ona/source/kym/bytecode.zig

const coral = @import("coral");

const tokens = @import("./tokens.zig");

pub const Chunk = struct {
	constant_buffer: Buffer,
	bytecode_buffer: Buffer,
	constants: Constants,
	locals: SmallStack(Local, .{.name = "", .depth = 0}) = .{},

	const Buffer = coral.stack.Dense(u8);

	const Constants = coral.stack.Dense(Constant);

	const Local = struct {
		name: []const u8,
		depth: u16,

		const empty = Local{ .name = "", .depth = 0 };
	};

	pub fn compile(self: *Chunk, script: []const u8) ParseError!void {
		self.reset();

		var tokenizer = tokens.Tokenizer{.source = script};

		errdefer self.reset();

		var parser = Parser{
			.chunk = self,
			.tokenizer = &tokenizer,
		};

		while (true) {
			parser.step() catch |step_error| switch (step_error) {
				error.UnexpectedEnd => return,
			};

			try parser.parse_statement();
		}
	}

	fn declare_local(self: *Chunk, name: []const u8) !void {
		return self.locals.push(.{
			.name = name,
			.depth = 0,
		});
	}

	pub fn deinit(self: *Chunk) void {
		self.bytecode_buffer.deinit();
		self.constant_buffer.deinit();
		self.constants.deinit();
	}

	fn emit_byte(self: *Chunk, byte: u8) !void {
		return self.bytecode_buffer.push_one(byte);
	}

	fn emit_opcode(self: *Chunk, opcode: Opcode) !void {
		return self.bytecode_buffer.push_one(@enumToInt(opcode));
	}

	fn emit_operand(self: *Chunk, operand: Operand) !void {
		return self.bytecode_buffer.push_all(coral.io.bytes_of(&operand));
	}

	pub fn fetch_byte(self: Chunk, cursor: *usize) ?u8 {
		if (cursor.* >= self.bytecode_buffer.values.len) return null;

		defer cursor.* += 1;

		return self.bytecode_buffer.values[cursor.*];
	}

	pub fn fetch_constant(self: Chunk, cursor: *usize) ?*Constant {
		return &self.constants.values[self.fetch_operand(cursor) orelse return null];
	}

	pub fn fetch_opcode(self: Chunk, cursor: *usize) ?Opcode {
		return @intToEnum(Opcode, self.fetch_byte(cursor) orelse return null);
	}

	pub fn fetch_operand(self: Chunk, cursor: *usize) ?Operand {
		const operand_size = @sizeOf(Operand);
		const updated_cursor = cursor.* + operand_size;

		if (updated_cursor > self.bytecode_buffer.values.len) return null;

		var operand_bytes align(@alignOf(Operand)) = [_]u8{0} ** operand_size;

		coral.io.copy(&operand_bytes, self.bytecode_buffer.values[cursor.* .. updated_cursor]);

		cursor.* = updated_cursor;

		return @bitCast(Operand, operand_bytes);
	}

	pub fn init(allocator: coral.io.MemoryAllocator) !Chunk {
		const page_size = 1024;
		var constant_buffer = try Buffer.init(allocator, page_size);

		errdefer constant_buffer.deinit();

		const assumed_average_bytecode_size = 1024;
		var bytecode_buffer = try Buffer.init(allocator, assumed_average_bytecode_size);

		errdefer bytecode_buffer.deinit();

		const assumed_average_constant_count = 512;
		var constants = try Constants.init(allocator, assumed_average_constant_count);

		errdefer constants.deinit();

		return Chunk{
			.constant_buffer = constant_buffer,
			.bytecode_buffer = bytecode_buffer,
			.constants = constants,
		};
	}

	fn intern_string(self: *Chunk, string: []const u8) !u64 {
		var constant_slot = @as(u64, 0);

		for (self.constants.values) |interned_constant| {
			switch (interned_constant) {
				.string => |interned_string| if (coral.io.equals(interned_string, string)) return constant_slot,
			}

			constant_slot += 1;
		}

		const constant_allocator = coral.stack.as_dense_allocator(&self.constant_buffer);
		const allocation = constant_allocator.allocate_many(u8, string.len + 1) orelse return error.OutOfMemory;

		errdefer constant_allocator.deallocate(allocation);

		// Zero-terminate string.
		allocation[string.len] = 0;

		// Write string contents.
		{
			const allocated_string = allocation[0 .. string.len];

			coral.io.copy(allocated_string, string);
			try self.constants.push_one(.{.string = @ptrCast([:0]u8, allocated_string)});
		}

		return constant_slot;
	}

	pub fn reset(self: *Chunk) void {
		self.bytecode_buffer.clear();
		self.constant_buffer.clear();
	}

	pub fn resolve_local(self: *Chunk, name: []const u8) ?u8 {
		var count = @as(u8, self.locals.buffer.len);

		while (count != 0) {
			const index = count - 1;

			if (coral.io.equals(name, self.locals.buffer[index].name)) return index;

			count = index;
		}

		return null;
	}
};

pub const Constant = union (enum) {
	string: [:0]u8,
};

pub const Opcode = enum(u8) {
	pop,
	push_nil,
	push_true,
	push_false,
	push_zero,
	push_integer,
	push_float,
	push_string,
	push_array,
	push_table,

	not,
	neg,
	add,
	sub,
	div,
	mul,

	call,
	get_index,
	set_index,
	get_x,
	set_x,
	get_y,
	set_y,
	get_global,
	set_global,
	get_local,
	set_local,
};

pub const Operand = u64;

pub const ParseError = Parser.StepError || tokens.Token.ExpectError || error {
	OutOfMemory,
	IntOverflow,
	UndefinedLocal,
};

const Parser = struct {
	chunk: *Chunk,
	tokenizer: *tokens.Tokenizer,
	current_token: tokens.Token = .newline,
	previous_token: tokens.Token = .newline,

	const Operator = enum {
		not,
		negate,
		add,
		subtract,
		divide,
		multiply,

		const Self = @This();

		fn opcode(self: Self) Opcode {
			return switch (self) {
				.not => .not,
				.negate => .neg,
				.add => .add,
				.subtract => .sub,
				.multiply => .mul,
				.divide => .div,
			};
		}

		fn precedence(self: Self) isize {
			return switch (self) {
				.not => 13,
				.negate => 13,
				.add => 11,
				.subtract => 11,
				.divide => 12,
				.multiply => 12,
			};
		}
	};

	const OperatorStack = SmallStack(Operator, .not);

	const StepError = error {
		UnexpectedEnd,
	};

	const operator_tokens = &.{.symbol_assign, .symbol_plus, .symbol_dash, .symbol_asterisk, .symbol_forward_slash};

	fn parse_expression(self: *Parser) ParseError!void {
		var operators = OperatorStack{};
		var local_depth = @as(usize, 0);

		while (true) {
			switch (self.current_token) {
				.keyword_nil => {
					try self.previous_token.expect_any(operator_tokens);
					try self.chunk.emit_opcode(.push_nil);

					self.step() catch |step_error| switch (step_error) {
						error.UnexpectedEnd => return,
					};
				},

				.keyword_true => {
					try self.previous_token.expect_any(operator_tokens);
					try self.chunk.emit_opcode(.push_true);

					self.step() catch |step_error| switch (step_error) {
						error.UnexpectedEnd => return,
					};
				},

				.keyword_false => {
					try self.previous_token.expect_any(operator_tokens);
					try self.chunk.emit_opcode(.push_false);

					self.step() catch |step_error| switch (step_error) {
						error.UnexpectedEnd => return,
					};
				},

				.integer_literal => |literal| {
					try self.previous_token.expect_any(operator_tokens);

					const value = coral.utf8.parse_signed(@bitSizeOf(i64), literal)
						catch |parse_error| switch (parse_error) {
							error.BadSyntax => unreachable,
							error.IntOverflow => return error.IntOverflow,
						};

					if (value == 0) {
						try self.chunk.emit_opcode(.push_zero);
					} else {
						try self.chunk.emit_opcode(.push_integer);
						try self.chunk.emit_operand(@bitCast(u64, value));
					}

					try self.step();
				},

				.real_literal => |literal| {
					try self.previous_token.expect_any(operator_tokens);

					try self.chunk.emit_operand(@bitCast(u64, coral.utf8.parse_float(@bitSizeOf(f64), literal)
						catch |parse_error| switch (parse_error) {
							// Already validated to be a real by the tokenizer so this cannot fail, as real syntax is a
							// subset of float syntax.
							error.BadSyntax => unreachable,
						}));

					try self.step();
				},

				.string_literal => |literal| {
					try self.previous_token.expect_any(operator_tokens);
					try self.chunk.emit_opcode(.push_string);
					try self.chunk.emit_operand(try self.chunk.intern_string(literal));
					try self.step();
				},

				.global_identifier, .local_identifier => {
					try self.previous_token.expect_any(&.{.symbol_assign, .symbol_plus,
						.symbol_dash, .symbol_asterisk, .symbol_forward_slash, .symbol_period});

					try self.step();
				},

				.symbol_bang => {
					try self.previous_token.expect_any(operator_tokens);
					try operators.push(.not);
					try self.step();

					local_depth = 0;
				},

				.symbol_plus => {
					try self.parse_operator(&operators, .add);

					local_depth = 0;
				},

				.symbol_dash => {
					try self.parse_operator(&operators, .subtract);

					local_depth = 0;
				},

				.symbol_asterisk => {
					try self.parse_operator(&operators, .multiply);

					local_depth = 0;
				},

				.symbol_forward_slash => {
					try self.parse_operator(&operators, .divide);

					local_depth = 0;
				},

				.symbol_period => {
					switch (self.previous_token) {
						.global_identifier => |identifier| {
							try self.chunk.emit_opcode(.get_global);
							try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
						},

						.local_identifier => |identifier| {
							if (local_depth == 0) {
								try self.chunk.emit_byte(self.chunk.resolve_local(identifier) orelse {
									return error.UndefinedLocal;
								});
							} else {
								try self.chunk.emit_opcode(.get_index);
								try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
							}
						},

						else => return error.UnexpectedToken,
					}

					try self.step();

					local_depth += 1;
				},

				.symbol_paren_left => {
					switch (self.previous_token) {
						.local_identifier => |identifier| {
							if (local_depth == 0) {
								try self.chunk.emit_byte(self.chunk.resolve_local(identifier) orelse {
									return error.UndefinedLocal;
								});
							} else {
								try self.chunk.emit_opcode(.get_index);
								try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
							}
						},

						.global_identifier => |identifier| {
							try self.chunk.emit_opcode(.get_global);
							try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
						},

						else => {
							try self.parse_expression();
							try self.previous_token.expect(.symbol_paren_right);
							try self.step();

							local_depth = 0;

							continue;
						},
					}

					local_depth += 1;

					var argument_count = @as(Operand, 0);

					while (true) {
						try self.step();

						try switch (self.current_token) {
							.symbol_paren_right => break,
							else => self.parse_expression(),
						};

						switch (self.previous_token) {
							.symbol_paren_right => break,
							.symbol_comma => {},
							else => return error.UnexpectedToken,
						}

						argument_count += 1;
					}

					try self.chunk.emit_opcode(.call);
					try self.chunk.emit_operand(argument_count);
					try self.step();

					local_depth = 0;
				},

				.symbol_brace_left => {
					const is_call_argument = switch (self.previous_token) {
						.local_identifier, .global_identifier => true,
						else => false,
					};

					var field_count = @as(Operand, 0);

					while (true) {
						try self.step();

						switch (self.current_token) {
							.newline => {},

							.local_identifier => {
								// Create local copy of identifier because step() will overwrite captures.
								const interned_identifier =
									try self.chunk.intern_string(self.current_token.local_identifier);

								try self.chunk.emit_opcode(.push_string);
								try self.chunk.emit_operand(interned_identifier);
								try self.step();

								switch (self.current_token) {
									.symbol_assign => {
										try self.parse_expression();

										field_count += 1;
									},

									.symbol_brace_right => {
										try self.chunk.emit_opcode(.push_string);
										try self.chunk.emit_operand(interned_identifier);

										field_count += 1;

										break;
									},

									.symbol_comma => {
										try self.chunk.emit_opcode(.push_string);
										try self.chunk.emit_operand(interned_identifier);

										field_count += 1;
									},

									else => return error.UnexpectedToken,
								}
							},

							.symbol_brace_right => break,
							else => return error.UnexpectedToken,
						}
					}

					if (is_call_argument) {
						try self.chunk.emit_opcode(.call);
						try self.chunk.emit_operand(1);
					}
				},

				else => {
					try self.previous_token.expect_any(&.{.keyword_nil, .keyword_true, .keyword_false, .integer_literal,
						.real_literal, .string_literal, .global_identifier, .local_identifier, .symbol_brace_right,
							.symbol_paren_right});

					while (operators.pop()) |operator| {
						try self.chunk.emit_opcode(operator.opcode());
					}

					return;
				},
			}
		}
	}

	fn parse_operator(self: *Parser, operators: *OperatorStack, rhs_operator: Operator) ParseError!void {
		try self.previous_token.expect_any(operator_tokens);

		while (operators.pop()) |lhs_operator| {
			if (rhs_operator.precedence() < lhs_operator.precedence()) break try operators.push(lhs_operator);

			try self.chunk.emit_opcode(lhs_operator.opcode());
		}

		try operators.push(rhs_operator);
		try self.step();
	}

	fn parse_statement(self: *Parser) ParseError!void {
		var local_depth = @as(usize, 0);

		while (true) {
			switch (self.current_token) {
				.newline => self.step() catch |step_error| switch (step_error) {
					error.UnexpectedEnd => return,
				},

				.keyword_return => {
					try self.previous_token.expect(.newline);

					self.step() catch |step_error| switch (step_error) {
						error.UnexpectedEnd => return,
					};

					try self.parse_expression();

					while (true) {
						self.step() catch |step_error| switch (step_error) {
							error.UnexpectedEnd => return,
						};

						try self.current_token.expect(.newline);
					}
				},

				.local_identifier => {
					try self.previous_token.expect_any(&.{.newline, .symbol_period});
					try self.step();
				},

				.global_identifier => {
					try self.previous_token.expect(.newline);
					try self.step();
				},

				.symbol_period => switch (self.previous_token) {
					.global_identifier => {
						// Create local copy of identifier because step() will overwrite captures.
						const identifier = self.previous_token.local_identifier;

						try self.step();
						try self.current_token.expect(.local_identifier);
						try self.chunk.emit_opcode(.get_global);
						try self.chunk.emit_operand(try self.chunk.intern_string(identifier));

						local_depth += 1;
					},

					.local_identifier => {
						// Create local copy of identifier because step() will overwrite captures.
						const identifier = self.previous_token.global_identifier;

						try self.step();
						try self.current_token.expect(.local_identifier);

						if (local_depth == 0) {
							try self.chunk.emit_byte(self.chunk.resolve_local(identifier) orelse {
								return error.UndefinedLocal;
							});
						} else {
							try self.chunk.emit_opcode(.get_index);
							try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
						}

						local_depth += 1;
					},

					else => return error.UnexpectedToken,
				},

				.symbol_assign => {
					try self.previous_token.expect(.local_identifier);

					const identifier = self.previous_token.local_identifier;

					if (local_depth == 0) {
						if (self.chunk.resolve_local(identifier)) |local_slot| {
							try self.chunk.emit_opcode(.set_local);
							try self.chunk.emit_byte(local_slot);
						} else {
							try self.chunk.declare_local(identifier);
						}
					} else {
						try self.chunk.emit_opcode(.set_index);
						try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
					}

					try self.step();
					try self.parse_expression();

					local_depth = 0;
				},

				.symbol_paren_left => {
					switch (self.previous_token) {
						.local_identifier => |identifier| {
							if (local_depth == 0) {
								try self.chunk.emit_byte(self.chunk.resolve_local(identifier) orelse {
									return error.UndefinedLocal;
								});
							} else {
								try self.chunk.emit_opcode(.get_index);
								try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
							}
						},

						.global_identifier => |identifier| {
							try self.chunk.emit_opcode(.get_global);
							try self.chunk.emit_operand(try self.chunk.intern_string(identifier));
						},

						else => return error.UnexpectedToken,
					}

					var argument_count = @as(Operand, 0);

					while (true) {
						try self.step();

						try switch (self.current_token) {
							.symbol_paren_right => break,
							else => self.parse_expression(),
						};

						argument_count += 1;

						switch (self.current_token) {
							.symbol_paren_right => break,
							.symbol_comma => {},
							else => return error.UnexpectedToken,
						}
					}

					try self.chunk.emit_opcode(.call);
					try self.chunk.emit_operand(argument_count);
					try self.chunk.emit_opcode(.pop);

					self.step() catch |step_error| switch (step_error) {
						error.UnexpectedEnd => return,
					};

					local_depth = 0;
				},

				else => return error.UnexpectedToken,
			}
		}
	}

	fn step(self: *Parser) StepError!void {
		self.previous_token = self.current_token;
		self.current_token = self.tokenizer.next() orelse return error.UnexpectedEnd;

		@import("std").debug.print("{s}\n", .{self.current_token.text()});
	}
};

fn SmallStack(comptime Element: type, comptime default: Element) type {
	const maximum = 255;

	return struct {
		buffer: [maximum]Element = [_]Element{default} ** maximum,
		count: u8 = 0,

		const Self = @This();

		fn peek(self: Self) ?Element {
			if (self.count == 0) return null;

			return self.buffer[self.count - 1];
		}

		fn pop(self: *Self) ?Element {
			if (self.count == 0) return null;

			self.count -= 1;

			return self.buffer[self.count];
		}

		fn push(self: *Self, element: Element) !void {
			if (self.count == maximum) return error.OutOfMemory;

			self.buffer[self.count] = element;
			self.count += 1;
		}
	};
}

const SymbolTable = coral.table.Hashed(coral.table.string_key, usize);