Implement Control Flow Statements in Kym #37

Merged
kayomn merged 8 commits from kym-control-flow into main 2023-08-26 00:01:25 +02:00
5 changed files with 890 additions and 630 deletions

View File

@ -1,24 +1,23 @@
# Test comment.
i = 0
pos = @vec3(10, 20, 0.3)
coord = pos.xz
while i < 5:
@print("hello, world")
options = {
.title = "Afterglow",
i = i + 1
end
if i > 6:
@print("`i` greater than `6`")
elif i == 4:
@print("`i` is equal to `4`")
else:
@print("i'unno")
end
return {
.title = "Game",
.width = 1280,
.height = 800,
.tick_rate = 60,
["foo"] = "bar",
[42] = "42",
}
options["foo"] = "rab"
options[42] = "24"
@print(options.title)
@print(options["foo"])
@print(options[42])
return options

View File

@ -1,4 +1,4 @@
const Ast = @import("./kym/Ast.zig");
const ast = @import("./kym/ast.zig");
const coral = @import("coral");
@ -36,6 +36,7 @@ pub const RuntimeEnv = struct {
};
const Opcode = union (enum) {
pop,
push_nil,
push_true,
push_false,
@ -61,6 +62,9 @@ pub const RuntimeEnv = struct {
clt,
cge,
cle,
jt: u32,
jf: u32,
};
const OpcodeList = coral.list.Stack(Opcode);
@ -69,7 +73,7 @@ pub const RuntimeEnv = struct {
local_identifiers_buffer: [255][]const coral.io.Byte = [_][]const coral.io.Byte{""} ** 255,
local_identifiers_count: u8 = 0,
fn compile_expression(self: *CompilationUnit, chunk: *Chunk, expression: Ast.Expression) RuntimeError!void {
fn compile_expression(self: *CompilationUnit, chunk: *Chunk, expression: ast.Expression) RuntimeError!void {
const number_format = coral.utf8.DecimalFormat{
.delimiter = "_",
.positive_prefix = .none,
@ -132,9 +136,9 @@ pub const RuntimeEnv = struct {
.subtraction => .sub,
.multiplication => .mul,
.divsion => .div,
.greater_equals_comparison => .eql,
.greater_equals_comparison => .cge,
.greater_than_comparison => .cgt,
.equals_comparison => .cge,
.equals_comparison => .eql,
.less_than_comparison => .clt,
.less_equals_comparison => .cle,
});
@ -200,21 +204,23 @@ pub const RuntimeEnv = struct {
.local_get => |local_get| {
try chunk.opcodes.push_one(.{
.push_local = self.resolve_local(local_get) orelse {
.push_local = self.resolve_local(local_get.identifier) orelse {
return chunk.env.raise(error.OutOfMemory, "undefined local");
},
});
},
.local_set => |local_set| {
if (self.resolve_local(local_set)) |index| {
try self.compile_expression(chunk, local_set.value_expression.*);
if (self.resolve_local(local_set.identifier)) |index| {
try chunk.opcodes.push_one(.{.local_set = index});
} else {
if (self.local_identifiers_count == self.local_identifiers_buffer.len) {
return chunk.env.raise(error.BadSyntax, "chunks may have a maximum of 255 locals");
}
self.local_identifiers_buffer[self.local_identifiers_count] = local_set;
self.local_identifiers_buffer[self.local_identifiers_count] = local_set.identifier;
self.local_identifiers_count += 1;
}
},
@ -255,17 +261,62 @@ pub const RuntimeEnv = struct {
}
}
fn compile_statement(self: *CompilationUnit, chunk: *Chunk, statement: Ast.Statement) RuntimeError!void {
fn compile_statement(self: *CompilationUnit, chunk: *Chunk, statement: ast.Statement) RuntimeError!void {
switch (statement) {
.@"return" => |@"return"| {
if (@"return".expression) |expression| {
if (@"return") |expression| {
try self.compile_expression(chunk, expression);
} else {
try chunk.opcodes.push_one(.push_nil);
}
},
.expression => |expression| try self.compile_expression(chunk, expression),
.block => |block| {
for (block.values) |block_statement| {
try self.compile_statement(chunk, block_statement);
}
},
.@"while" => |@"while"| {
try self.compile_expression(chunk, @"while".condition_expression);
try chunk.opcodes.push_one(.{.jf = 0});
const origin_index = @as(u32, @intCast(chunk.opcodes.values.len - 1));
for (@"while".block_statements.values) |block_statement| {
try self.compile_statement(chunk, block_statement);
}
chunk.opcodes.values[origin_index].jf = @intCast(chunk.opcodes.values.len - 1);
try self.compile_expression(chunk, @"while".condition_expression);
try chunk.opcodes.push_one(.{.jt = origin_index});
},
.@"if" => |@"if"| {
try self.compile_expression(chunk, @"if".condition_expression);
try chunk.opcodes.push_one(.{.jf = 0});
const origin_index = @as(u32, @intCast(chunk.opcodes.values.len - 1));
for (@"if".block_statements.values) |block_statement| {
try self.compile_statement(chunk, block_statement);
}
chunk.opcodes.values[origin_index].jf = @intCast(chunk.opcodes.values.len - 1);
if (@"if".else_statement) |else_statement| {
try self.compile_statement(chunk, else_statement.*);
}
},
.expression => |expression| {
try self.compile_expression(chunk, expression);
if (expression == .invoke) {
try chunk.opcodes.push_one(.pop);
}
},
}
}
@ -288,10 +339,10 @@ pub const RuntimeEnv = struct {
}
};
fn compile(self: *Chunk, ast: Ast) RuntimeError!void {
fn compile(self: *Chunk, statements: []const ast.Statement) RuntimeError!void {
var unit = CompilationUnit{};
for (ast.list_statements()) |statement| {
for (statements) |statement| {
try unit.compile_statement(self, statement);
}
}
@ -322,8 +373,16 @@ pub const RuntimeEnv = struct {
defer coral.debug.assert(self.env.frames.pop() != null);
for (self.opcodes.values) |opcode| {
switch (opcode) {
var opcode_cursor = @as(u32, 0);
while (opcode_cursor < self.opcodes.values.len) : (opcode_cursor += 1) {
switch (self.opcodes.values[opcode_cursor]) {
.pop => {
if (try self.pop_local()) |ref| {
self.env.discard(ref);
}
},
.push_nil => try self.env.locals.push_one(null),
.push_true => try self.env.locals.push_one(try self.env.new_boolean(true)),
.push_false => try self.env.locals.push_one(try self.env.new_boolean(false)),
@ -337,6 +396,10 @@ pub const RuntimeEnv = struct {
},
.push_local => |push_local| {
if (push_local >= self.env.locals.values.len) {
return self.env.raise(error.IllegalState, "invalid local");
}
if (self.env.locals.values[push_local]) |local| {
try self.env.locals.push_one(try self.env.acquire(local));
} else {
@ -401,15 +464,11 @@ pub const RuntimeEnv = struct {
},
.object_get => {
const index = (try self.pop_local()) orelse {
return self.env.raise(error.TypeMismatch, "nil is not a valid index");
};
const index = try self.env.expect(try self.pop_local());
defer self.env.discard(index);
const indexable = (try self.pop_local()) orelse {
return self.env.raise(error.TypeMismatch, "nil is not a valid indexable");
};
const indexable = try self.env.expect(try self.pop_local());
defer self.env.discard(indexable);
@ -791,6 +850,29 @@ pub const RuntimeEnv = struct {
}));
},
.jf => |jf| {
const condition = (try self.pop_local()) orelse {
opcode_cursor = jf;
continue;
};
self.env.discard(condition);
if (!condition.is_truthy()) {
opcode_cursor = jf;
}
},
.jt => |jt| {
const condition = (try self.pop_local()) orelse continue;
self.env.discard(condition);
if (condition.is_truthy()) {
opcode_cursor = jt;
}
},
}
}
@ -1044,20 +1126,18 @@ pub const RuntimeEnv = struct {
defer self.allocator.deallocate(file_data);
const file_name = file_path.to_string() orelse "<script>";
var ast = Ast.make(self.allocator, file_name);
defer ast.free();
ast.parse(file_data) catch |parse_error| return switch (parse_error) {
error.BadSyntax => self.raise(error.BadSyntax, ast.error_message()),
error.OutOfMemory => error.OutOfMemory,
};
var chunk = try Chunk.make(self, file_name);
defer chunk.free();
try chunk.compile(ast);
var ast_tree = ast.Tree.make(self.allocator, file_name);
defer ast_tree.free();
try chunk.compile(ast_tree.parse(file_data) catch |parse_error| return switch (parse_error) {
error.BadSyntax => self.raise(error.BadSyntax, ast_tree.error_message()),
error.OutOfMemory => error.OutOfMemory,
});
return chunk.execute();
}
@ -1512,6 +1592,21 @@ pub const RuntimeRef = opaque {
.dynamic => true,
};
}
pub fn typename(self: *const RuntimeRef) []const coral.io.Byte {
return switch (self.object().payload) {
.false => "false",
.true => "true",
.float => "float",
.fixed => "fixed",
.symbol => "symbol",
.vector2 => "vector2",
.vector3 => "vector3",
.syscall => "syscall",
.string => "string",
.dynamic => "dynamic",
};
}
};
pub const Syscall = fn (env: *RuntimeEnv) RuntimeError!?*RuntimeRef;

View File

@ -1,567 +0,0 @@
const coral = @import("coral");
const tokens = @import("./tokens.zig");
name: []const coral.io.Byte,
allocator: coral.io.Allocator,
arena: coral.arena.Stacking,
statements: Statement.List,
error_buffer: coral.list.ByteStack,
tokenizer: tokens.Tokenizer,
pub const Expression = union (enum) {
nil_literal,
true_literal,
false_literal,
builtin: []const coral.io.Byte,
number_literal: []const coral.io.Byte,
string_literal: []const coral.io.Byte,
symbol_literal: []const coral.io.Byte,
table_literal: TableLiteral,
grouped_expression: *Expression,
local_get: []const coral.io.Byte,
local_set: []const coral.io.Byte,
field_get: struct {
object_expression: *Expression,
identifier: []const coral.io.Byte,
},
field_set: struct {
object_expression: *Expression,
identifier: []const coral.io.Byte,
value_expression: *Expression,
},
subscript_get: struct {
object_expression: *Expression,
subscript_expression: *Expression,
},
subscript_set: struct {
object_expression: *Expression,
subscript_expression: *Expression,
value_expression: *Expression,
},
binary_operation: struct {
operator: BinaryOperator,
lhs_expression: *Expression,
rhs_expression: *Expression,
},
unary_operation: struct {
operator: UnaryOperator,
expression: *Expression,
},
invoke: struct {
object_expression: *Expression,
argument_expressions: List,
},
pub const BinaryOperator = enum {
addition,
subtraction,
multiplication,
divsion,
equals_comparison,
greater_than_comparison,
greater_equals_comparison,
less_than_comparison,
less_equals_comparison,
fn token(self: BinaryOperator) tokens.Token {
return switch (self) {
.addition => .symbol_plus,
.subtraction => .symbol_minus,
.multiplication => .symbol_asterisk,
.divsion => .symbol_forward_slash,
.equals_comparison => .symbol_double_equals,
.greater_than_comparison => .symbol_greater_than,
.greater_equals_comparison => .symbol_greater_equals,
.less_than_comparison => .symbol_less_than,
.less_equals_comparison => .symbol_less_equals,
};
}
};
pub const TableLiteral = coral.list.Stack(struct {
key_expression: Expression,
value_expression: Expression,
});
pub const List = coral.list.Stack(Expression);
};
const ExpressionParser = fn (self: *Self) ParseError!Expression;
pub const ParseError = error {
OutOfMemory,
BadSyntax,
};
const Self = @This();
pub const Statement = union (enum) {
@"return": struct {
expression: ?Expression,
},
expression: Expression,
const List = coral.list.Stack(Statement);
};
const UnaryOperator = enum {
boolean_negation,
numeric_negation,
};
fn binary_operation_parser(
comptime parse_next: ExpressionParser,
comptime operators: []const Expression.BinaryOperator) ExpressionParser {
const BinaryOperationParser = struct {
fn parse(self: *Self) ParseError!Expression {
const allocator = self.arena.as_allocator();
var expression = try parse_next(self);
inline for (operators) |operator| {
const token = comptime operator.token();
if (self.tokenizer.token == coral.io.tag_of(token)) {
self.tokenizer.step();
if (self.tokenizer.token == .end) {
return self.report("expected other half of expression after `" ++ comptime token.text() ++ "`");
}
expression = .{
.binary_operation = .{
.operator = operator,
.lhs_expression = try coral.io.allocate_one(allocator, expression),
.rhs_expression = try coral.io.allocate_one(allocator, try parse_next(self)),
},
};
}
}
return expression;
}
};
return BinaryOperationParser.parse;
}
pub fn error_message(self: Self) []const coral.io.Byte {
return self.error_buffer.values;
}
pub fn free(self: *Self) void {
self.arena.free();
self.statements.free();
self.error_buffer.free();
}
pub fn make(allocator: coral.io.Allocator, ast_name: []const coral.io.Byte) Self {
return Self{
.arena = coral.arena.Stacking.make(allocator, 4096),
.error_buffer = coral.list.ByteStack.make(allocator),
.statements = Statement.List.make(allocator),
.tokenizer = .{.source = ""},
.allocator = allocator,
.name = ast_name,
};
}
fn report(self: *Self, message: []const coral.io.Byte) ParseError {
coral.utf8.print_formatted(coral.list.stack_as_writer(&self.error_buffer), "{name}@{line}: {message}", .{
.name = self.name,
.line = self.tokenizer.lines_stepped,
.message = message,
}) catch return error.OutOfMemory;
return error.BadSyntax;
}
pub fn list_statements(self: Self) []const Statement {
return self.statements.values;
}
pub fn parse(self: *Self, data: []const coral.io.Byte) ParseError!void {
self.tokenizer = .{.source = data};
var has_returned = false;
self.tokenizer.skip_newlines();
while (true) {
try self.statements.push_one(parse_statement: {
switch (self.tokenizer.token) {
.end => return,
.keyword_return => {
if (has_returned) {
return self.report("multiple returns in function scope but expected only one");
}
self.tokenizer.step();
if (self.tokenizer.token != .end and self.tokenizer.token != .newline) {
break: parse_statement .{
.@"return" = .{
.expression = try self.parse_expression(),
},
};
}
if (self.tokenizer.token != .end and self.tokenizer.token != .newline) {
return self.report("expected end or newline after return statement");
}
has_returned = true;
break: parse_statement .{
.@"return" = .{
.expression = null,
},
};
},
else => {
break: parse_statement .{
.expression = try self.parse_expression()
};
},
}
});
}
}
const parse_additive = binary_operation_parser(parse_equality, &.{
.addition,
.subtraction,
});
const parse_comparison = binary_operation_parser(parse_term, &.{
.greater_than_comparison,
.greater_equals_comparison,
.less_than_comparison,
.less_equals_comparison
});
const parse_equality = binary_operation_parser(parse_comparison, &.{
.equals_comparison,
});
pub fn parse_expression(self: *Self) ParseError!Expression {
const allocator = self.arena.as_allocator();
const expression = try parse_additive(self);
if (self.tokenizer.token == .symbol_equals) {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected assignment after `=`");
}
return switch (expression) {
.local_get => |local_get| .{.local_set = local_get},
.field_get => |field_get| .{
.field_set = .{
.object_expression = field_get.object_expression,
.identifier = field_get.identifier,
.value_expression = try coral.io.allocate_one(allocator, try self.parse_expression()),
},
},
.subscript_get => |subscript_get| .{
.subscript_set = .{
.object_expression = subscript_get.object_expression,
.subscript_expression = subscript_get.subscript_expression,
.value_expression = try coral.io.allocate_one(allocator, try self.parse_expression()),
},
},
else => self.report("expected local or field on left-hand side of expression"),
};
}
return expression;
}
fn parse_factor(self: *Self) ParseError!Expression {
const allocator = self.arena.as_allocator();
var expression = @as(Expression, parse: {
switch (self.tokenizer.token) {
.symbol_paren_left => {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected an expression after `(`");
}
const expression = try self.parse_expression();
if (self.tokenizer.token != .symbol_paren_right) {
return self.report("expected a closing `)` after expression");
}
self.tokenizer.skip_newlines();
break: parse .{.grouped_expression = try coral.io.allocate_one(allocator, expression)};
},
.keyword_nil => {
self.tokenizer.skip_newlines();
break: parse .nil_literal;
},
.keyword_true => {
self.tokenizer.skip_newlines();
break: parse .true_literal;
},
.keyword_false => {
self.tokenizer.skip_newlines();
break: parse .false_literal;
},
.number => |value| {
self.tokenizer.skip_newlines();
break: parse .{.number_literal = value};
},
.string => |value| {
self.tokenizer.skip_newlines();
break: parse .{.string_literal = value};
},
.identifier => |identifier| {
self.tokenizer.skip_newlines();
break: parse .{.local_get = identifier};
},
.builtin => |builtin| {
self.tokenizer.skip_newlines();
break: parse .{.builtin = builtin};
},
.symbol_brace_left => {
var table_literal = Expression.TableLiteral.make(allocator);
self.tokenizer.skip_newlines();
while (true) {
switch (self.tokenizer.token) {
.symbol_brace_right => {
self.tokenizer.skip_newlines();
break: parse .{.table_literal = table_literal};
},
.symbol_period => {
self.tokenizer.step();
const identifier = switch (self.tokenizer.token) {
.identifier => |identifier| identifier,
else => return self.report("expected identifier after `.`"),
};
self.tokenizer.skip_newlines();
if (self.tokenizer.token != .symbol_equals) {
return self.report("expected `=` after symbol");
}
self.tokenizer.skip_newlines();
try table_literal.push_one(.{
.value_expression = try self.parse_expression(),
.key_expression = .{.symbol_literal = identifier},
});
switch (self.tokenizer.token) {
.symbol_comma => self.tokenizer.skip_newlines(),
.symbol_brace_right => {
self.tokenizer.skip_newlines();
break: parse .{.table_literal = table_literal};
},
else => return self.report("expected `,` or `}` after expression"),
}
},
.symbol_bracket_left => {
self.tokenizer.step();
const subscript_expression = try self.parse_expression();
if (self.tokenizer.token != .symbol_bracket_right) {
return self.report("expected `]` after subscript expression");
}
self.tokenizer.skip_newlines();
if (self.tokenizer.token != .symbol_equals) {
return self.report("expected `=` after `]`");
}
self.tokenizer.skip_newlines();
try table_literal.push_one(.{
.value_expression = try self.parse_expression(),
.key_expression = subscript_expression,
});
switch (self.tokenizer.token) {
.symbol_comma => self.tokenizer.skip_newlines(),
.symbol_brace_right => {
self.tokenizer.skip_newlines();
break: parse .{.table_literal = table_literal};
},
else => return self.report("expected `,` or `}` after expression"),
}
},
else => return self.report("expected `}` or fields in table literal"),
}
}
},
.symbol_minus => {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected expression after numeric negation (`-`)");
}
break: parse .{
.unary_operation = .{
.expression = try coral.io.allocate_one(allocator, try self.parse_factor()),
.operator = .numeric_negation,
},
};
},
.symbol_bang => {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected expression after boolean negation (`!`)");
}
break: parse .{
.unary_operation = .{
.expression = try coral.io.allocate_one(allocator, try self.parse_factor()),
.operator = .boolean_negation,
},
};
},
else => return self.report("unexpected token in expression"),
}
});
while (true) {
switch (self.tokenizer.token) {
.symbol_period => {
self.tokenizer.skip_newlines();
// TODO: Remove when Zig fixes miscompilation with in-place struct re-assignment.
const unnecessary_temp = try coral.io.allocate_one(allocator, expression);
expression = .{
.field_get = .{
.identifier = switch (self.tokenizer.token) {
.identifier => |field_identifier| field_identifier,
else => return self.report("expected identifier after `.`"),
},
.object_expression = unnecessary_temp,
},
};
self.tokenizer.skip_newlines();
},
.symbol_bracket_left => {
self.tokenizer.skip_newlines();
// TODO: Remove when Zig fixes miscompilation with in-place struct re-assignment.
const unnecessary_temp = try coral.io.allocate_one(allocator, expression);
expression = .{
.subscript_get = .{
.subscript_expression = try coral.io.allocate_one(allocator, try self.parse_expression()),
.object_expression = unnecessary_temp,
},
};
if (self.tokenizer.token != .symbol_bracket_right) {
return self.report("expected `]` subscript expression");
}
self.tokenizer.skip_newlines();
},
.symbol_paren_left => {
var argument_expressions = Expression.List.make(allocator);
while (true) {
self.tokenizer.skip_newlines();
switch (self.tokenizer.token) {
.symbol_paren_right => break,
else => {
try argument_expressions.push_one(try self.parse_expression());
switch (self.tokenizer.token) {
.symbol_comma => continue,
.symbol_paren_right => break,
else => return self.report("expected `,` or `)` after function argument expression"),
}
},
}
}
self.tokenizer.skip_newlines();
// TODO: Remove when Zig fixes miscompilation with in-place struct re-assignment.
const unnecessary_temp = try coral.io.allocate_one(allocator, expression);
expression = .{
.invoke = .{
.argument_expressions = argument_expressions,
.object_expression = unnecessary_temp,
},
};
},
else => break,
}
}
return expression;
}
const parse_term = binary_operation_parser(parse_factor, &.{
.multiplication,
.divsion,
});

677
source/ona/kym/ast.zig Normal file
View File

@ -0,0 +1,677 @@
const coral = @import("coral");
const tokens = @import("./tokens.zig");
pub const BinaryOperator = enum {
addition,
subtraction,
multiplication,
divsion,
equals_comparison,
greater_than_comparison,
greater_equals_comparison,
less_than_comparison,
less_equals_comparison,
fn builder(comptime build_next: ExpressionBuilder, comptime operators: []const BinaryOperator) ExpressionBuilder {
const Builder = struct {
fn build(self: *Tree) ParseError!Expression {
const allocator = self.arena.as_allocator();
var expression = try build_next(self);
inline for (operators) |operator| {
const token = @as(tokens.Token, switch (operator) {
.addition => .symbol_plus,
.subtraction => .symbol_minus,
.multiplication => .symbol_asterisk,
.divsion => .symbol_forward_slash,
.equals_comparison => .symbol_double_equals,
.greater_than_comparison => .symbol_greater_than,
.greater_equals_comparison => .symbol_greater_equals,
.less_than_comparison => .symbol_less_than,
.less_equals_comparison => .symbol_less_equals,
});
if (self.tokenizer.token == coral.io.tag_of(token)) {
self.tokenizer.step();
if (self.tokenizer.token == .end) {
return self.report(
"expected other half of expression after `" ++
comptime token.text() ++
"`");
}
// TODO: Remove once Zig has fixed struct self-reassignment.
const unnecessary_temp = try coral.io.allocate_one(allocator, expression);
expression = .{
.binary_operation = .{
.operator = operator,
.lhs_expression = unnecessary_temp,
.rhs_expression = try coral.io.allocate_one(allocator, try build_next(self)),
},
};
}
}
return expression;
}
};
return Builder.build;
}
};
pub const Expression = union (enum) {
nil_literal,
true_literal,
false_literal,
builtin: []const coral.io.Byte,
number_literal: []const coral.io.Byte,
string_literal: []const coral.io.Byte,
symbol_literal: []const coral.io.Byte,
table_literal: TableLiteral,
grouped_expression: *Expression,
local_get: struct {
identifier: []const coral.io.Byte,
},
local_set: struct {
identifier: []const coral.io.Byte,
value_expression: *Expression,
},
field_get: struct {
object_expression: *Expression,
identifier: []const coral.io.Byte,
},
field_set: struct {
object_expression: *Expression,
identifier: []const coral.io.Byte,
value_expression: *Expression,
},
subscript_get: struct {
object_expression: *Expression,
subscript_expression: *Expression,
},
subscript_set: struct {
object_expression: *Expression,
subscript_expression: *Expression,
value_expression: *Expression,
},
binary_operation: struct {
operator: BinaryOperator,
lhs_expression: *Expression,
rhs_expression: *Expression,
},
unary_operation: struct {
operator: UnaryOperator,
expression: *Expression,
},
invoke: struct {
object_expression: *Expression,
argument_expressions: List,
},
const List = coral.list.Stack(Expression);
const TableLiteral = coral.list.Stack(struct {
key_expression: Expression,
value_expression: Expression,
});
};
const ExpressionBuilder = fn (self: *Tree) ParseError!Expression;
pub const ParseError = error {
OutOfMemory,
BadSyntax,
};
pub const Statement = union (enum) {
@"return": ?Expression,
@"if": struct {
condition_expression: Expression,
block_statements: List,
else_statement: ?*Statement,
},
@"while": struct {
condition_expression: Expression,
block_statements: List,
},
block: List,
expression: Expression,
const List = coral.list.Stack(Statement);
};
pub const Tree = struct {
name: []const coral.io.Byte,
allocator: coral.io.Allocator,
arena: coral.arena.Stacking,
error_buffer: coral.list.ByteStack,
tokenizer: tokens.Tokenizer,
parsed_statements: Statement.List,
has_returned: bool,
pub fn error_message(self: Tree) []const coral.io.Byte {
return self.error_buffer.values;
}
pub fn free(self: *Tree) void {
self.parsed_statements.free();
self.error_buffer.free();
self.arena.free();
}
pub fn make(allocator: coral.io.Allocator, ast_name: []const coral.io.Byte) Tree {
return .{
.arena = coral.arena.Stacking.make(allocator, 4096),
.error_buffer = coral.list.ByteStack.make(allocator),
.parsed_statements = Statement.List.make(allocator),
.tokenizer = .{.source = ""},
.allocator = allocator,
.name = ast_name,
.has_returned = false,
};
}
pub fn parse(self: *Tree, data: []const coral.io.Byte) ParseError![]const Statement {
self.free();
self.tokenizer = .{.source = data};
self.has_returned = false;
self.tokenizer.skip_newlines();
while (self.tokenizer.token != .end) {
try self.parsed_statements.push_one(try self.parse_statement());
}
return self.parsed_statements.values;
}
const parse_additive = BinaryOperator.builder(parse_equality, &.{
.addition,
.subtraction,
});
fn parse_branch(self: *Tree) ParseError!Statement {
const allocator = self.arena.as_allocator();
defer self.tokenizer.skip_newlines();
self.tokenizer.step();
const condition_expression = try self.parse_expression();
if (self.tokenizer.token != .symbol_colon) {
return self.report("expected `:` after `if` statement condition");
}
var statements = Statement.List.make(allocator);
self.tokenizer.skip_newlines();
while (true) {
switch (self.tokenizer.token) {
.keyword_end => {
return .{
.@"if" = .{
.condition_expression = condition_expression,
.block_statements = statements,
.else_statement = null,
},
};
},
.keyword_else => {
self.tokenizer.step();
if (self.tokenizer.token != .symbol_colon) {
return self.report("expected `:` after `if` statement condition");
}
var else_statements = Statement.List.make(allocator);
self.tokenizer.skip_newlines();
while (self.tokenizer.token != .keyword_end) {
try else_statements.push_one(try self.parse_statement());
}
return .{
.@"if" = .{
.else_statement = try coral.io.allocate_one(allocator, Statement{.block = else_statements}),
.condition_expression = condition_expression,
.block_statements = statements,
},
};
},
.keyword_elif => {
return .{
.@"if" = .{
.else_statement = try coral.io.allocate_one(allocator, try self.parse_branch()),
.condition_expression = condition_expression,
.block_statements = statements,
},
};
},
else => try statements.push_one(try self.parse_statement()),
}
}
}
const parse_comparison = BinaryOperator.builder(parse_term, &.{
.greater_than_comparison,
.greater_equals_comparison,
.less_than_comparison,
.less_equals_comparison
});
const parse_equality = BinaryOperator.builder(parse_comparison, &.{
.equals_comparison,
});
pub fn parse_expression(self: *Tree) ParseError!Expression {
const allocator = self.arena.as_allocator();
const expression = try self.parse_additive();
if (self.tokenizer.token == .symbol_equals) {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected assignment after `=`");
}
return switch (expression) {
.local_get => |local_get| .{
.local_set = .{
.identifier = local_get.identifier,
.value_expression = try coral.io.allocate_one(allocator, try self.parse_expression()),
},
},
.field_get => |field_get| .{
.field_set = .{
.object_expression = field_get.object_expression,
.identifier = field_get.identifier,
.value_expression = try coral.io.allocate_one(allocator, try self.parse_expression()),
},
},
.subscript_get => |subscript_get| .{
.subscript_set = .{
.object_expression = subscript_get.object_expression,
.subscript_expression = subscript_get.subscript_expression,
.value_expression = try coral.io.allocate_one(allocator, try self.parse_expression()),
},
},
else => self.report("expected local or field on left-hand side of expression"),
};
}
return expression;
}
fn parse_factor(self: *Tree) ParseError!Expression {
const allocator = self.arena.as_allocator();
var expression = @as(Expression, parse: {
switch (self.tokenizer.token) {
.symbol_paren_left => {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected an expression after `(`");
}
const expression = try self.parse_expression();
if (self.tokenizer.token != .symbol_paren_right) {
return self.report("expected a closing `)` after expression");
}
self.tokenizer.skip_newlines();
break: parse .{.grouped_expression = try coral.io.allocate_one(allocator, expression)};
},
.keyword_nil => {
self.tokenizer.skip_newlines();
break: parse .nil_literal;
},
.keyword_true => {
self.tokenizer.skip_newlines();
break: parse .true_literal;
},
.keyword_false => {
self.tokenizer.skip_newlines();
break: parse .false_literal;
},
.number => |value| {
self.tokenizer.skip_newlines();
break: parse .{.number_literal = value};
},
.string => |value| {
self.tokenizer.skip_newlines();
break: parse .{.string_literal = value};
},
.identifier => |identifier| {
self.tokenizer.skip_newlines();
break: parse .{.local_get = .{.identifier = identifier}};
},
.builtin => |builtin| {
self.tokenizer.skip_newlines();
break: parse .{.builtin = builtin};
},
.symbol_brace_left => {
var table_literal = Expression.TableLiteral.make(allocator);
self.tokenizer.skip_newlines();
while (true) {
switch (self.tokenizer.token) {
.symbol_brace_right => {
self.tokenizer.skip_newlines();
break: parse .{.table_literal = table_literal};
},
.symbol_period => {
self.tokenizer.step();
const identifier = switch (self.tokenizer.token) {
.identifier => |identifier| identifier,
else => return self.report("expected identifier after `.`"),
};
self.tokenizer.skip_newlines();
if (self.tokenizer.token != .symbol_equals) {
return self.report("expected `=` after symbol");
}
self.tokenizer.skip_newlines();
try table_literal.push_one(.{
.value_expression = try self.parse_expression(),
.key_expression = .{.symbol_literal = identifier},
});
switch (self.tokenizer.token) {
.symbol_comma => self.tokenizer.skip_newlines(),
.symbol_brace_right => {
self.tokenizer.skip_newlines();
break: parse .{.table_literal = table_literal};
},
else => return self.report("expected `,` or `}` after expression"),
}
},
.symbol_bracket_left => {
self.tokenizer.step();
const subscript_expression = try self.parse_expression();
if (self.tokenizer.token != .symbol_bracket_right) {
return self.report("expected `]` after subscript expression");
}
self.tokenizer.skip_newlines();
if (self.tokenizer.token != .symbol_equals) {
return self.report("expected `=` after `]`");
}
self.tokenizer.skip_newlines();
try table_literal.push_one(.{
.value_expression = try self.parse_expression(),
.key_expression = subscript_expression,
});
switch (self.tokenizer.token) {
.symbol_comma => self.tokenizer.skip_newlines(),
.symbol_brace_right => {
self.tokenizer.skip_newlines();
break: parse .{.table_literal = table_literal};
},
else => return self.report("expected `,` or `}` after expression"),
}
},
else => return self.report("expected `}` or fields in table literal"),
}
}
},
.symbol_minus => {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected expression after numeric negation (`-`)");
}
break: parse .{
.unary_operation = .{
.expression = try coral.io.allocate_one(allocator, try self.parse_factor()),
.operator = .numeric_negation,
},
};
},
.symbol_bang => {
self.tokenizer.skip_newlines();
if (self.tokenizer.token == .end) {
return self.report("expected expression after boolean negation (`!`)");
}
break: parse .{
.unary_operation = .{
.expression = try coral.io.allocate_one(allocator, try self.parse_factor()),
.operator = .boolean_negation,
},
};
},
else => return self.report("unexpected token in expression"),
}
});
while (true) {
switch (self.tokenizer.token) {
.symbol_period => {
self.tokenizer.skip_newlines();
// TODO: Remove when Zig fixes miscompilation with in-place struct re-assignment.
const unnecessary_temp = try coral.io.allocate_one(allocator, expression);
expression = .{
.field_get = .{
.identifier = switch (self.tokenizer.token) {
.identifier => |field_identifier| field_identifier,
else => return self.report("expected identifier after `.`"),
},
.object_expression = unnecessary_temp,
},
};
self.tokenizer.skip_newlines();
},
.symbol_bracket_left => {
self.tokenizer.skip_newlines();
// TODO: Remove when Zig fixes miscompilation with in-place struct re-assignment.
const unnecessary_temp = try coral.io.allocate_one(allocator, expression);
expression = .{
.subscript_get = .{
.subscript_expression = try coral.io.allocate_one(allocator, try self.parse_expression()),
.object_expression = unnecessary_temp,
},
};
if (self.tokenizer.token != .symbol_bracket_right) {
return self.report("expected `]` subscript expression");
}
self.tokenizer.skip_newlines();
},
.symbol_paren_left => {
var argument_expressions = Expression.List.make(allocator);
while (true) {
self.tokenizer.skip_newlines();
switch (self.tokenizer.token) {
.symbol_paren_right => break,
else => {
try argument_expressions.push_one(try self.parse_expression());
switch (self.tokenizer.token) {
.symbol_comma => continue,
.symbol_paren_right => break,
else => return self.report("expected `,` or `)` after function argument expression"),
}
},
}
}
self.tokenizer.skip_newlines();
// TODO: Remove when Zig fixes miscompilation with in-place struct re-assignment.
const unnecessary_temp = try coral.io.allocate_one(allocator, expression);
expression = .{
.invoke = .{
.argument_expressions = argument_expressions,
.object_expression = unnecessary_temp,
},
};
},
else => break,
}
}
return expression;
}
fn parse_statement(self: *Tree) ParseError!Statement {
const allocator = self.arena.as_allocator();
switch (self.tokenizer.token) {
.keyword_return => {
defer self.tokenizer.skip_newlines();
if (self.has_returned) {
return self.report("multiple returns in lambda scope but expected only one");
}
self.tokenizer.step();
if (self.tokenizer.token != .end and self.tokenizer.token != .newline) {
return .{.@"return" = try self.parse_expression()};
}
if (self.tokenizer.token != .end and self.tokenizer.token != .newline) {
return self.report("expected end or newline after return statement");
}
self.has_returned = true;
return .{.@"return" = null};
},
.keyword_while => {
defer self.tokenizer.skip_newlines();
self.tokenizer.step();
const condition_expression = try self.parse_expression();
if (self.tokenizer.token != .symbol_colon) {
return self.report("expected `:` after `while` statement");
}
var statements = Statement.List.make(allocator);
self.tokenizer.skip_newlines();
while (self.tokenizer.token != .keyword_end) {
try statements.push_one(try self.parse_statement());
}
return .{
.@"while" = .{
.block_statements = statements,
.condition_expression = condition_expression,
},
};
},
.keyword_if => return self.parse_branch(),
else => return .{.expression = try self.parse_expression()},
}
}
const parse_term = BinaryOperator.builder(parse_factor, &.{
.multiplication,
.divsion,
});
fn report(self: *Tree, message: []const coral.io.Byte) ParseError {
coral.utf8.print_formatted(coral.list.stack_as_writer(&self.error_buffer), "{name}@{line}: {message}", .{
.name = self.name,
.line = self.tokenizer.lines_stepped,
.message = message,
}) catch return error.OutOfMemory;
return error.BadSyntax;
}
};
pub const UnaryOperator = enum {
boolean_negation,
numeric_negation,
};

View File

@ -21,7 +21,7 @@ pub const Token = union(enum) {
symbol_bracket_left,
symbol_bracket_right,
symbol_period,
symbol_lambda,
symbol_colon,
symbol_less_than,
symbol_less_equals,
symbol_greater_than,
@ -38,6 +38,12 @@ pub const Token = union(enum) {
keyword_return,
keyword_self,
keyword_const,
keyword_if,
keyword_do,
keyword_end,
keyword_while,
keyword_else,
keyword_elif,
pub fn text(self: Token) []const coral.io.Byte {
return switch (self) {
@ -62,7 +68,7 @@ pub const Token = union(enum) {
.symbol_bracket_left => "[",
.symbol_bracket_right => "]",
.symbol_period => ".",
.symbol_lambda => "=>",
.symbol_colon => ":",
.symbol_less_than => "<",
.symbol_less_equals => "<=",
.symbol_greater_than => ">",
@ -79,6 +85,12 @@ pub const Token = union(enum) {
.keyword_true => "true",
.keyword_return => "return",
.keyword_self => "self",
.keyword_if => "if",
.keyword_do => "do",
.keyword_end => "end",
.keyword_while => "while",
.keyword_elif => "elif",
.keyword_else => "else",
};
}
};
@ -86,7 +98,7 @@ pub const Token = union(enum) {
pub const Tokenizer = struct {
source: []const coral.io.Byte,
lines_stepped: usize = 1,
token: Token = .end,
token: Token = .newline,
pub fn skip_newlines(self: *Tokenizer) void {
self.step();
@ -173,9 +185,29 @@ pub const Tokenizer = struct {
}
},
'n' => {
if (coral.io.ends_with(identifier, "il")) {
self.token = .keyword_nil;
'd' => {
if (coral.io.ends_with(identifier, "o")) {
self.token = .keyword_do;
return;
}
},
'e' => {
if (coral.io.ends_with(identifier, "lse")) {
self.token = .keyword_else;
return;
}
if (coral.io.ends_with(identifier, "lif")) {
self.token = .keyword_elif;
return;
}
if (coral.io.ends_with(identifier, "nd")) {
self.token = .keyword_end;
return;
}
@ -189,9 +221,17 @@ pub const Tokenizer = struct {
}
},
't' => {
if (coral.io.ends_with(identifier, "rue")) {
self.token = .keyword_true;
'i' => {
if (coral.io.ends_with(identifier, "f")) {
self.token = .keyword_if;
return;
}
},
'n' => {
if (coral.io.ends_with(identifier, "il")) {
self.token = .keyword_nil;
return;
}
@ -213,6 +253,22 @@ pub const Tokenizer = struct {
}
},
't' => {
if (coral.io.ends_with(identifier, "rue")) {
self.token = .keyword_true;
return;
}
},
'w' => {
if (coral.io.ends_with(identifier, "hile")) {
self.token = .keyword_while;
return;
}
},
else => {},
}
@ -338,6 +394,13 @@ pub const Tokenizer = struct {
return;
},
':' => {
self.token = .symbol_colon;
cursor += 1;
return;
},
'=' => {
cursor += 1;
@ -350,13 +413,6 @@ pub const Tokenizer = struct {
return;
},
'>' => {
cursor += 1;
self.token = .symbol_lambda;
return;
},
else => {},
}
}