three.js/examples/jsm/loaders/usd/USDAParser.js

// Pre-compiled regex patterns for performance
const DEF_MATCH_REGEX = /^def\s+(?:(\w+)\s+)?"?([^"]+)"?$/;
const VARIANT_STRING_REGEX = /^string\s+(\w+)$/;
const ATTR_MATCH_REGEX = /^(?:uniform\s+)?(\w+(?:\[\])?)\s+(.+)$/;

// Spec types (must match USDCParser/USDComposer)
const SpecType = {
	Attribute: 1,
	Prim: 6,
	Relationship: 8
};

class USDAParser {

	parseText( text ) {

		// Preprocess: strip comments and normalize multiline values
		text = this._preprocess( text );

		const root = {};

		const lines = text.split( '\n' );

		let string = null;
		let target = root;

		const stack = [ root ];

		for ( const line of lines ) {

			if ( line.includes( '=' ) ) {

				// Find the first '=' that's not inside quotes
				const eqIdx = this._findAssignmentOperator( line );

				if ( eqIdx === - 1 ) {

					string = line.trim();
					continue;

				}

				const lhs = line.slice( 0, eqIdx ).trim();
				const rhs = line.slice( eqIdx + 1 ).trim();

				if ( rhs.endsWith( '{' ) ) {

					const group = {};
					stack.push( group );

					target[ lhs ] = group;
					target = group;

				} else if ( rhs.endsWith( '(' ) ) {

					// see #28631

					const values = rhs.slice( 0, - 1 );
					target[ lhs ] = values;

					const meta = {};
					stack.push( meta );

					target = meta;

				} else {

					target[ lhs ] = rhs;

				}

			} else if ( line.includes( ':' ) && ! line.includes( '=' ) ) {

				// Handle dictionary entries like "0: [(...)...]" for timeSamples
				const colonIdx = line.indexOf( ':' );
				const key = line.slice( 0, colonIdx ).trim();
				const value = line.slice( colonIdx + 1 ).trim();

				// Only process if key looks like a number (timeSamples frame)
				if ( /^[\d.]+$/.test( key ) ) {

					target[ key ] = value;

				}

			} else if ( line.endsWith( '{' ) ) {

				const group = target[ string ] || {};
				stack.push( group );

				target[ string ] = group;
				target = group;

			} else if ( line.endsWith( '}' ) ) {

				stack.pop();

				if ( stack.length === 0 ) continue;

				target = stack[ stack.length - 1 ];

			} else if ( line.endsWith( '(' ) ) {

				const meta = {};
				stack.push( meta );

				string = line.split( '(' )[ 0 ].trim() || string;

				target[ string ] = meta;
				target = meta;

			} else if ( line.endsWith( ')' ) ) {

				stack.pop();

				target = stack[ stack.length - 1 ];

			} else if ( line.trim() ) {

				string = line.trim();

			}

		}

		return root;

	}

	_preprocess( text ) {

		// Remove block comments /* ... */
		text = this._stripBlockComments( text );

		// Collapse triple-quoted strings into single lines
		text = this._collapseTripleQuotedStrings( text );

		// Remove line comments # ... (but preserve #usda header)
		// Only remove # comments that aren't at the start of a line or after whitespace
		const lines = text.split( '\n' );
		const processed = [];

		let inMultilineValue = false;
		let bracketDepth = 0;
		let parenDepth = 0;
		let accumulated = '';

		for ( let i = 0; i < lines.length; i ++ ) {

			let line = lines[ i ];

			// Strip inline comments (but not inside strings)
			line = this._stripInlineComment( line );

			// Track bracket/paren depth for multiline values
			const trimmed = line.trim();

			if ( inMultilineValue ) {

				// Continue accumulating multiline value
				accumulated += ' ' + trimmed;

				// Update depths
				for ( const ch of trimmed ) {

					if ( ch === '[' ) bracketDepth ++;
					else if ( ch === ']' ) bracketDepth --;
					else if ( ch === '(' && bracketDepth > 0 ) parenDepth ++;
					else if ( ch === ')' && bracketDepth > 0 ) parenDepth --;

				}

				// Check if multiline value is complete
				if ( bracketDepth === 0 && parenDepth === 0 ) {

					processed.push( accumulated );
					accumulated = '';
					inMultilineValue = false;

				}

			} else {

				// Check if this line starts a multiline array value
				// Look for patterns like "attr = [" or "attr = @path@[" without closing ]
				if ( trimmed.includes( '=' ) ) {

					const eqIdx = this._findAssignmentOperator( trimmed );

					if ( eqIdx !== - 1 ) {

						const rhs = trimmed.slice( eqIdx + 1 ).trim();

						// Count brackets in the value part
						let openBrackets = 0;
						let closeBrackets = 0;

						for ( const ch of rhs ) {

							if ( ch === '[' ) openBrackets ++;
							else if ( ch === ']' ) closeBrackets ++;

						}

						if ( openBrackets > closeBrackets ) {

							// Multiline array detected
							inMultilineValue = true;
							bracketDepth = openBrackets - closeBrackets;
							parenDepth = 0;
							accumulated = trimmed;
							continue;

						}

					}

				}

				processed.push( trimmed );

			}

		}

		return processed.join( '\n' );

	}

	_stripBlockComments( text ) {

		// Iteratively remove /* ... */ comments without regex backtracking
		let result = '';
		let i = 0;

		while ( i < text.length ) {

			// Check for block comment start
			if ( text[ i ] === '/' && i + 1 < text.length && text[ i + 1 ] === '*' ) {

				// Find the closing */
				let j = i + 2;

				while ( j < text.length ) {

					if ( text[ j ] === '*' && j + 1 < text.length && text[ j + 1 ] === '/' ) {

						// Found closing, skip past it
						j += 2;
						break;

					}

					j ++;

				}

				// Move past the comment (or to end if unclosed)
				i = j;

			} else {

				result += text[ i ];
				i ++;

			}

		}

		return result;

	}

	_collapseTripleQuotedStrings( text ) {

		let result = '';
		let i = 0;

		while ( i < text.length ) {

			if ( i + 2 < text.length ) {

				const triple = text.slice( i, i + 3 );

				if ( triple === '\'\'\'' || triple === '"""' ) {

					const quoteChar = triple;
					result += quoteChar;
					i += 3;

					while ( i < text.length ) {

						if ( i + 2 < text.length && text.slice( i, i + 3 ) === quoteChar ) {

							result += quoteChar;
							i += 3;
							break;

						} else {

							if ( text[ i ] === '\n' ) {

								result += '\\n';

							} else if ( text[ i ] !== '\r' ) {

								result += text[ i ];

							}

							i ++;

						}

					}

					continue;

				}

			}

			result += text[ i ];
			i ++;

		}

		return result;

	}

	_stripInlineComment( line ) {

		// Don't strip if line starts with #usda
		if ( line.trim().startsWith( '#usda' ) ) return line;

		// Find # that's not inside a string
		let inString = false;
		let stringChar = null;
		let escaped = false;

		for ( let i = 0; i < line.length; i ++ ) {

			const ch = line[ i ];

			if ( escaped ) {

				escaped = false;
				continue;

			}

			if ( ch === '\\' ) {

				escaped = true;
				continue;

			}

			if ( ! inString && ( ch === '"' || ch === '\'' ) ) {

				inString = true;
				stringChar = ch;

			} else if ( inString && ch === stringChar ) {

				inString = false;
				stringChar = null;

			} else if ( ! inString && ch === '#' ) {

				// Found comment start outside of string
				return line.slice( 0, i ).trimEnd();

			}

		}

		return line;

	}

	_findAssignmentOperator( line ) {

		// Find the first '=' that's not inside quotes
		let inString = false;
		let stringChar = null;
		let escaped = false;

		for ( let i = 0; i < line.length; i ++ ) {

			const ch = line[ i ];

			if ( escaped ) {

				escaped = false;
				continue;

			}

			if ( ch === '\\' ) {

				escaped = true;
				continue;

			}

			if ( ! inString && ( ch === '"' || ch === '\'' ) ) {

				inString = true;
				stringChar = ch;

			} else if ( inString && ch === stringChar ) {

				inString = false;
				stringChar = null;

			} else if ( ! inString && ch === '=' ) {

				return i;

			}

		}

		return - 1;

	}

	/**
	 * Parse USDA text and return raw spec data in specsByPath format.
	 * Used by USDComposer for unified scene composition.
	 */
	parseData( text ) {

		const root = this.parseText( text );
		const specsByPath = {};

		// Parse root metadata
		const rootFields = {};
		if ( '#usda 1.0' in root ) {

			const header = root[ '#usda 1.0' ];

			if ( header.upAxis ) {

				rootFields.upAxis = header.upAxis.replace( /"/g, '' );

			}

			if ( header.defaultPrim ) {

				rootFields.defaultPrim = header.defaultPrim.replace( /"/g, '' );

			}

			if ( header.metersPerUnit !== undefined ) {

				rootFields.metersPerUnit = parseFloat( header.metersPerUnit );

			}

			if ( header.framesPerSecond !== undefined ) {

				rootFields.framesPerSecond = parseFloat( header.framesPerSecond );

			}

			if ( header.timeCodesPerSecond !== undefined ) {

				rootFields.timeCodesPerSecond = parseFloat( header.timeCodesPerSecond );

			}

		}

		specsByPath[ '/' ] = { specType: SpecType.Prim, fields: rootFields };

		// Walk the tree and build specsByPath
		const walkTree = ( data, parentPath ) => {

			const primChildren = [];

			for ( const key in data ) {

				// Skip metadata
				if ( key === '#usda 1.0' ) continue;
				if ( key === 'variants' ) continue;

				// Check for primitive definitions
				// Matches both 'def TypeName "name"' and 'def "name"' (no type)
				const defMatch = key.match( DEF_MATCH_REGEX );
				if ( defMatch ) {

					const typeName = defMatch[ 1 ] || '';
					const name = defMatch[ 2 ];
					const path = parentPath === '/' ? '/' + name : parentPath + '/' + name;

					primChildren.push( name );

					const primFields = { typeName };
					const primData = data[ key ];

					// Extract attributes and relationships from this prim
					this._extractPrimData( primData, path, primFields, specsByPath, SpecType );

					specsByPath[ path ] = { specType: SpecType.Prim, fields: primFields };

					// Recurse into children
					walkTree( primData, path );

				}

			}

			// Add primChildren to parent spec
			if ( primChildren.length > 0 && specsByPath[ parentPath ] ) {

				specsByPath[ parentPath ].fields.primChildren = primChildren;

			}

		};

		walkTree( root, '/' );

		// Fallback: infer elementSize for primvars:skel:jointIndices/jointWeights
		// when not explicitly declared in the USDA text
		this._inferSkelElementSize( specsByPath );

		return { specsByPath };

	}

	_inferSkelElementSize( specsByPath ) {

		// For each mesh prim with primvars:skel:jointIndices/jointWeights but no
		// elementSize, infer it from the data: elementSize = array.length / numVertices.
		for ( const path in specsByPath ) {

			const spec = specsByPath[ path ];
			if ( spec.specType !== SpecType.Prim || spec.fields.typeName !== 'Mesh' ) continue;

			const pointsSpec = specsByPath[ path + '.points' ];
			if ( ! pointsSpec || ! pointsSpec.fields.default ) continue;

			const numVertices = pointsSpec.fields.default.length / 3;
			if ( numVertices === 0 ) continue;

			this._inferElementSize( specsByPath[ path + '.primvars:skel:jointIndices' ], numVertices );
			this._inferElementSize( specsByPath[ path + '.primvars:skel:jointWeights' ], numVertices );

		}

	}

	_inferElementSize( attrSpec, numVertices ) {

		if ( ! attrSpec || attrSpec.fields.elementSize !== undefined || ! attrSpec.fields.default ) return;

		const len = attrSpec.fields.default.length;
		if ( len > 0 && len % numVertices === 0 ) attrSpec.fields.elementSize = len / numVertices;

	}

	_extractPrimData( data, path, primFields, specsByPath, SpecType ) {

		if ( ! data || typeof data !== 'object' ) return;

		for ( const key in data ) {

			// Skip nested defs (handled by walkTree)
			if ( key.startsWith( 'def ' ) ) continue;

			if ( key === 'prepend references' ) {

				primFields.references = [ data[ key ] ];
				continue;

			}

			if ( key === 'payload' ) {

				primFields.payload = data[ key ];
				continue;

			}

			if ( key === 'variants' ) {

				const variantSelection = {};
				const variants = data[ key ];

				for ( const vKey in variants ) {

					const match = vKey.match( VARIANT_STRING_REGEX );
					if ( match ) {

						const variantSetName = match[ 1 ];
						const variantValue = variants[ vKey ].replace( /"/g, '' );
						variantSelection[ variantSetName ] = variantValue;

					}

				}

				if ( Object.keys( variantSelection ).length > 0 ) {

					primFields.variantSelection = variantSelection;

				}

				continue;

			}

			if ( key.startsWith( 'rel ' ) ) {

				const relName = key.slice( 4 );
				const relPath = path + '.' + relName;
				const target = data[ key ].replace( /[<>]/g, '' );
				specsByPath[ relPath ] = {
					specType: SpecType.Relationship,
					fields: { targetPaths: [ target ] }
				};
				continue;

			}

			// Handle xformOpOrder
			if ( key.includes( 'xformOpOrder' ) ) {

				const ops = data[ key ]
					.replace( /[\[\]]/g, '' )
					.split( ',' )
					.map( s => s.trim().replace( /"/g, '' ) );
				primFields.xformOpOrder = ops;
				continue;

			}

			// Handle typed attributes
			// Format: [qualifier] type attrName (e.g., "uniform token[] joints", "float3 position")
			const attrMatch = key.match( ATTR_MATCH_REGEX );
			if ( attrMatch ) {

				const valueType = attrMatch[ 1 ];
				const attrName = attrMatch[ 2 ];
				const rawValue = data[ key ];

				// Handle connection attributes (e.g., "inputs:normal.connect = </path>")
				if ( attrName.endsWith( '.connect' ) ) {

					const baseAttrName = attrName.slice( 0, - 8 ); // Remove '.connect'
					const attrPath = path + '.' + baseAttrName;

					// Parse connection path - extract from <path> format
					let connPath = String( rawValue ).trim();
					if ( connPath.startsWith( '<' ) ) connPath = connPath.slice( 1 );
					if ( connPath.endsWith( '>' ) ) connPath = connPath.slice( 0, - 1 );

					// Get or create the attribute spec
					if ( ! specsByPath[ attrPath ] ) {

						specsByPath[ attrPath ] = {
							specType: SpecType.Attribute,
							fields: { typeName: valueType }
						};

					}

					specsByPath[ attrPath ].fields.connectionPaths = [ connPath ];
					continue;

				}

				// Handle timeSamples attributes specially
				if ( attrName.endsWith( '.timeSamples' ) && typeof rawValue === 'object' ) {

					const baseAttrName = attrName.slice( 0, - 12 ); // Remove '.timeSamples'
					const attrPath = path + '.' + baseAttrName;

					// Parse timeSamples dictionary into times and values arrays
					const times = [];
					const values = [];

					for ( const frameKey in rawValue ) {

						const frame = parseFloat( frameKey );
						if ( isNaN( frame ) ) continue;

						times.push( frame );
						values.push( this._parseAttributeValue( valueType, rawValue[ frameKey ] ) );

					}

					// Sort by time
					const sorted = times.map( ( t, i ) => ( { t, v: values[ i ] } ) ).sort( ( a, b ) => a.t - b.t );

					specsByPath[ attrPath ] = {
						specType: SpecType.Attribute,
						fields: {
							timeSamples: { times: sorted.map( s => s.t ), values: sorted.map( s => s.v ) },
							typeName: valueType
						}
					};

				} else {

					// Parse value based on type
					const parsedValue = this._parseAttributeValue( valueType, rawValue );

					// Store as attribute spec, preserving any existing fields
					// (e.g. connectionPaths set by an earlier `.connect` form)
					const attrPath = path + '.' + attrName;

					if ( specsByPath[ attrPath ] ) {

						specsByPath[ attrPath ].fields.default = parsedValue;
						specsByPath[ attrPath ].fields.typeName = valueType;

					} else {

						specsByPath[ attrPath ] = {
							specType: SpecType.Attribute,
							fields: { default: parsedValue, typeName: valueType }
						};

					}

				}

			}

		}

	}

	_parseAttributeValue( valueType, rawValue ) {

		if ( rawValue === undefined || rawValue === null ) return undefined;

		const str = String( rawValue ).trim();

		// Array types
		if ( valueType.endsWith( '[]' ) ) {

			let result;

			// Parse JSON-like arrays
			try {

				// Handle arrays with parentheses like [(1,2,3), (4,5,6)]
				// Remove trailing comma (valid in USDA but not JSON)
				let cleaned = str.replace( /\(/g, '[' ).replace( /\)/g, ']' );
				if ( cleaned.endsWith( ',' ) ) cleaned = cleaned.slice( 0, - 1 );
				const parsed = JSON.parse( cleaned );

				// Flatten nested arrays for types like point3f[]
				result = Array.isArray( parsed ) && Array.isArray( parsed[ 0 ] ) ? parsed.flat() : parsed;

			} catch ( e ) {

				// Try simple array parsing
				const cleaned = str.replace( /[\[\]]/g, '' );
				result = cleaned.split( ',' ).map( s => {

					const trimmed = s.trim();
					const num = parseFloat( trimmed );
					return isNaN( num ) ? trimmed.replace( /"/g, '' ) : num;

				} );

			}

			//reorder (w, x, y, z) to (x, y, z, w)
			if ( valueType.startsWith( 'quat' ) ) {

				for ( let i = 0; i < result.length; i += 4 ) {

					const w = result[ i ];
					result[ i ] = result[ i + 1 ];
					result[ i + 1 ] = result[ i + 2 ];
					result[ i + 2 ] = result[ i + 3 ];
					result[ i + 3 ] = w;

				}

			}

			return result;

		}

		// Vector types (double3, float3, point3f, etc.)
		if ( valueType.includes( '3' ) || valueType.includes( '2' ) || valueType.includes( '4' ) ) {

			// Parse (x, y, z) format
			const cleaned = str.replace( /[()]/g, '' );
			const values = cleaned.split( ',' ).map( s => parseFloat( s.trim() ) );
			return values;

		}

		// Quaternion types (quatf, quatd, quath)
		// Text format is (w, x, y, z), convert to (x, y, z, w)
		if ( valueType.startsWith( 'quat' ) ) {

			const cleaned = str.replace( /[()]/g, '' );
			const values = cleaned.split( ',' ).map( s => parseFloat( s.trim() ) );
			return [ values[ 1 ], values[ 2 ], values[ 3 ], values[ 0 ] ];

		}

		// Matrix types
		if ( valueType.includes( 'matrix' ) ) {

			const cleaned = str.replace( /[()]/g, '' );
			const values = cleaned.split( ',' ).map( s => parseFloat( s.trim() ) );
			return values;

		}

		// Scalar numeric types
		if ( valueType === 'float' || valueType === 'double' || valueType === 'int' ) {

			return parseFloat( str );

		}

		// String/token types
		if ( valueType === 'string' || valueType === 'token' ) {

			return this._parseString( str );

		}

		// Asset path
		if ( valueType === 'asset' ) {

			return str.replace( /@/g, '' ).replace( /"/g, '' );

		}

		// Default: return as string with quotes removed
		return this._parseString( str );

	}

	_parseString( str ) {

		// Remove surrounding quotes
		if ( ( str.startsWith( '"' ) && str.endsWith( '"' ) ) ||
			( str.startsWith( '\'' ) && str.endsWith( '\'' ) ) ) {

			str = str.slice( 1, - 1 );

		}

		// Handle escape sequences
		let result = '';
		let i = 0;

		while ( i < str.length ) {

			if ( str[ i ] === '\\' && i + 1 < str.length ) {

				const next = str[ i + 1 ];

				switch ( next ) {

					case 'n': result += '\n'; break;
					case 't': result += '\t'; break;
					case 'r': result += '\r'; break;
					case '\\': result += '\\'; break;
					case '"': result += '"'; break;
					case '\'': result += '\''; break;
					default: result += next; break;

				}

				i += 2;

			} else {

				result += str[ i ];
				i ++;

			}

		}

		return result;

	}

}

export { USDAParser };