Merge pull request #4 from ryleelyman/upstream

fix!: `addMethod` now passes errors it receives; added test

src/block.zig

@@ -36,6 +36,7 @@ pub fn Block(
         const Self = @This();
         const captures_info = @typeInfo(Captures).Struct;
         const InvokeFn = FnType(anyopaque);
+        const signature: [:0]const u8 = objc.comptimeEncode(InvokeFn);
 
         /// This is the function type that is called back.
         pub const Fn = FnType(Context);
@@ -64,8 +65,6 @@ pub fn Block(
                 @field(ctx, field.name) = @field(captures, field.name);
             }
 
-            const signature = try createFnSignature(Return, @typeInfo(InvokeFn).Fn.params);
-            errdefer alloc.free(signature);
             var descriptor = try alloc.create(Descriptor);
             errdefer alloc.destroy(descriptor);
             descriptor.* = .{
@@ -205,61 +204,6 @@ fn BlockContext(comptime Captures: type, comptime InvokeFn: type) type {
     });
 }
 
-/// Creates the function signature expected by the descriptor.
-pub fn createFnSignature(
-    comptime Return: type,
-    comptime Args: []const std.builtin.Type.Fn.Param,
-) ![:0]const u8 {
-    var list = try std.ArrayList(u8).initCapacity(alloc, 1024);
-    defer list.deinit();
-    try typeSignature(&list, Return);
-    inline for (Args) |arg| try typeSignature(&list, arg.type.?);
-    return try list.toOwnedSliceSentinel(0);
-}
-
-fn typeSignature(list: *std.ArrayList(u8), comptime T: type) !void {
-    switch (T) {
-        objc.Object, objc.c.id => try list.append('@'),
-        objc.Class, objc.c.Class => try list.append('#'),
-        objc.Sel, objc.c.SEL => try list.append(':'),
-        bool => try list.append('B'),
-        u8 => try list.append('C'),
-        i8 => try list.append('c'),
-        u16 => try list.append('S'),
-        i16 => try list.append('s'),
-        u32, c_uint => try list.append('I'),
-        i32, c_int => try list.append('i'),
-        u64, c_ulong, c_ulonglong => try list.append('Q'),
-        i64, c_long, c_longlong => try list.append('q'),
-        f32 => try list.append('f'),
-        f64 => try list.append('d'),
-        [:0]const u8, [*c]const u8, [:0]u8, [*c]u8 => try list.append('*'),
-        void => try list.append('v'),
-        else => {
-            const type_info = @typeInfo(T);
-            switch (type_info) {
-                .Struct => |s| {
-                    try list.appendSlice("{?=");
-                    inline for (s.fields) |field| try typeSignature(list, field.type);
-                    try list.appendSlice("}");
-                },
-                .Union => |u| {
-                    try list.appendSlice("(?=");
-                    inline for (u.fields) |field| try typeSignature(list, field.type);
-                    try list.appendSlice(")");
-                },
-                .Pointer => |p| {
-                    try list.append('^');
-                    try typeSignature(list, p.child);
-                },
-                .Fn => try list.append('?'),
-                .Opaque => try list.append('v'),
-                else => @compileError("unsupported type for typeSignature: " ++ @typeName(T)),
-            }
-        },
-    }
-}
-
 const NSConcreteStackBlock = @extern(*anyopaque, .{ .name = "_NSConcreteStackBlock" });
 extern "C" fn _Block_object_assign(dst: *anyopaque, src: *const anyopaque, flag: c_int) void;
 extern "C" fn _Block_object_dispose(src: *const anyopaque, flag: c_int) void;

src/c.zig

@@ -2,3 +2,14 @@ pub usingnamespace @cImport({
     @cInclude("objc/runtime.h");
     @cInclude("objc/message.h");
 });
+
+/// This is a funky helper to help with the fact that some macOS
+/// SDKs have an i8 return value for bools and some have stdbool.
+pub fn boolResult(comptime Fn: type, result: anytype) bool {
+    const fn_info = @typeInfo(Fn).Fn;
+    return switch (fn_info.return_type.?) {
+        bool => result,
+        i8 => result == 1,
+        else => @compileError("unhandled class_addIvar return type"),
+    };
+}

src/class.zig

@@ -63,29 +63,29 @@ pub const Class = struct {
     /// allows adding new methods; returns true on success.
     // imp should be a function with C calling convention
     // whose first two arguments are a `c.id` and a `c.SEL`.
-    pub fn addMethod(self: Class, name: [:0]const u8, imp: anytype) bool {
-        const fn_info = @typeInfo(@TypeOf(imp)).Fn;
+    pub fn addMethod(self: Class, name: [:0]const u8, imp: anytype) !bool {
+        const Fn = @TypeOf(imp);
+        const fn_info = @typeInfo(Fn).Fn;
         assert(fn_info.calling_convention == .C);
         assert(fn_info.is_var_args == false);
         assert(fn_info.params.len >= 2);
         assert(fn_info.params[0].type == c.id);
         assert(fn_info.params[1].type == c.SEL);
-        const str = objc.encodeFn(fn_info.return_type.?, fn_info.params) catch @panic("OOM!");
-        defer std.heap.raw_c_allocator.free(str);
-        return c.class_addMethod(self.value, objc.sel(name).value, @ptrCast(&imp), str.ptr);
+        const encoding = comptime objc.comptimeEncode(Fn);
+        return c.boolResult(@TypeOf(c.class_addMethod), c.class_addMethod(
+            self.value,
+            objc.sel(name).value,
+            @ptrCast(&imp),
+            encoding.ptr,
+        ));
     }
 
     // only call this function between allocateClassPair and registerClassPair
     // this adds an Ivar of type `id`.
     pub fn addIvar(self: Class, name: [:0]const u8) bool {
         // The return type is i8 when we're cross compiling, unsure why.
-        const fn_info = @typeInfo(@TypeOf(c.class_addIvar)).Fn;
         const result = c.class_addIvar(self.value, name, @sizeOf(c.id), @alignOf(c.id), "@");
-        return switch (fn_info.return_type.?) {
-            bool => result,
-            i8 => result == 1,
-            else => @compileError("unhandled class_addIvar return type"),
-        };
+        return c.boolResult(@TypeOf(c.class_addIvar), result);
     }
 };
 
@@ -195,3 +195,23 @@ test "Ivars" {
     const slice = std.mem.sliceTo(my_ivar.getProperty([*c]const u8, "UTF8String"), 0);
     try testing.expectEqualSlices(u8, "69---nice", slice);
 }
+
+test "addMethod" {
+    const testing = std.testing;
+    const My_Class = setup: {
+        const My_Class = allocateClassPair(objc.getClass("NSObject").?, "my_class").?;
+        defer registerClassPair(My_Class);
+        std.debug.assert(try My_Class.addMethod("my_addition", struct {
+            fn imp(target: objc.c.id, sel: objc.c.SEL, a: i32, b: i32) callconv(.C) i32 {
+                _ = sel;
+                _ = target;
+                return a + b;
+            }
+        }.imp));
+        break :setup My_Class;
+    };
+    const result = My_Class.msgSend(objc.Object, "alloc", .{})
+        .msgSend(objc.Object, "init", .{})
+        .msgSend(i32, "my_addition", .{ @as(i32, 2), @as(i32, 3) });
+    try testing.expectEqual(@as(i32, 5), result);
+}

src/encoding.zig

@@ -0,0 +1,364 @@
+const std = @import("std");
+const objc = @import("main.zig");
+const c = @import("c.zig");
+const assert = std.debug.assert;
+const testing = std.testing;
+
+/// Encode a type into a comptime string.
+pub fn comptimeEncode(comptime T: type) [:0]const u8 {
+    comptime {
+        const encoding = objc.Encoding.init(T);
+
+        // Figure out how much space we need
+        var counting = std.io.countingWriter(std.io.null_writer);
+        try std.fmt.format(counting.writer(), "{}", .{encoding});
+
+        // Build our final signature
+        var buf: [counting.bytes_written + 1]u8 = undefined;
+        var fbs = std.io.fixedBufferStream(buf[0..counting.bytes_written]);
+        try std.fmt.format(fbs.writer(), "{}", .{encoding});
+        buf[counting.bytes_written] = 0;
+
+        return buf[0..counting.bytes_written :0];
+    }
+}
+
+/// Encoding union which parses type information and turns it into Obj-C
+/// runtime Type Encodings.
+///
+/// https://developer.apple.com/library/archive/documentation/Cocoa/Conceptual/ObjCRuntimeGuide/Articles/ocrtTypeEncodings.html
+pub const Encoding = union(enum) {
+    char,
+    int,
+    short,
+    long,
+    longlong,
+    uchar,
+    uint,
+    ushort,
+    ulong,
+    ulonglong,
+    float,
+    double,
+    bool,
+    void,
+    char_string,
+    object,
+    class,
+    selector,
+    array: struct { arr_type: type, len: usize },
+    structure: struct { struct_type: type, show_type_spec: bool },
+    @"union": struct { union_type: type, show_type_spec: bool },
+    bitfield: u32,
+    pointer: struct { ptr_type: type, size: std.builtin.Type.Pointer.Size },
+    function: std.builtin.Type.Fn,
+    unknown,
+
+    pub fn init(comptime T: type) Encoding {
+        return switch (T) {
+            i8, c_char => .char,
+            c_short => .short,
+            i32, c_int => .int,
+            c_long => .long,
+            i64, c_longlong => .longlong,
+            u8 => .uchar,
+            c_ushort => .ushort,
+            u32, c_uint => .uint,
+            c_ulong => .ulong,
+            u64, c_ulonglong => .ulonglong,
+            f32 => .float,
+            f64 => .double,
+            bool => .bool,
+            void, anyopaque => .void,
+            [*c]u8, [*c]const u8 => .char_string,
+            c.SEL, objc.Sel => .selector,
+            c.Class, objc.Class => .class,
+            c.id, objc.Object => .object,
+            else => switch (@typeInfo(T)) {
+                .Array => |arr| .{ .array = .{ .len = arr.len, .arr_type = arr.child } },
+                .Struct => .{ .structure = .{ .struct_type = T, .show_type_spec = true } },
+                .Union => .{ .@"union" = .{
+                    .union_type = T,
+                    .show_type_spec = true,
+                } },
+                .Pointer => |ptr| .{ .pointer = .{ .ptr_type = T, .size = ptr.size } },
+                .Fn => |fn_info| .{ .function = fn_info },
+                else => @compileError("unsupported type: " ++ @typeName(T)),
+            },
+        };
+    }
+
+    pub fn format(
+        comptime self: Encoding,
+        comptime fmt: []const u8,
+        options: std.fmt.FormatOptions,
+        writer: anytype,
+    ) !void {
+        switch (self) {
+            .char => try writer.writeAll("c"),
+            .int => try writer.writeAll("i"),
+            .short => try writer.writeAll("s"),
+            .long => try writer.writeAll("l"),
+            .longlong => try writer.writeAll("q"),
+            .uchar => try writer.writeAll("C"),
+            .uint => try writer.writeAll("I"),
+            .ushort => try writer.writeAll("S"),
+            .ulong => try writer.writeAll("L"),
+            .ulonglong => try writer.writeAll("Q"),
+            .float => try writer.writeAll("f"),
+            .double => try writer.writeAll("d"),
+            .bool => try writer.writeAll("B"),
+            .void => try writer.writeAll("v"),
+            .char_string => try writer.writeAll("*"),
+            .object => try writer.writeAll("@"),
+            .class => try writer.writeAll("#"),
+            .selector => try writer.writeAll(":"),
+            .array => |a| {
+                try writer.print("[{}", .{a.len});
+                const encode_type = init(a.arr_type);
+                try encode_type.format(fmt, options, writer);
+                try writer.writeAll("]");
+            },
+            .structure => |s| {
+                const struct_info = @typeInfo(s.struct_type);
+                assert(struct_info.Struct.layout == .Extern);
+
+                // Strips the fully qualified type name to leave just the
+                // type name. Used in naming the Struct in an encoding.
+                var type_name_iter = std.mem.splitBackwardsScalar(u8, @typeName(s.struct_type), '.');
+                const type_name = type_name_iter.first();
+                try writer.print("{{{s}", .{type_name});
+
+                // if the encoding should show the internal type specification
+                // of the struct (determined by levels of pointer indirection)
+                if (s.show_type_spec) {
+                    try writer.writeAll("=");
+                    inline for (struct_info.Struct.fields) |field| {
+                        const field_encode = init(field.type);
+                        try field_encode.format(fmt, options, writer);
+                    }
+                }
+
+                try writer.writeAll("}");
+            },
+            .@"union" => |u| {
+                const union_info = @typeInfo(u.union_type);
+                assert(union_info.Union.layout == .Extern);
+
+                // Strips the fully qualified type name to leave just the
+                // type name. Used in naming the Union in an encoding
+                var type_name_iter = std.mem.splitBackwardsScalar(u8, @typeName(u.union_type), '.');
+                const type_name = type_name_iter.first();
+                try writer.print("({s}", .{type_name});
+
+                // if the encoding should show the internal type specification
+                // of the Union (determined by levels of pointer indirection)
+                if (u.show_type_spec) {
+                    try writer.writeAll("=");
+                    inline for (union_info.Union.fields) |field| {
+                        const field_encode = init(field.type);
+                        try field_encode.format(fmt, options, writer);
+                    }
+                }
+
+                try writer.writeAll(")");
+            },
+            .bitfield => |b| try writer.print("b{}", .{b}), // not sure if needed from Zig -> Obj-C
+            .pointer => |p| {
+                switch (p.size) {
+                    .One => {
+                        // get the pointer info (count of levels of direction
+                        // and the underlying type)
+                        const pointer_info = indirectionCountAndType(p.ptr_type);
+                        for (0..pointer_info.indirection_levels) |_| {
+                            try writer.writeAll("^");
+                        }
+
+                        // create a new Encoding union from the pointers child
+                        // type, giving an encoding of the underlying pointer type
+                        comptime var encoding = init(pointer_info.child);
+
+                        // if the indirection levels are greater than 1, for
+                        // certain types that means getting rid of it's
+                        // internal type specification
+                        //
+                        // https://developer.apple.com/library/archive/documentation/Cocoa/Conceptual/ObjCRuntimeGuide/Articles/ocrtTypeEncodings.html#//apple_ref/doc/uid/TP40008048-CH100
+                        if (pointer_info.indirection_levels > 1) {
+                            switch (encoding) {
+                                .structure => |*s| s.show_type_spec = false,
+                                .@"union" => |*u| u.show_type_spec = false,
+                                else => {},
+                            }
+                        }
+
+                        // call this format function again, this time with the child type encoding
+                        try encoding.format(fmt, options, writer);
+                    },
+                    else => @compileError("Pointer size not supported for encoding"),
+                }
+            },
+            .function => |fn_info| {
+                assert(fn_info.calling_convention == .C);
+
+                // Return type is first in a method encoding
+                const ret_type_enc = init(fn_info.return_type.?);
+                try ret_type_enc.format(fmt, options, writer);
+                inline for (fn_info.params) |param| {
+                    const param_enc = init(param.type.?);
+                    try param_enc.format(fmt, options, writer);
+                }
+            },
+            .unknown => {},
+        }
+    }
+};
+
+/// This comptime function gets the levels of indirection from a type. If the type is a pointer type it
+/// returns the underlying type from the pointer (the child) by walking the pointer to that child.
+/// Returns the type and 0 for count if the type isn't a pointer
+fn indirectionCountAndType(comptime T: type) struct {
+    child: type,
+    indirection_levels: comptime_int,
+} {
+    var WalkType = T;
+    var count: usize = 0;
+    while (@typeInfo(WalkType) == .Pointer) : (count += 1) {
+        WalkType = @typeInfo(WalkType).Pointer.child;
+    }
+
+    return .{ .child = WalkType, .indirection_levels = count };
+}
+
+fn encodingMatchesType(comptime T: type, expected_encoding: []const u8) !void {
+    var buf: [200]u8 = undefined;
+    const enc = Encoding.init(T);
+    const enc_string = try std.fmt.bufPrint(&buf, "{s}", .{enc});
+    try testing.expectEqualStrings(expected_encoding, enc_string);
+}
+
+test "i8 to Encoding.char encoding" {
+    try encodingMatchesType(i8, "c");
+}
+
+test "c_char to Encoding.char encoding" {
+    try encodingMatchesType(c_char, "c");
+}
+
+test "c_short to Encoding.short encoding" {
+    try encodingMatchesType(c_short, "s");
+}
+
+test "c_int to Encoding.int encoding" {
+    try encodingMatchesType(c_int, "i");
+}
+
+test "c_long to Encoding.long encoding" {
+    try encodingMatchesType(c_long, "l");
+}
+
+test "c_longlong to Encoding.longlong encoding" {
+    try encodingMatchesType(c_longlong, "q");
+}
+
+test "u8 to Encoding.uchar encoding" {
+    try encodingMatchesType(u8, "C");
+}
+
+test "c_ushort to Encoding.ushort encoding" {
+    try encodingMatchesType(c_ushort, "S");
+}
+
+test "c_uint to Encoding.uint encoding" {
+    try encodingMatchesType(c_uint, "I");
+}
+
+test "c_ulong to Encoding.ulong encoding" {
+    try encodingMatchesType(c_ulong, "L");
+}
+
+test "c_ulonglong to Encoding.ulonglong encoding" {
+    try encodingMatchesType(c_ulonglong, "Q");
+}
+
+test "f32 to Encoding.float encoding" {
+    try encodingMatchesType(f32, "f");
+}
+
+test "f64 to Encoding.double encoding" {
+    try encodingMatchesType(f64, "d");
+}
+
+test "[4]i8 to Encoding.array encoding" {
+    try encodingMatchesType([4]i8, "[4c]");
+}
+
+test "*u8 to Encoding.pointer encoding" {
+    try encodingMatchesType(*u8, "^C");
+}
+
+test "**u8 to Encoding.pointer encoding" {
+    try encodingMatchesType(**u8, "^^C");
+}
+
+test "*TestStruct to Encoding.pointer encoding" {
+    const TestStruct = extern struct {
+        float: f32,
+        char: u8,
+    };
+    try encodingMatchesType(*TestStruct, "^{TestStruct=fC}");
+}
+
+test "**TestStruct to Encoding.pointer encoding" {
+    const TestStruct = extern struct {
+        float: f32,
+        char: u8,
+    };
+    try encodingMatchesType(**TestStruct, "^^{TestStruct}");
+}
+
+test "*TestStruct with 2 level indirection NestedStruct to Encoding.pointer encoding" {
+    const NestedStruct = extern struct {
+        char: i8,
+    };
+    const TestStruct = extern struct {
+        float: f32,
+        char: u8,
+        nested: **NestedStruct,
+    };
+    try encodingMatchesType(*TestStruct, "^{TestStruct=fC^^{NestedStruct}}");
+}
+
+test "Union to Encoding.union encoding" {
+    const TestUnion = extern union {
+        int: c_int,
+        short: c_short,
+        long: c_long,
+    };
+    try encodingMatchesType(TestUnion, "(TestUnion=isl)");
+}
+
+test "*Union to Encoding.union encoding" {
+    const TestUnion = extern union {
+        int: c_int,
+        short: c_short,
+        long: c_long,
+    };
+    try encodingMatchesType(*TestUnion, "^(TestUnion=isl)");
+}
+
+test "**Union to Encoding.union encoding" {
+    const TestUnion = extern union {
+        int: c_int,
+        short: c_short,
+        long: c_long,
+    };
+    try encodingMatchesType(**TestUnion, "^^(TestUnion)");
+}
+
+test "Fn to Encoding.function encoding" {
+    const test_fn = struct {
+        fn add(_: c.id, _: c.SEL, _: i8) callconv(.C) void {}
+    };
+
+    try encodingMatchesType(@TypeOf(test_fn.add), "v@:c");
+}

src/main.zig

@@ -2,12 +2,13 @@ const std = @import("std");
 
 pub const c = @import("c.zig");
 pub usingnamespace @import("autorelease.zig");
+pub usingnamespace @import("block.zig");
 pub usingnamespace @import("class.zig");
+pub usingnamespace @import("encoding.zig");
 pub usingnamespace @import("object.zig");
 pub usingnamespace @import("property.zig");
 pub usingnamespace @import("protocol.zig");
 pub usingnamespace @import("sel.zig");
-pub usingnamespace @import("block.zig");
 
 /// This just calls the C allocator free. Some things need to be freed
 /// and this is how they can be freed for objc.