Merge pull request #2181 from lightpanda-io/blob_constructor

Support more types in new Blob(...)

src/browser/js/Value.zig

@@ -163,6 +163,53 @@ pub fn isFloat64Array(self: Value) bool {
     return v8.v8__Value__IsFloat64Array(self.handle);
 }
 
+// A few places in the code take various types, but want a string. This is a
+// type-aware version of toString(). If you do:
+//    (new ArrayBuffer(100)).toString()
+// You'll get "[object ArrayBuffer]". But this `toStringSmart()` knows about
+// buffers, and Blobs, etc and will try to return the real underlying string
+// value. It _does_ ultimately fallback to toString() - callers should check
+// for types they _don't_ want before calling this. For example, `Response`
+// checks for null or undefined before calling this to apply specific handling
+// to those cases.
+pub fn toStringSmart(self: Value) ![]const u8 {
+    if (self.isString()) |js_str| {
+        return try js_str.toSlice();
+    }
+
+    const Blob = @import("../webapi/Blob.zig");
+    if (self.local.jsValueToZig(*Blob, self)) |blob_obj| {
+        return blob_obj._slice;
+    } else |_| {}
+
+    var byte_offset: usize = 0;
+    var byte_len: usize = undefined;
+    var array_buffer: ?*const v8.ArrayBuffer = null;
+
+    if (self.isTypedArray() or self.isArrayBufferView()) {
+        const buffer_handle: *const v8.ArrayBufferView = @ptrCast(self.handle);
+        byte_len = v8.v8__ArrayBufferView__ByteLength(buffer_handle);
+        byte_offset = v8.v8__ArrayBufferView__ByteOffset(buffer_handle);
+        array_buffer = v8.v8__ArrayBufferView__Buffer(buffer_handle);
+    } else if (self.isArrayBuffer()) {
+        array_buffer = @ptrCast(self.handle);
+        byte_len = v8.v8__ArrayBuffer__ByteLength(array_buffer);
+    } else {
+        return self.toStringSlice();
+    }
+
+    const backing_store_ptr = v8.v8__ArrayBuffer__GetBackingStore(array_buffer orelse return "");
+    if (byte_len == 0) {
+        return &[_]u8{};
+    }
+
+    const backing_store_handle = v8.std__shared_ptr__v8__BackingStore__get(&backing_store_ptr) orelse return "";
+    const data = v8.v8__BackingStore__Data(backing_store_handle) orelse return "";
+    const base = @as([*]const u8, @ptrCast(data)) + byte_offset;
+
+    return base[0..byte_len];
+}
+
 pub fn isPromise(self: Value) bool {
     return v8.v8__Value__IsPromise(self.handle);
 }

src/browser/tests/blob.html

@@ -79,6 +79,100 @@
   }
 </script>
 
+<script id=parts>
+  // Blob as a blob part - contents are copied in, not stringified.
+  {
+    const inner = new Blob(["hello "], { type: "text/plain" });
+    const blob = new Blob([inner, "world"]);
+    testing.expectEqual(11, blob.size);
+    testing.async(async () => {
+      testing.expectEqual("hello world", await blob.text());
+    });
+  }
+
+  // Uint8Array as a blob part.
+  {
+    const bytes = new Uint8Array([104, 101, 108, 108, 111]); // "hello"
+    const blob = new Blob([bytes, " ", "world"]);
+    testing.expectEqual(11, blob.size);
+    testing.async(async () => {
+      testing.expectEqual("hello world", await blob.text());
+    });
+  }
+
+  // Non-byte TypedArrays contribute their raw backing bytes.
+  {
+    // Uint16Array of one element (0x0041 = 'A' little-endian) produces 2 bytes.
+    const u16 = new Uint16Array([0x0041]);
+    const blob = new Blob([u16]);
+    testing.expectEqual(2, blob.size);
+  }
+
+  {
+    // Float32Array: 4 bytes per element.
+    const f32 = new Float32Array([1.0, 2.0, 3.0]);
+    const blob = new Blob([f32]);
+    testing.expectEqual(12, blob.size);
+  }
+
+  // ArrayBuffer as a blob part.
+  {
+    const buf = new Uint8Array([1, 2, 3, 4]).buffer;
+    const blob = new Blob([buf]);
+    testing.expectEqual(4, blob.size);
+    testing.async(async () => {
+      const result = await blob.bytes();
+      testing.expectEqual(new Uint8Array([1, 2, 3, 4]), result);
+    });
+  }
+
+  // DataView (ArrayBufferView) as a blob part.
+  {
+    const buf = new Uint8Array([10, 20, 30, 40, 50]).buffer;
+    const view = new DataView(buf, 1, 3); // bytes [20, 30, 40]
+    const blob = new Blob([view]);
+    testing.expectEqual(3, blob.size);
+    testing.async(async () => {
+      const result = await blob.bytes();
+      testing.expectEqual(new Uint8Array([20, 30, 40]), result);
+    });
+  }
+
+  // Mixed types in a single parts array.
+  {
+    const inner = new Blob(["bb"]);
+    const bytes = new Uint8Array([99, 99]); // "cc"
+    const buf = new Uint8Array([100, 100]).buffer; // "dd"
+    const blob = new Blob(["aa", inner, bytes, buf, "ee"]);
+    testing.expectEqual(10, blob.size);
+    testing.async(async () => {
+      testing.expectEqual("aabbccddee", await blob.text());
+    });
+  }
+
+  // Number coerces to string.
+  {
+    const blob = new Blob([42]);
+    testing.expectEqual(2, blob.size);
+    testing.async(async () => {
+      testing.expectEqual("42", await blob.text());
+    });
+  }
+
+  // Empty parts array.
+  {
+    const blob = new Blob([]);
+    testing.expectEqual(0, blob.size);
+    testing.expectEqual("", blob.type);
+  }
+
+  // No arguments.
+  {
+    const blob = new Blob();
+    testing.expectEqual(0, blob.size);
+  }
+</script>
+
 <script id=stream>
   {
     const parts = ["may", "thy", "knife", "chip", "and", "shatter"];

src/browser/webapi/Blob.zig

@@ -59,69 +59,23 @@ const InitOptions = struct {
     endings: []const u8 = "transparent",
 };
 
-/// Creates a new Blob (JS constructor).
-pub fn init(
-    maybe_blob_parts: ?[]const []const u8,
-    maybe_options: ?InitOptions,
-    page: *Page,
-) !*Blob {
-    return initWithMimeValidation(maybe_blob_parts, maybe_options, false, page);
-}
-
-/// Creates a new Blob with optional MIME validation.
-/// When validate_mime is true, uses full MIME parsing (for Response/Request).
-/// When false, uses simple ASCII validation per FileAPI spec (for Blob constructor).
-pub fn initWithMimeValidation(
-    maybe_blob_parts: ?[]const []const u8,
-    maybe_options: ?InitOptions,
-    validate_mime: bool,
-    page: *Page,
-) !*Blob {
-    const data_len = blk: {
-        const parts = maybe_blob_parts orelse break :blk 0;
-        var size: usize = 0;
-        for (parts) |p| {
-            size += p.len;
-        }
-        break :blk size;
-    };
-    const arena = try page.getArena(256 + data_len, "Blob");
+/// Creates a new Blob from JS values with optional MIME validation.
+/// This is the JS Constructor
+pub fn init(parts_: ?[]const js.Value, opts_: ?InitOptions, page: *Page) !*Blob {
+    const arena = try page.getArena(.large, "Blob");
     errdefer page.releaseArena(arena);
 
-    const options: InitOptions = maybe_options orelse .{};
-
-    const mime: []const u8 = blk: {
-        const t = options.type;
-        if (t.len == 0) {
-            break :blk "";
-        }
-
-        const buf = try arena.dupe(u8, t);
-
-        if (validate_mime) {
-            // Full MIME parsing per MIME sniff spec (for Content-Type headers)
-            _ = Mime.parse(buf) catch break :blk "";
-        } else {
-            // Simple validation per FileAPI spec (for Blob constructor):
-            // - If any char is outside U+0020-U+007E, return empty string
-            // - Otherwise lowercase
-            for (t) |c| {
-                if (c < 0x20 or c > 0x7E) {
-                    break :blk "";
-                }
-            }
-            _ = std.ascii.lowerString(buf, buf);
-        }
-
-        break :blk buf;
-    };
+    const opts: InitOptions = opts_ orelse .{};
+    const mime = try validateMimeType(arena, opts.type, false);
 
     const data = blk: {
-        if (maybe_blob_parts) |blob_parts| {
+        if (parts_) |blob_parts| {
+            const use_native_endings = std.mem.eql(u8, opts.endings, "native");
             var w: Writer.Allocating = .init(arena);
-            const use_native_endings = std.mem.eql(u8, options.endings, "native");
-            try writeBlobParts(&w.writer, blob_parts, use_native_endings);
-
+            for (blob_parts) |js_val| {
+                const part = try js_val.toStringSmart();
+                try writePartWithEndings(part, use_native_endings, &w.writer);
+            }
             break :blk w.written();
         }
 
@@ -139,6 +93,50 @@ pub fn initWithMimeValidation(
     return self;
 }
 
+/// Creates a new Blob from raw byte slices (for internal Zig use).
+pub fn initFromBytes(data: []const u8, content_type: []const u8, validate_mime: bool, page: *Page) !*Blob {
+    const arena = try page.getArena(.large, "Blob");
+    errdefer page.releaseArena(arena);
+
+    const mime = try validateMimeType(arena, content_type, validate_mime);
+
+    const self = try arena.create(Blob);
+    self.* = .{
+        ._rc = .{},
+        ._arena = arena,
+        ._type = .generic,
+        ._slice = try arena.dupe(u8, data),
+        ._mime = mime,
+    };
+    return self;
+}
+
+/// Validates and normalizes MIME type according to spec.
+fn validateMimeType(arena: Allocator, mime_type: []const u8, full_validation: bool) ![]const u8 {
+    if (mime_type.len == 0) {
+        return "";
+    }
+
+    const buf = try arena.dupe(u8, mime_type);
+
+    if (full_validation) {
+        // Full MIME parsing per MIME sniff spec (for Content-Type headers)
+        _ = Mime.parse(buf) catch return "";
+    } else {
+        // Simple validation per FileAPI spec (for Blob constructor):
+        // - If any char is outside U+0020-U+007E, return empty string
+        // - Otherwise lowercase
+        for (mime_type) |c| {
+            if (c < 0x20 or c > 0x7E) {
+                return "";
+            }
+        }
+        _ = std.ascii.lowerString(buf, buf);
+    }
+
+    return buf;
+}
+
 pub fn deinit(self: *Blob, session: *Session) void {
     session.releaseArena(self._arena);
 }
@@ -171,18 +169,11 @@ const vector_sizes = blk: {
     break :blk items;
 };
 
-/// Writes blob parts to given `Writer` with desired endings.
-fn writeBlobParts(
-    writer: *Writer,
-    blob_parts: []const []const u8,
-    use_native_endings: bool,
-) !void {
-    // Transparent.
+/// Writes a single part with optional line ending normalization.
+fn writePartWithEndings(part: []const u8, use_native_endings: bool, writer: *Writer) !void {
+    // Transparent - no conversion needed.
     if (!use_native_endings) {
-        for (blob_parts) |part| {
-            try writer.writeAll(part);
-        }
-
+        try writer.writeAll(part);
         return;
     }
 
@@ -204,68 +195,66 @@ fn writeBlobParts(
     // ```
     // "the quick\n\nbrown fox"
     // ```
-    scan_parts: for (blob_parts) |part| {
-        var end: usize = 0;
+    var end: usize = 0;
 
-        inline for (vector_sizes) |vector_len| {
-            const Vec = @Vector(vector_len, u8);
+    inline for (vector_sizes) |vector_len| {
+        const Vec = @Vector(vector_len, u8);
 
-            while (end + vector_len <= part.len) : (end += vector_len) {
-                const cr: Vec = @splat('\r');
-                // Load chunk as vectors.
-                const data = part[end..][0..vector_len];
-                const chunk: Vec = data.*;
-                // Look for CR.
-                const match = chunk == cr;
+        while (end + vector_len <= part.len) : (end += vector_len) {
+            const cr: Vec = @splat('\r');
+            // Load chunk as vectors.
+            const data = part[end..][0..vector_len];
+            const chunk: Vec = data.*;
+            // Look for CR.
+            const match = chunk == cr;
 
-                // Create a bitset out of match vector.
-                const bitset = std.bit_set.IntegerBitSet(vector_len){
-                    .mask = @bitCast(@intFromBool(match)),
-                };
+            // Create a bitset out of match vector.
+            const bitset = std.bit_set.IntegerBitSet(vector_len){
+                .mask = @bitCast(@intFromBool(match)),
+            };
 
-                var iter = bitset.iterator(.{});
-                var relative_start: usize = 0;
-                while (iter.next()) |index| {
-                    _ = try writer.writeVec(&.{ data[relative_start..index], "\n" });
+            var iter = bitset.iterator(.{});
+            var relative_start: usize = 0;
+            while (iter.next()) |index| {
+                _ = try writer.writeVec(&.{ data[relative_start..index], "\n" });
 
-                    if (index + 1 != data.len and data[index + 1] == '\n') {
-                        relative_start = index + 2;
-                    } else {
-                        relative_start = index + 1;
-                    }
-                }
-
-                _ = try writer.writeVec(&.{data[relative_start..]});
-            }
-        }
-
-        // Scalar scan fallback.
-        var relative_start: usize = end;
-        while (end < part.len) {
-            if (part[end] == '\r') {
-                _ = try writer.writeVec(&.{ part[relative_start..end], "\n" });
-
-                // Part ends with CR. We can continue to next part.
-                if (end + 1 == part.len) {
-                    continue :scan_parts;
-                }
-
-                // If next char is LF, skip it too.
-                if (part[end + 1] == '\n') {
-                    relative_start = end + 2;
+                if (index + 1 != data.len and data[index + 1] == '\n') {
+                    relative_start = index + 2;
                 } else {
-                    relative_start = end + 1;
+                    relative_start = index + 1;
                 }
             }
 
-            end += 1;
+            _ = try writer.writeVec(&.{data[relative_start..]});
+        }
+    }
+
+    // Scalar scan fallback.
+    var relative_start: usize = end;
+    while (end < part.len) {
+        if (part[end] == '\r') {
+            _ = try writer.writeVec(&.{ part[relative_start..end], "\n" });
+
+            // Part ends with CR. We need to remember this for next part.
+            if (end + 1 == part.len) {
+                return;
+            }
+
+            // If next char is LF, skip it too.
+            if (part[end + 1] == '\n') {
+                relative_start = end + 2;
+            } else {
+                relative_start = end + 1;
+            }
         }
 
-        // Write the remaining. We get this in such situations:
-        // `the quick brown\rfox`
-        // `the quick brown\r\nfox`
-        try writer.writeAll(part[relative_start..end]);
+        end += 1;
     }
+
+    // Write the remaining. We get this in such situations:
+    // `the quick brown\rfox`
+    // `the quick brown\r\nfox`
+    try writer.writeAll(part[relative_start..end]);
 }
 
 /// Returns a Promise that resolves with the contents of the blob
@@ -323,7 +312,7 @@ pub fn slice(
         break :blk @min(data.len, @max(start, @as(u31, @intCast(requested_end))));
     };
 
-    return Blob.init(&.{data[start..end]}, .{ .type = content_type_ orelse "" }, page);
+    return Blob.initFromBytes(data[start..end], content_type_ orelse "", false, page);
 }
 
 /// Returns the size of the Blob in bytes.

src/browser/webapi/net/Request.zig

@@ -174,12 +174,7 @@ pub fn blob(self: *Request, page: *Page) !js.Promise {
     const headers = try self.getHeaders(page);
     const content_type = try headers.get("content-type", page) orelse "";
 
-    const b = try Blob.initWithMimeValidation(
-        &.{body},
-        .{ .type = content_type },
-        true,
-        page,
-    );
+    const b = try Blob.initFromBytes(body, content_type, true, page);
 
     return page.js.local.?.resolvePromise(b);
 }

src/browser/webapi/net/Response.zig

@@ -83,29 +83,10 @@ pub fn init(body_: ?BodyInit, opts_: ?InitOpts, page: *Page) !*Response {
             .bytes => |body_bytes| break :blk .{ .bytes = try arena.dupe(u8, body_bytes) },
             .stream => |stream| break :blk .{ .stream = stream },
             .js_val => |js_val| {
-                const local = page.js.local.?;
-
-                if (local.jsValueToZig(*ReadableStream, js_val)) |stream| {
-                    break :blk .{ .stream = stream };
-                } else |_| {}
-
-                if (js_val.isString()) |js_str| {
-                    break :blk .{ .bytes = try js_str.toSliceWithAlloc(arena) };
-                }
-
-                if (js_val.isArrayBuffer() or js_val.isTypedArray() or js_val.isArrayBufferView()) {
-                    if (local.jsValueToZig([]u8, js_val)) |data| {
-                        break :blk .{ .bytes = try arena.dupe(u8, data) };
-                    } else |_| {}
-                }
-
-                if (local.jsValueToZig(*Blob, js_val)) |blob_obj| {
-                    break :blk .{ .bytes = try arena.dupe(u8, blob_obj._slice) };
-                } else |_| {}
-
-                if (js_val.isNullOrUndefined() == false) {
-                    break :blk .{ .bytes = try js_val.toStringSliceWithAlloc(arena) };
+                if (js_val.isNullOrUndefined()) {
+                    break :blk .empty;
                 }
+                break :blk .{ .bytes = try arena.dupe(u8, try js_val.toStringSmart()) };
             },
         }
         break :blk .empty;
@@ -335,14 +316,7 @@ pub fn blob(self: *const Response, page: *Page) !js.Promise {
         .stream => return local.rejectPromise(.{ .type_error = "Cannot read blob from stream body" }),
     };
     const content_type = try self._headers.get("content-type", page) orelse "";
-
-    const b = try Blob.initWithMimeValidation(
-        &.{body},
-        .{ .type = content_type },
-        true,
-        page,
-    );
-
+    const b = try Blob.initFromBytes(body, content_type, true, page);
     return local.resolvePromise(b);
 }
 

src/browser/webapi/net/WebSocket.zig

@@ -466,7 +466,7 @@ fn dispatchMessageEvent(self: *WebSocket, data: []const u8, frame_type: http.WsF
             switch (self._binary_type) {
                 .arraybuffer => .{ .arraybuffer = .{ .values = data } },
                 .blob => blk: {
-                    const blob = try Blob.init(&.{data}, .{}, page);
+                    const blob = try Blob.initFromBytes(data, "", false, page);
                     blob.acquireRef();
                     break :blk .{ .blob = blob };
                 },