rollup-plugin-glsl-optimize
Version:
Import GLSL source files as strings. Pre-processed, validated and optimized with Khronos Group SPIRV-Tools. Supports glslify.
320 lines (304 loc) • 9.41 kB
JavaScript
/**
* @internal
* Very basic GLSL parsing handling #version and #extension preprocessor directives
* @param {string} input
*/
export function* simpleParse(input) {
yield* parser(lexer(input));
}
/**
* @typedef {1|2|3|4|5|6|7|8} TokenID Token ID
* @return {TokenID}
* @param {Number} n
*/
const _T = (n) => /** @type {TokenID} */(n);
/** @internal */
export const TOK = Object.freeze({EOL: _T(1), EOF: _T(2), Line: _T(3), Comment: _T(4),
Version: _T(5), Extension: _T(6), LineNo: _T(7), Directive: _T(8)});
/** @internal */
export const TOKENNAMES = Object.freeze({1: 'EOL', 2: 'EOF', 3: 'Line', 4: 'Comment',
5: 'Version', 6: 'Extension', 7: 'LineNo', 8: 'Directive'});
/**
* @typedef {Object} LexerToken
* @property {TokenID} type
* @property {string} text
* @property {string} value
* @property {number} line
* @property {number} col
*/
/**
*
* @param {string} input
* @return {Generator<LexerToken>}
*/
function* lexer(input) {
let skipOne = false;
let line = 1, col = 0;
let afterLineContinuation = false, inCommentSingleLine = false, inCommentMultiLine = false;
/** @type {LexerToken} */
let curToken = undefined;
/** @type {string} */
let curText = undefined;
/** @param {TokenID} type */
const setTokenIf = (type) => {
if (!curToken) {
setToken(type);
}
};
/** @param {TokenID} type */
const setToken = (type) => {
curToken = {type, col, line, value: '', text: ''};
};
const emitToken = function* () {
if (curToken.type === TOK.Line) {
yield curToken;
} else {
yield curToken;
}
curToken = undefined;
};
const emitTokenIf = function* () {
if (curToken) {
yield* emitToken();
}
};
const appendTokenValue = () => {
curToken.text += curText;
curToken.value += curText;
};
const appendToken = () => {
curToken.text += curText;
};
const handleEOL = function* () {
if (afterLineContinuation) {
appendToken();
afterLineContinuation = false;
} else if (inCommentMultiLine) {
appendToken();
curToken.value += '\n';
} else {
yield* emitTokenIf();
inCommentSingleLine = false;
yield {type: TOK.EOL, text: curText, col, line, value: '\n'};
}
line++; col = 0;
};
if (input.length > 0) {
let next = input[0];
let cur;
/* Even though GLSL uses UTF-8 encoding, the actual syntactic charset is ASCII
So we can use JS indexed iteration (UTF-16) for a speedup compared to full Unicode iteration (for of) */
for (let i = 1; i <= input.length; i++) {
cur = next;
next = i < input.length ? input[i] : undefined;
col++;
if (skipOne) {
skipOne = false;
continue;
}
/** Current char */
curText = cur;
switch (cur) {
case '\\':
switch (next) {
case '\r': case '\n':
setTokenIf(TOK.Line);
appendToken();
afterLineContinuation = true;
break;
default:
setTokenIf(TOK.Line);
appendTokenValue();
}
break;
case '\r':
if (next === '\n') {
curText += next; skipOne = true;
}
yield* handleEOL();
break;
case '\n':
if (next === '\r') {
curText += next; skipOne = true;
}
yield* handleEOL();
break;
default:
if (inCommentSingleLine) {
appendTokenValue();
} else if (inCommentMultiLine) {
if (cur === '*' && next === '/') {
curText += next; skipOne = true;
appendToken();
yield* emitToken();
inCommentMultiLine = false;
} else {
appendTokenValue();
}
} else {
switch (cur) {
case '/':
switch (next) {
case '/':
curText += next; skipOne = true;
yield* emitTokenIf();
setToken(TOK.Comment);
appendToken();
inCommentSingleLine = true;
break;
case '*':
curText += next; skipOne = true;
yield* emitTokenIf();
setToken(TOK.Comment);
appendToken();
inCommentMultiLine = true;
break;
default:
setTokenIf(TOK.Line);
appendTokenValue();
}
break;
default:
setTokenIf(TOK.Line);
appendTokenValue();
}
}
} // End main switch
} // End for
yield* emitTokenIf();
} // End if input
yield {type: TOK.EOF, text: '', col, line, value: ''};
}
/**
* @typedef {Object} ExtensionToken
* @property {string} ExtensionName
* @property {string} ExtensionBehavior
* @typedef {Object} VersionToken
* @property {string} Version
* @typedef {LexerToken & Partial<ExtensionToken> & Partial<VersionToken>} ParserToken
*/
/**
*
* @param {Generator<LexerToken>} input
* @return {Generator<ParserToken>}
*/
function* parser(input) {
/* Coalesce comment tokens, since we could have
#directive <COMMENT> value1 value2 */
/** @type {ParserToken[]} */
let LineTokens = [];
for (const token of input) {
switch (token.type) {
case TOK.Line:
LineTokens.push(token);
break;
case TOK.Comment:
if (LineTokens.length > 0) {
LineTokens.push(token);
} else {
yield token;
}
break;
case TOK.EOL: case TOK.EOF:
if (LineTokens.length > 0) {
/** @type {ParserToken} */
const combinedToken = {...LineTokens[0], type: TOK.Line,
// Comments are treated syntactically as a single space:
value: LineTokens.map((token) => token.type === TOK.Comment ? ' ' : token.value).join(''),
text: LineTokens.map((token) => token.text).join(''),
};
const matchPreprocessor = /^[ \t]*#[ \t]*([^ \t].*)?$/u.exec(combinedToken.value);
if (matchPreprocessor && matchPreprocessor.length === 2) {
// Preprocessor directive
const directiveLine = matchPreprocessor[1];
if (directiveLine !== undefined) {
const directiveParts = directiveLine.split(/[ \t]+/u);
if (directiveParts.length > 0) {
let [directive, ...body] = directiveParts;
body = body.filter(Boolean); // Filter whitespace
switch (directive.toLowerCase()) {
case 'version':
combinedToken.type = TOK.Version;
combinedToken.Version = body.join(' ');
break;
case 'line':
combinedToken.type = TOK.LineNo;
break;
case 'extension': {
combinedToken.type = TOK.Extension;
if (body.length === 3 && body[1] === ':') {
combinedToken.ExtensionName = body[0];
const extensionBehavior = body[2].toLowerCase();
switch (extensionBehavior) {
case 'require': case 'enable': case 'warn': case 'disable':
combinedToken.ExtensionBehavior = extensionBehavior;
break;
default:
combinedToken.ExtensionBehavior = body[2];
warnParse(`#extension directive: unknown behavior '${body[2]}'`, combinedToken);
}
} else {
warnParse('#extension directive: parse error', combinedToken);
}
}
break;
default:
combinedToken.type = TOK.Directive;
break;
}
}
}
}
yield combinedToken;
LineTokens = [];
}
yield token;
break;
default:
yield token;
}
}
}
const warnParse = (message, token) => console.error(`Warning: ${formatParseError(message, token)}`);
/**
* @internal
* @param {string} message
* @param {LexerToken} token
*/
export const formatParseError = (message, token) => `${message}\nLine ${
token.line} col ${token.col}:\n${formatLine(token.text)}`;
const formatLine = (line) => {
let lineF = '';
for (let i = 0; i < line.length; i++) {
switch (line[i]) {
case '\r':
lineF += '<CR>';
break;
case '\n':
lineF += '<EOL>';
break;
case '\t':
lineF += '<TAB>';
break;
default:
lineF += line[i];
}
}
return lineF;
};
/**
* @return {string}
* @param {ParserToken} t
*/
const printToken = (t) => {
const {type, line, col, value, text, ...rest} = t;
const restKV = Object.entries(rest);
return `<${TOKENNAMES[type]} L${line}:${col}> v:'${
formatLine(value)}' t:'${formatLine(text)}'${
restKV.length ? ` | ${restKV.map(([k, v]) => `${k} : '${v}'`).join(' | ')}`: ''}`;
};
export const test = {
lexer,
parser,
printToken,
};