nprasad2077/oxc
1
0 Fork 0
mirror of https://github.com/oxc-project/oxc.git synced 2026-11-04 10:29:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

perf(napi/parser): deserializing tokens use only int32 array (#21138)

Browse Source 
Small perf optimization. When deserializing tokens in `oxc-parser`, use only the `Int32Array` view of the buffer, instead of both `Int32Array` and `Uint8Array`. Less values to keep in registers, and `int32` doesn't need to be obtained with a property lookup on each call.
This commit is contained in:
overlookmotel committed 2026-04-07 23:14:36 +00:00
1 parent f163d10859
commit 5b35a53a0b
1 file changed
+18 -13
Download Patch File Download Diff File Expand all files Collapse all files
napi/parser/src-js/raw-transfer/eager.js
+18 -13
View File
@@ -162,10 +162,11 @@ const TOKEN_TYPES = [
"JSXIdentifier",
];
// Mask for active bits in `ESTreeKind` discriminants
const TOKEN_KIND_MASK = 15;
// Details of Rust `Token` type
const TOKEN_SIZE = 16;
const KIND_FIELD_OFFSET = 8;
const IS_ESCAPED_FIELD_OFFSET = 10;
/**
* Deserialize tokens from buffer.
@@ -183,7 +184,7 @@ function deserializeTokens(buffer, sourceText, isJs) {
const tokens = [];
while (pos < endPos) {
tokens.push(deserializeToken(pos, buffer, sourceText, isJs));
tokens.push(deserializeToken(pos, int32, sourceText, isJs));
pos += TOKEN_SIZE;
}
return tokens;
@@ -192,21 +193,24 @@ function deserializeTokens(buffer, sourceText, isJs) {
/**
* Deserialize a token from buffer at position `pos`.
* @param {number} pos - Position in buffer containing Rust `Token` type
* @param {Uint8Array} buffer - Buffer containing AST in raw form
* @param {Int32Array} int32 - Buffer containing AST in raw form as an `Int32Array`
* @param {string} sourceText - Source for the file
* @param {boolean} isJs - `true` if parsing in JS mode
* @returns {Object} - Token object
*/
function deserializeToken(pos, buffer, sourceText, isJs) {
const { int32 } = buffer;
const pos32 = pos >> 2;
const start = int32[pos32],
end = int32[pos32 + 1];
function deserializeToken(pos, int32, sourceText, isJs) {
const pos32 = pos >> 2,
start = int32[pos32],
end = int32[pos32 + 1],
kindAndFlags = int32[pos32 + 2];
let value = sourceText.slice(start, end);
const kind = buffer[pos + KIND_FIELD_OFFSET];
// `Kind` is byte at index 8 in `Token`.
// `Kind` has 12 variants numbered from 0 to 11.
// We have to mask the bottom byte (`& 0xFF`), so may as well mask off bits which can't be set in `Kind` at same time.
// This may allow V8 to generate more efficient code for `TOKEN_TYPES[kind]`.
const kind = kindAndFlags & TOKEN_KIND_MASK;
if (kind === REGEXP_KIND) {
const patternEnd = value.lastIndexOf("/");
@@ -225,8 +229,9 @@ function deserializeToken(pos, buffer, sourceText, isJs) {
// Strip leading `#` from private identifiers
if (kind === PRIVATE_IDENTIFIER_KIND) value = value.slice(1);
// Unescape identifiers, keywords, and private identifiers in JS mode
if (isJs && kind <= PRIVATE_IDENTIFIER_KIND && buffer[pos + IS_ESCAPED_FIELD_OFFSET] === 1) {
// Unescape identifiers, keywords, and private identifiers in JS mode.
// `is_escaped` flag is in byte 10 of `Token`, and is a `bool`.
if (isJs && kind <= PRIVATE_IDENTIFIER_KIND && (kindAndFlags & 0x10000) !== 0) {
value = unescapeIdentifier(value);
}