You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
pegjs/lib/compiler/passes/generate-bytecode.js

681 lines
18 KiB
JavaScript

"use strict";
const asts = require( "../asts" );
const js = require( "../js" );
const op = require( "../opcodes" );
const visitor = require( "../visitor" );
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
// Generates bytecode.
//
// Instructions
// ============
//
// Stack Manipulation
// ------------------
//
// [0] PUSH c
//
// stack.push(consts[c]);
//
// [1] PUSH_UNDEFINED
//
// stack.push(undefined);
//
// [2] PUSH_NULL
//
// stack.push(null);
//
// [3] PUSH_FAILED
//
// stack.push(FAILED);
//
// [4] PUSH_EMPTY_ARRAY
//
// stack.push([]);
//
// [5] PUSH_CURR_POS
//
// stack.push(currPos);
//
// [6] POP
//
// stack.pop();
//
// [7] POP_CURR_POS
//
// currPos = stack.pop();
//
// [8] POP_N n
//
// stack.pop(n);
//
// [9] NIP
//
// value = stack.pop();
// stack.pop();
// stack.push(value);
//
// [10] APPEND
//
// value = stack.pop();
// array = stack.pop();
// array.push(value);
// stack.push(array);
//
// [11] WRAP n
//
// stack.push(stack.pop(n));
//
// [12] TEXT
//
// stack.push(input.substring(stack.pop(), currPos));
//
// Conditions and Loops
// --------------------
//
// [13] IF t, f
//
// if (stack.top()) {
// interpret(ip + 3, ip + 3 + t);
// } else {
// interpret(ip + 3 + t, ip + 3 + t + f);
// }
//
// [14] IF_ERROR t, f
//
// if (stack.top() === FAILED) {
// interpret(ip + 3, ip + 3 + t);
// } else {
// interpret(ip + 3 + t, ip + 3 + t + f);
// }
//
// [15] IF_NOT_ERROR t, f
//
// if (stack.top() !== FAILED) {
// interpret(ip + 3, ip + 3 + t);
// } else {
// interpret(ip + 3 + t, ip + 3 + t + f);
// }
//
// [16] WHILE_NOT_ERROR b
//
// while(stack.top() !== FAILED) {
// interpret(ip + 2, ip + 2 + b);
// }
//
// Matching
// --------
//
// [17] MATCH_ANY a, f, ...
//
// if (input.length > currPos) {
// interpret(ip + 3, ip + 3 + a);
// } else {
// interpret(ip + 3 + a, ip + 3 + a + f);
// }
//
// [18] MATCH_STRING s, a, f, ...
//
// if (input.substr(currPos, consts[s].length) === consts[s]) {
// interpret(ip + 4, ip + 4 + a);
// } else {
// interpret(ip + 4 + a, ip + 4 + a + f);
// }
//
// [19] MATCH_STRING_IC s, a, f, ...
//
// if (input.substr(currPos, consts[s].length).toLowerCase() === consts[s]) {
// interpret(ip + 4, ip + 4 + a);
// } else {
// interpret(ip + 4 + a, ip + 4 + a + f);
// }
//
// [20] MATCH_REGEXP r, a, f, ...
//
// if (consts[r].test(input.charAt(currPos))) {
// interpret(ip + 4, ip + 4 + a);
// } else {
// interpret(ip + 4 + a, ip + 4 + a + f);
// }
//
// [21] ACCEPT_N n
//
// stack.push(input.substring(currPos, n));
// currPos += n;
//
// [22] ACCEPT_STRING s
//
// stack.push(consts[s]);
// currPos += consts[s].length;
//
// [23] FAIL e
//
// stack.push(FAILED);
// fail(consts[e]);
//
// Calls
// -----
//
// [24] LOAD_SAVED_POS p
//
// savedPos = stack[p];
//
// [25] UPDATE_SAVED_POS
//
// savedPos = currPos;
//
// [26] CALL f, n, pc, p1, p2, ..., pN
//
// value = consts[f](stack[p1], ..., stack[pN]);
// stack.pop(n);
// stack.push(value);
//
// Rules
// -----
//
// [27] RULE r
//
// stack.push(parseRule(r));
//
// Failure Reporting
// -----------------
//
// [28] SILENT_FAILS_ON
//
// silentFails++;
//
// [29] SILENT_FAILS_OFF
//
// silentFails--;
function generateBytecode( ast ) {
const consts = [];
let generate;
function addConst( value ) {
const index = consts.indexOf( value );
return index === -1 ? consts.push( value ) - 1 : index;
}
function addFunctionConst( params, code ) {
return addConst( `function(${ params.join( ", " ) }) {${ code }}` );
}
function cloneEnv( env ) {
const clone = {};
Object.keys( env ).forEach( name => {
clone[ name ] = env[ name ];
} );
return clone;
}
function buildSequence() {
return Array.prototype.concat.apply( [], arguments );
}
function buildCondition( condCode, thenCode, elseCode ) {
return condCode.concat(
[ thenCode.length, elseCode.length ],
thenCode,
elseCode
);
}
function buildLoop( condCode, bodyCode ) {
return condCode.concat( [ bodyCode.length ], bodyCode );
}
function buildCall( functionIndex, delta, env, sp ) {
const params = Object.keys( env ).map( name => sp - env[ name ] );
return [ op.CALL, functionIndex, delta, params.length ].concat( params );
}
function buildSimplePredicate( expression, negative, context ) {
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
return buildSequence(
[ op.PUSH_CURR_POS ],
[ op.SILENT_FAILS_ON ],
generate( expression, {
sp: context.sp + 1,
env: cloneEnv( context.env ),
action: null
} ),
[ op.SILENT_FAILS_OFF ],
buildCondition(
[ negative ? op.IF_ERROR : op.IF_NOT_ERROR ],
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
buildSequence(
[ op.POP ],
[ negative ? op.POP : op.POP_CURR_POS ],
[ op.PUSH_UNDEFINED ]
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
),
buildSequence(
[ op.POP ],
[ negative ? op.POP_CURR_POS : op.POP ],
[ op.PUSH_FAILED ]
)
)
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
);
}
function buildSemanticPredicate( code, negative, context ) {
const functionIndex = addFunctionConst( Object.keys( context.env ), code );
return buildSequence(
[ op.UPDATE_SAVED_POS ],
buildCall( functionIndex, 0, context.env, context.sp ),
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
buildCondition(
[ op.IF ],
buildSequence( [ op.POP ], negative ? [ op.PUSH_FAILED ] : [ op.PUSH_UNDEFINED ] ),
buildSequence( [ op.POP ], negative ? [ op.PUSH_UNDEFINED ] : [ op.PUSH_FAILED ] )
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
)
);
}
function buildAppendLoop( expressionCode ) {
return buildLoop(
[ op.WHILE_NOT_ERROR ],
buildSequence( [ op.APPEND ], expressionCode )
);
}
generate = visitor.build( {
grammar( node ) {
node.rules.forEach( generate );
node.consts = consts;
},
rule( node ) {
node.bytecode = generate( node.expression, {
sp: -1, // stack pointer
env: { }, // mapping of label names to stack positions
action: null // action nodes pass themselves to children here
} );
},
named( node, context ) {
const nameIndex = addConst(
`peg$otherExpectation("${ js.stringEscape( node.name ) }")`
);
// The code generated below is slightly suboptimal because |FAIL| pushes
// to the stack, so we need to stick a |POP| in front of it. We lack a
// dedicated instruction that would just report the failure and not touch
// the stack.
return buildSequence(
[ op.SILENT_FAILS_ON ],
generate( node.expression, context ),
[ op.SILENT_FAILS_OFF ],
buildCondition( [ op.IF_ERROR ], [ op.FAIL, nameIndex ], [] )
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
);
},
choice( node, context ) {
function buildAlternativesCode( alternatives, context ) {
return buildSequence(
generate( alternatives[ 0 ], {
sp: context.sp,
env: cloneEnv( context.env ),
action: null
} ),
alternatives.length < 2
? []
: buildCondition(
[ op.IF_ERROR ],
buildSequence(
[ op.POP ],
buildAlternativesCode( alternatives.slice( 1 ), context )
),
[]
)
);
}
return buildAlternativesCode( node.alternatives, context );
},
action( node, context ) {
const env = cloneEnv( context.env );
const emitCall = node.expression.type !== "sequence" || node.expression.elements.length === 0;
const expressionCode = generate( node.expression, {
sp: context.sp + ( emitCall ? 1 : 0 ),
env: env,
action: node
} );
const functionIndex = addFunctionConst( Object.keys( env ), node.code );
return emitCall === false
? expressionCode
: buildSequence(
[ op.PUSH_CURR_POS ],
expressionCode,
buildCondition(
[ op.IF_NOT_ERROR ],
buildSequence(
[ op.LOAD_SAVED_POS, 1 ],
buildCall( functionIndex, 1, env, context.sp + 2 )
),
[]
),
[ op.NIP ]
);
},
sequence( node, context ) {
function buildElementsCode( elements, context ) {
if ( elements.length > 0 ) {
const processedCount = node.elements.length - elements.slice( 1 ).length;
return buildSequence(
generate( elements[ 0 ], {
sp: context.sp,
env: context.env,
action: null
} ),
buildCondition(
[ op.IF_NOT_ERROR ],
buildElementsCode( elements.slice( 1 ), {
sp: context.sp + 1,
env: context.env,
action: context.action
} ),
buildSequence(
processedCount > 1 ? [ op.POP_N, processedCount ] : [ op.POP ],
[ op.POP_CURR_POS ],
[ op.PUSH_FAILED ]
)
)
);
} else if ( context.action ) {
const functionIndex = addFunctionConst(
Object.keys( context.env ),
context.action.code
);
return buildSequence(
[ op.LOAD_SAVED_POS, node.elements.length ],
buildCall(
functionIndex,
node.elements.length + 1,
context.env,
context.sp
)
);
}
return buildSequence( [ op.WRAP, node.elements.length ], [ op.NIP ] );
}
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
return buildSequence(
[ op.PUSH_CURR_POS ],
buildElementsCode( node.elements, {
sp: context.sp + 1,
env: context.env,
action: context.action
} )
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
);
},
labeled( node, context ) {
const env = cloneEnv( context.env );
context.env[ node.label ] = context.sp + 1;
return generate( node.expression, {
sp: context.sp,
env: env,
action: null
} );
},
text( node, context ) {
return buildSequence(
[ op.PUSH_CURR_POS ],
generate( node.expression, {
sp: context.sp + 1,
env: cloneEnv( context.env ),
action: null
} ),
buildCondition(
[ op.IF_NOT_ERROR ],
buildSequence( [ op.POP ], [ op.TEXT ] ),
[ op.NIP ]
)
);
},
simple_and( node, context ) {
return buildSimplePredicate( node.expression, false, context );
},
simple_not( node, context ) {
return buildSimplePredicate( node.expression, true, context );
},
optional( node, context ) {
return buildSequence(
generate( node.expression, {
sp: context.sp,
env: cloneEnv( context.env ),
action: null
} ),
buildCondition(
[ op.IF_ERROR ],
buildSequence( [ op.POP ], [ op.PUSH_NULL ] ),
[]
)
);
},
zero_or_more( node, context ) {
const expressionCode = generate( node.expression, {
sp: context.sp + 1,
env: cloneEnv( context.env ),
action: null
} );
return buildSequence(
[ op.PUSH_EMPTY_ARRAY ],
expressionCode,
buildAppendLoop( expressionCode ),
[ op.POP ]
);
},
one_or_more( node, context ) {
const expressionCode = generate( node.expression, {
sp: context.sp + 1,
env: cloneEnv( context.env ),
action: null
} );
return buildSequence(
[ op.PUSH_EMPTY_ARRAY ],
expressionCode,
buildCondition(
[ op.IF_NOT_ERROR ],
buildSequence( buildAppendLoop( expressionCode ), [ op.POP ] ),
buildSequence( [ op.POP ], [ op.POP ], [ op.PUSH_FAILED ] )
)
);
},
group( node, context ) {
return generate( node.expression, {
sp: context.sp,
env: cloneEnv( context.env ),
action: null
} );
},
semantic_and( node, context ) {
return buildSemanticPredicate( node.code, false, context );
},
semantic_not( node, context ) {
return buildSemanticPredicate( node.code, true, context );
},
rule_ref( node ) {
return [ op.RULE, asts.indexOfRule( ast, node.name ) ];
},
literal( node ) {
if ( node.value.length > 0 ) {
const stringIndex = addConst( `"${ js.stringEscape(
node.ignoreCase ? node.value.toLowerCase() : node.value
) }"` );
const expectedIndex = addConst(
"peg$literalExpectation("
+ `"${ js.stringEscape( node.value ) }", `
+ node.ignoreCase
+ ")"
);
// For case-sensitive strings the value must match the beginning of the
// remaining input exactly. As a result, we can use |ACCEPT_STRING| and
// save one |substr| call that would be needed if we used |ACCEPT_N|.
return buildCondition(
node.ignoreCase
? [ op.MATCH_STRING_IC, stringIndex ]
: [ op.MATCH_STRING, stringIndex ],
node.ignoreCase
? [ op.ACCEPT_N, node.value.length ]
: [ op.ACCEPT_STRING, stringIndex ],
[ op.FAIL, expectedIndex ]
);
}
const stringIndex = addConst( "\"\"" );
return [ op.PUSH, stringIndex ];
},
class( node ) {
const regexp = "/^["
+ ( node.inverted ? "^" : "" )
+ node.parts
.map( part =>
( Array.isArray( part )
? js.regexpClassEscape( part[ 0 ] )
+ "-"
+ js.regexpClassEscape( part[ 1 ] )
: js.regexpClassEscape( part ) )
)
.join( "" )
+ "]/"
+ ( node.ignoreCase ? "i" : "" );
const parts = "["
+ node.parts
.map( part =>
( Array.isArray( part )
? `["${ js.stringEscape( part[ 0 ] ) }", "${ js.stringEscape( part[ 1 ] ) }"]`
: "\"" + js.stringEscape( part ) + "\"" )
)
.join( ", " )
+ "]";
const regexpIndex = addConst( regexp );
const expectedIndex = addConst(
"peg$classExpectation("
+ parts + ", "
+ node.inverted + ", "
+ node.ignoreCase
+ ")"
);
return buildCondition(
[ op.MATCH_REGEXP, regexpIndex ],
[ op.ACCEPT_N, 1 ],
[ op.FAIL, expectedIndex ]
);
},
any() {
const expectedIndex = addConst( "peg$anyExpectation()" );
return buildCondition(
[ op.MATCH_ANY ],
[ op.ACCEPT_N, 1 ],
[ op.FAIL, expectedIndex ]
);
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
}
} );
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
generate( ast );
Code generator rewrite This is a complete rewrite of the PEG.js code generator. Its goals are: 1. Allow optimizing the generated parser code for code size as well as for parsing speed. 2. Prepare ground for future optimizations and big features (like incremental parsing). 2. Replace the old template-based code-generation system with something more lightweight and flexible. 4. General code cleanup (structure, style, variable names, ...). New Architecture ---------------- The new code generator consists of two steps: * Bytecode generator -- produces bytecode for an abstract virtual machine * JavaScript generator -- produces JavaScript code based on the bytecode The abstract virtual machine is stack-based. Originally I wanted to make it register-based, but it turned out that all the code related to it would be more complex and the bytecode itself would be longer (because of explicit register specifications in instructions). The only downsides of the stack-based approach seem to be few small inefficiencies (see e.g. the |NIP| instruction), which seem to be insignificant. The new generator allows optimizing for parsing speed or code size (you can choose using the |optimize| option of the |PEG.buildParser| method or the --optimize/-o option on the command-line). When optimizing for size, the JavaScript generator emits the bytecode together with its constant table and a generic bytecode interpreter. Because the interpreter is small and the bytecode and constant table grow only slowly with size of the grammar, the resulting parser is also small. When optimizing for speed, the JavaScript generator just compiles the bytecode into JavaScript. The generated code is relatively efficient, so the resulting parser is fast. Internal Identifiers -------------------- As a small bonus, all internal identifiers visible to user code in the initializer, actions and predicates are prefixed by |peg$|. This lowers the chance that identifiers in user code will conflict with the ones from PEG.js. It also makes using any internals in user code ugly, which is a good thing. This solves GH-92. Performance ----------- The new code generator improved parsing speed and parser code size significantly. The generated parsers are now: * 39% faster when optimizing for speed * 69% smaller when optimizing for size (without minification) * 31% smaller when optimizing for size (with minification) (Parsing speed was measured using the |benchmark/run| script. Code size was measured by generating parsers for examples in the |examples| directory and adding up the file sizes. Minification was done by |uglify --ascii| in version 1.3.4.) Final Note ---------- This is just a beginning! The new code generator lays a foundation upon which many optimizations and improvements can (and will) be made. Stay tuned :-)
12 years ago
}
module.exports = generateBytecode;