use std::{borrow::Cow, cmp, slice};
use cow_utils::CowUtils;
use oxc_ast::ast::*;
use oxc_data_structures::{code_buffer::CodeBuffer, stack::Stack};
use oxc_index::IndexVec;
use oxc_semantic::Scoping;
use oxc_span::{GetSpan, Span};
use oxc_str::CompactStr;
use oxc_syntax::{
class::ClassId,
identifier::{is_identifier_part, is_identifier_part_ascii},
operator::{BinaryOperator, UnaryOperator, UpdateOperator},
precedence::Precedence,
};
use rustc_hash::FxHashMap;
mod binary_expr_visitor;
mod comment;
mod context;
mod r#gen;
mod operator;
mod options;
#[cfg(feature = "sourcemap")]
mod sourcemap_builder;
mod str;
use binary_expr_visitor::BinaryExpressionVisitor;
use comment::CommentsMap;
use operator::Operator;
#[cfg(feature = "sourcemap")]
use sourcemap_builder::SourcemapBuilder;
use str::{Quote, cold_branch, is_script_close_tag};
pub use context::Context;
pub use r#gen::{Gen, GenExpr};
pub use options::{CodegenOptions, CommentOptions, LegalComment};
pub use oxc_data_structures::code_buffer::IndentChar;
#[non_exhaustive]
pub struct CodegenReturn {
pub code: String,
#[cfg(feature = "sourcemap")]
pub map: Option<oxc_sourcemap::SourceMap>,
pub legal_comments: Vec<Comment>,
}
pub struct Codegen<'a> {
pub(crate) options: CodegenOptions,
source_text: Option<&'a str>,
scoping: Option<Scoping>,
private_member_mappings: Option<IndexVec<ClassId, FxHashMap<String, CompactStr>>>,
code: CodeBuffer,
prev_op_end: usize,
prev_reg_exp_end: usize,
need_space_before_dot: usize,
print_next_indent_as_space: bool,
binary_expr_stack: Stack<BinaryExpressionVisitor<'a>>,
class_stack: Stack<ClassId>,
next_class_id: ClassId,
is_jsx: bool,
needs_semicolon: bool,
prev_op: Option<Operator>,
start_of_stmt: usize,
start_of_arrow_expr: usize,
start_of_default_export: usize,
indent: u32,
quote: Quote,
comments: CommentsMap,
#[cfg(feature = "sourcemap")]
sourcemap_builder: Option<SourcemapBuilder<'a>>,
}
impl Default for Codegen<'_> {
fn default() -> Self {
Self::new()
}
}
impl<'a> From<Codegen<'a>> for String {
fn from(val: Codegen<'a>) -> Self {
val.into_source_text()
}
}
impl<'a> From<Codegen<'a>> for Cow<'a, str> {
fn from(val: Codegen<'a>) -> Self {
Cow::Owned(val.into_source_text())
}
}
impl<'a> Codegen<'a> {
#[must_use]
pub fn new() -> Self {
let options = CodegenOptions::default();
Self {
options,
source_text: None,
scoping: None,
private_member_mappings: None,
code: CodeBuffer::default(),
needs_semicolon: false,
need_space_before_dot: 0,
print_next_indent_as_space: false,
binary_expr_stack: Stack::with_capacity(12),
class_stack: Stack::with_capacity(4),
next_class_id: ClassId::from_usize(0),
prev_op_end: 0,
prev_reg_exp_end: 0,
prev_op: None,
start_of_stmt: 0,
start_of_arrow_expr: 0,
start_of_default_export: 0,
is_jsx: false,
indent: 0,
quote: Quote::Double,
comments: CommentsMap::default(),
#[cfg(feature = "sourcemap")]
sourcemap_builder: None,
}
}
#[must_use]
pub fn with_options(mut self, options: CodegenOptions) -> Self {
self.quote = if options.single_quote { Quote::Single } else { Quote::Double };
self.code = CodeBuffer::with_indent(options.indent_char, options.indent_width);
self.options = options;
self
}
#[must_use]
pub fn with_source_text(mut self, source_text: &'a str) -> Self {
self.source_text = Some(source_text);
self
}
#[must_use]
pub fn with_scoping(mut self, scoping: Option<Scoping>) -> Self {
self.scoping = scoping;
self
}
#[must_use]
pub fn with_private_member_mappings(
mut self,
mappings: Option<IndexVec<ClassId, FxHashMap<String, CompactStr>>>,
) -> Self {
self.private_member_mappings = mappings;
self
}
#[must_use]
pub fn build(mut self, program: &Program<'a>) -> CodegenReturn {
self.quote = if self.options.single_quote { Quote::Single } else { Quote::Double };
self.source_text = Some(program.source_text);
self.indent = self.options.initial_indent;
self.code.reserve(program.source_text.len());
self.build_comments(&program.comments);
#[cfg(feature = "sourcemap")]
if let Some(path) = &self.options.source_map_path {
self.sourcemap_builder = Some(SourcemapBuilder::new(path, program.source_text));
}
program.print(&mut self, Context::default());
let legal_comments = self.handle_eof_linked_or_external_comments(program);
let code = self.code.into_string();
#[cfg(feature = "sourcemap")]
let map = self.sourcemap_builder.map(SourcemapBuilder::into_sourcemap);
CodegenReturn {
code,
#[cfg(feature = "sourcemap")]
map,
legal_comments,
}
}
#[must_use]
pub fn into_source_text(self) -> String {
self.code.into_string()
}
#[inline]
pub fn print_ascii_byte(&mut self, byte: u8) {
self.code.print_ascii_byte(byte);
}
#[inline]
pub fn print_str(&mut self, s: &str) {
self.code.print_str(s);
}
#[inline]
pub fn print_str_escaping_script_close_tag(&mut self, s: &str) {
let bytes = s.as_bytes();
let mut consumed = 0;
let mut search_bytes = |mut ptr: *const u8, last_ptr| {
loop {
let byte = unsafe { *ptr.as_ref().unwrap_unchecked() };
if byte == b'<' {
let slice = unsafe { slice::from_raw_parts(ptr, 8) };
if is_script_close_tag(slice) {
unsafe {
let index = ptr.offset_from_unsigned(bytes.as_ptr());
let before = bytes.get_unchecked(consumed..=index);
self.code.print_bytes_unchecked(before);
consumed = index + 2;
}
self.print_str("\\/");
}
}
if ptr == last_ptr {
break;
}
ptr = unsafe { ptr.add(1) };
}
};
let mut chunks = bytes.chunks_exact(16);
for (chunk_index, chunk) in chunks.by_ref().enumerate() {
#[expect(clippy::missing_panics_doc, reason = "infallible")]
let chunk: &[u8; 16] = chunk.try_into().unwrap();
let mut contains_lt = false;
for &byte in chunk {
if byte == b'<' {
contains_lt = true;
}
}
if contains_lt {
cold_branch(|| unsafe {
let index = chunk_index * 16;
let remaining_bytes = bytes.len() - index;
let last_offset = cmp::min(remaining_bytes - 8, 15);
let ptr = bytes.as_ptr().add(index);
let last_ptr = ptr.add(last_offset);
search_bytes(ptr, last_ptr);
});
}
}
let last_chunk = chunks.remainder();
if last_chunk.len() >= 8 {
let ptr = last_chunk.as_ptr();
let last_ptr = unsafe { ptr.add(last_chunk.len() - 8) };
search_bytes(ptr, last_ptr);
}
unsafe {
let remaining = bytes.get_unchecked(consumed..);
self.code.print_bytes_unchecked(remaining);
}
}
#[inline]
pub fn print_expression(&mut self, expr: &Expression<'_>) {
expr.print_expr(self, Precedence::Lowest, Context::empty());
}
}
impl<'a> Codegen<'a> {
fn code(&self) -> &CodeBuffer {
&self.code
}
fn code_len(&self) -> usize {
self.code().len()
}
#[inline]
fn print_soft_space(&mut self) {
if !self.options.minify {
self.print_ascii_byte(b' ');
}
}
#[inline]
fn print_hard_space(&mut self) {
self.print_ascii_byte(b' ');
}
#[inline]
fn print_soft_newline(&mut self) {
if !self.options.minify {
self.print_ascii_byte(b'\n');
}
}
#[inline]
fn print_hard_newline(&mut self) {
self.print_ascii_byte(b'\n');
}
#[inline]
fn print_semicolon(&mut self) {
self.print_ascii_byte(b';');
}
#[inline]
fn print_comma(&mut self) {
self.print_ascii_byte(b',');
}
#[inline]
fn print_space_before_identifier(&mut self) {
let Some(byte) = self.last_byte() else { return };
if self.prev_reg_exp_end != self.code.len() {
let is_identifier = if byte.is_ascii() {
is_identifier_part_ascii(byte as char)
} else {
is_identifier_part(self.last_char().unwrap())
};
if !is_identifier {
return;
}
}
self.print_hard_space();
}
#[inline]
fn last_byte(&self) -> Option<u8> {
self.code.last_byte()
}
#[inline]
fn last_char(&self) -> Option<char> {
self.code.last_char()
}
#[inline]
fn indent(&mut self) {
if !self.options.minify {
self.indent += 1;
}
}
#[inline]
fn dedent(&mut self) {
if !self.options.minify {
self.indent -= 1;
}
}
#[inline]
fn enter_class(&mut self) {
let class_id = self.next_class_id;
self.next_class_id = ClassId::from_usize(self.next_class_id.index() + 1);
self.class_stack.push(class_id);
}
#[inline]
fn exit_class(&mut self) {
self.class_stack.pop();
}
#[inline]
fn current_class_ids(&self) -> impl Iterator<Item = ClassId> {
self.class_stack.iter().rev().copied()
}
#[inline]
fn wrap<F: FnMut(&mut Self)>(&mut self, wrap: bool, mut f: F) {
if wrap {
self.print_ascii_byte(b'(');
}
f(self);
if wrap {
self.print_ascii_byte(b')');
}
}
#[inline]
fn print_indent(&mut self) {
if self.options.minify {
return;
}
if self.print_next_indent_as_space {
self.print_hard_space();
self.print_next_indent_as_space = false;
return;
}
self.code.print_indent(self.indent as usize);
}
#[inline]
fn print_semicolon_after_statement(&mut self) {
if self.options.minify {
self.needs_semicolon = true;
} else {
self.print_str(";\n");
}
}
#[inline]
fn print_semicolon_if_needed(&mut self) {
if self.needs_semicolon {
self.print_semicolon();
self.needs_semicolon = false;
}
}
#[inline]
fn print_ellipsis(&mut self) {
self.print_str("...");
}
#[inline]
fn print_colon(&mut self) {
self.print_ascii_byte(b':');
}
#[inline]
fn print_equal(&mut self) {
self.print_ascii_byte(b'=');
}
fn print_curly_braces<F: FnOnce(&mut Self)>(&mut self, span: Span, single_line: bool, op: F) {
self.add_source_mapping(span);
self.print_ascii_byte(b'{');
if !single_line {
self.print_soft_newline();
self.indent();
}
op(self);
if !single_line {
self.dedent();
self.print_indent();
}
self.print_ascii_byte(b'}');
}
fn print_block_start(&mut self, span: Span) {
self.add_source_mapping(span);
self.print_ascii_byte(b'{');
self.print_soft_newline();
self.indent();
}
fn print_block_end(&mut self, _span: Span) {
self.dedent();
self.print_indent();
self.print_ascii_byte(b'}');
}
fn print_body(&mut self, stmt: &Statement<'_>, need_space: bool, ctx: Context) {
match stmt {
Statement::BlockStatement(stmt) => {
self.print_soft_space();
self.print_block_statement(stmt, ctx);
self.print_soft_newline();
}
Statement::EmptyStatement(_) => {
self.print_semicolon();
self.print_soft_newline();
}
stmt => {
if need_space && self.options.minify {
self.print_hard_space();
}
self.print_next_indent_as_space = true;
stmt.print(self, ctx);
}
}
}
fn print_block_statement(&mut self, stmt: &BlockStatement<'_>, ctx: Context) {
self.print_curly_braces(stmt.span, stmt.body.is_empty(), |p| {
for stmt in &stmt.body {
p.print_semicolon_if_needed();
stmt.print(p, ctx);
}
});
self.needs_semicolon = false;
}
fn print_directives_and_statements(
&mut self,
directives: &[Directive<'_>],
stmts: &[Statement<'_>],
ctx: Context,
) {
for directive in directives {
directive.print(self, ctx);
}
let Some((first, rest)) = stmts.split_first() else {
return;
};
let mut first_needs_parens = false;
if directives.is_empty()
&& !self.options.minify
&& let Statement::ExpressionStatement(s) = first
{
let s = s.expression.without_parentheses();
if matches!(s, Expression::StringLiteral(_)) {
first_needs_parens = true;
self.print_ascii_byte(b'(');
s.print_expr(self, Precedence::Lowest, ctx);
self.print_ascii_byte(b')');
self.print_semicolon_after_statement();
}
}
if !first_needs_parens {
first.print(self, ctx);
}
for stmt in rest {
self.print_semicolon_if_needed();
stmt.print(self, ctx);
}
}
#[inline]
fn print_list<T: Gen>(&mut self, items: &[T], ctx: Context) {
let Some((first, rest)) = items.split_first() else {
return;
};
first.print(self, ctx);
for item in rest {
self.print_comma();
self.print_soft_space();
item.print(self, ctx);
}
}
#[inline]
fn print_expressions<T: GenExpr>(&mut self, items: &[T], precedence: Precedence, ctx: Context) {
let Some((first, rest)) = items.split_first() else {
return;
};
first.print_expr(self, precedence, ctx);
for item in rest {
self.print_comma();
self.print_soft_space();
item.print_expr(self, precedence, ctx);
}
}
fn print_arguments(&mut self, span: Span, arguments: &[Argument<'_>], ctx: Context) {
self.print_ascii_byte(b'(');
let has_comment_before_right_paren = span.end > 0 && self.has_comment(span.end - 1);
let has_comment = has_comment_before_right_paren
|| arguments.iter().any(|item| self.has_comment(item.span().start));
if has_comment {
self.indent();
self.print_list_with_comments(arguments, ctx);
if !has_comment_before_right_paren
|| (span.end > 0 && !self.print_expr_comments(span.end - 1))
{
self.print_soft_newline();
}
self.dedent();
self.print_indent();
} else {
self.print_list(arguments, ctx);
}
self.print_ascii_byte(b')');
self.add_source_mapping_end(span);
}
fn print_list_with_comments(&mut self, items: &[Argument<'_>], ctx: Context) {
let Some((first, rest)) = items.split_first() else {
return;
};
if self.print_expr_comments(first.span().start) {
self.print_indent();
} else {
self.print_soft_newline();
self.print_indent();
}
first.print(self, ctx);
for item in rest {
self.print_comma();
if self.print_expr_comments(item.span().start) {
self.print_indent();
} else {
self.print_soft_newline();
self.print_indent();
}
item.print(self, ctx);
}
}
fn get_identifier_reference_name(&self, reference: &IdentifierReference<'a>) -> &'a str {
if let Some(scoping) = &self.scoping
&& let Some(reference_id) = reference.reference_id.get()
&& let Some(name) = scoping.get_reference_name(reference_id)
{
return unsafe { std::mem::transmute_copy(&name) };
}
reference.name.as_str()
}
fn get_binding_identifier_name(&self, ident: &BindingIdentifier<'a>) -> &'a str {
if let Some(scoping) = &self.scoping
&& let Some(symbol_id) = ident.symbol_id.get()
{
let name = scoping.symbol_name(symbol_id);
return unsafe { std::mem::transmute_copy(&name) };
}
ident.name.as_str()
}
fn print_space_before_operator(&mut self, next: Operator) {
if self.prev_op_end != self.code.len() {
return;
}
let Some(prev) = self.prev_op else { return };
let bin_op_add = Operator::Binary(BinaryOperator::Addition);
let bin_op_sub = Operator::Binary(BinaryOperator::Subtraction);
let un_op_pos = Operator::Unary(UnaryOperator::UnaryPlus);
let un_op_pre_inc = Operator::Update(UpdateOperator::Increment);
let un_op_neg = Operator::Unary(UnaryOperator::UnaryNegation);
let un_op_pre_dec = Operator::Update(UpdateOperator::Decrement);
let un_op_post_dec = Operator::Update(UpdateOperator::Decrement);
let bin_op_gt = Operator::Binary(BinaryOperator::GreaterThan);
let un_op_not = Operator::Unary(UnaryOperator::LogicalNot);
if ((prev == bin_op_add || prev == un_op_pos)
&& (next == bin_op_add || next == un_op_pos || next == un_op_pre_inc))
|| ((prev == bin_op_sub || prev == un_op_neg)
&& (next == bin_op_sub || next == un_op_neg || next == un_op_pre_dec))
|| (prev == un_op_post_dec && next == bin_op_gt)
|| (prev == un_op_not
&& next == un_op_pre_dec
&& self.code.peek_nth_byte_back(1) == Some(b'<'))
{
self.print_hard_space();
}
}
fn print_non_negative_float(&mut self, num: f64) {
let mut buffer = dragonbox_ecma::Buffer::new();
if num < 1000.0 && num.fract() == 0.0 {
self.print_str(buffer.format(num));
self.need_space_before_dot = self.code_len();
} else {
self.print_minified_number(num, &mut buffer);
}
}
fn print_decorators(&mut self, decorators: &[Decorator<'_>], ctx: Context) {
for decorator in decorators {
decorator.print(self, ctx);
self.print_hard_space();
}
}
#[expect(clippy::cast_possible_truncation, clippy::cast_sign_loss, clippy::cast_possible_wrap)]
fn print_minified_number(&mut self, num: f64, buffer: &mut dragonbox_ecma::Buffer) {
if num < 1000.0 && num.fract() == 0.0 {
self.print_str(buffer.format(num));
self.need_space_before_dot = self.code_len();
return;
}
let mut s = buffer.format(num);
if s.starts_with("0.") {
s = &s[1..];
}
let mut best_candidate = s.cow_replacen("e+", "e", 1);
let mut is_hex = false;
if num.fract() == 0.0 {
let hex_candidate = format!("0x{:x}", num as u128);
if hex_candidate.len() < best_candidate.len() {
is_hex = true;
best_candidate = hex_candidate.into();
}
}
else if best_candidate.starts_with(".0") {
if let Some(i) = best_candidate.bytes().skip(2).position(|c| c != b'0') {
let len = i + 2; let digits = &best_candidate[len..];
let exp = digits.len() + len - 1;
let exp_str_len = itoa::Buffer::new().format(exp).len();
let expected_len = digits.len() + 2 + exp_str_len;
if expected_len < best_candidate.len() {
best_candidate = format!("{digits}e-{exp}").into();
debug_assert_eq!(best_candidate.len(), expected_len);
}
}
}
if !is_hex
&& best_candidate.ends_with('0')
&& let Some(len) = best_candidate.bytes().rev().position(|c| c != b'0')
{
let base = &best_candidate[0..best_candidate.len() - len];
let exp_str_len = itoa::Buffer::new().format(len).len();
let expected_len = base.len() + 1 + exp_str_len;
if expected_len < best_candidate.len() {
best_candidate = format!("{base}e{len}").into();
debug_assert_eq!(best_candidate.len(), expected_len);
}
}
if let Some((integer, point, exponent)) = best_candidate
.split_once('.')
.and_then(|(a, b)| b.split_once('e').map(|e| (a, e.0, e.1)))
{
let new_expr = exponent.parse::<isize>().unwrap() - point.len() as isize;
let new_exp_str_len = itoa::Buffer::new().format(new_expr).len();
let expected_len = integer.len() + point.len() + 1 + new_exp_str_len;
if expected_len < best_candidate.len() {
best_candidate = format!("{integer}{point}e{new_expr}").into();
debug_assert_eq!(best_candidate.len(), expected_len);
}
}
self.print_str(&best_candidate);
if !best_candidate.bytes().any(|b| matches!(b, b'.' | b'e' | b'x')) {
self.need_space_before_dot = self.code_len();
}
}
#[cfg(feature = "sourcemap")]
fn add_source_mapping(&mut self, span: Span) {
if let Some(sourcemap_builder) = self.sourcemap_builder.as_mut()
&& !span.is_empty()
{
sourcemap_builder.add_source_mapping(self.code.as_bytes(), span.start, None);
}
}
#[cfg(not(feature = "sourcemap"))]
#[inline]
#[expect(clippy::needless_pass_by_ref_mut, clippy::unused_self)]
fn add_source_mapping(&mut self, _span: Span) {}
#[cfg(feature = "sourcemap")]
fn add_source_mapping_end(&mut self, span: Span) {
if let Some(sourcemap_builder) = self.sourcemap_builder.as_mut()
&& !span.is_empty()
{
if let Some(source_text) = self.source_text {
#[expect(clippy::cast_possible_truncation)]
if span.end >= source_text.len() as u32 {
return;
}
}
sourcemap_builder.add_source_mapping(self.code.as_bytes(), span.end, None);
}
}
#[cfg(not(feature = "sourcemap"))]
#[inline]
#[expect(clippy::needless_pass_by_ref_mut, clippy::unused_self)]
fn add_source_mapping_end(&mut self, _span: Span) {}
#[cfg(feature = "sourcemap")]
fn add_source_mapping_for_name(&mut self, span: Span, name: &str) {
if let Some(sourcemap_builder) = self.sourcemap_builder.as_mut()
&& !span.is_empty()
{
sourcemap_builder.add_source_mapping_for_name(self.code.as_bytes(), span, name);
}
}
#[cfg(not(feature = "sourcemap"))]
#[inline]
#[expect(clippy::needless_pass_by_ref_mut, clippy::unused_self)]
fn add_source_mapping_for_name(&mut self, _span: Span, _name: &str) {}
}