writer = index_file.writer(&index_file_buffer);
         {
             const index_prog_node = f.prog_node.start("Index pack", 0);
             defer index_prog_node.end();
-            var index_buffered_writer = std.io.bufferedWriter(index_file.deprecatedWriter());
-            try git.indexPack(gpa, object_format, &pack_file_reader, index_buffered_writer.writer());
-            try index_buffered_writer.flush();
+            try git.indexPack(gpa, object_format, &pack_file_reader, &index_file_writer);
         }
 
         {
+            var index_file_reader = index_file.reader(&index_file_buffer);
             const checkout_prog_node = f.prog_node.start("Checkout", 0);
             defer checkout_prog_node.end();
-            var repository = try git.Repository.init(gpa, object_format, &pack_file_reader, index_file);
+            var repository: git.Repository = undefined;
+            try repository.init(gpa, object_format, &pack_file_reader, &index_file_reader);
             defer repository.deinit();
             var diagnostics: git.Diagnostics = .{ .allocator = arena };
             try repository.checkout(out_dir, resource.want_oid, &diagnostics);