v8/test/mjsunit/harmony/public-instance-class-fields.js

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

716 lines
10 KiB
JavaScript
Raw Normal View History

// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
"use strict";
{
class C {
a;
}
assertEquals(undefined, C.a);
let c = new C;
let descriptor = Object.getOwnPropertyDescriptor(c, 'a');
assertTrue(c.hasOwnProperty('a'));
assertTrue(descriptor.writable);
assertTrue(descriptor.enumerable);
assertTrue(descriptor.configurable);
assertEquals(undefined, c.a);
}
{
class C {
x = 1;
constructor() {}
}
let c = new C;
assertEquals(1, c.x);
}
[class] Store class fields initializer on the constructor Previously, the class fields initializer function was stored on a synthetic context allocated variable. This approach had sevaral problems: - We didn't know that class literal had fields until after we had completely parsed the class literal. This meant that we had to go back and fix up the scope of the constructor to have this synthetic variable. This resulted in mismatch between parser and preparsed scope data. - This synthetic variable could potentially resolve to an initializer of an outer class. For ex: class X extends Object { c = 1; constructor() { var t = () => { class P extends Object { constructor() { var t = () => { super(); }; t(); } } super(); } t(); } } In this the inner class P could access the outer class X's initiliazer function. We would have to maintain extra metadata to make sure this doesn't happen. Instead this new approach uses a private symbol to store the initializer function on the class constructor itself. For the base constructor case, we can simply check for a bit on the constructor function literal to see if we need to emit code that loads and calls this initializer function. Therefore, we don't pay the cost of loading this function in case there are no class fields. For the derived constructor case, there are two possiblities: (a) We are in a super() call directly in the derived constructor: In this case we can do a check similar to the base constructor check, we can check for a bit on the derived constructor and emit code for loading and calling the initializer function. This is usually the common case and we don't pay any cost for not using class fields. (b) We are in a super() call inside an arrow function in the derived constructor: In this case, we /always/ emit code to load and call the initializer function. If the function doesn't exist then we have undefined and we don't call anything. Otherwise we call the function. super() can't be called twice so even if we emit code to load and call the initializer function multiple times, it doesn't matter because it would have already been an error. Bug: v8:5367 Change-Id: I7f77cd6493ff84cf0e430a8c1039bc9ac6941a88 Reviewed-on: https://chromium-review.googlesource.com/781660 Commit-Queue: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Georg Neis <neis@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Cr-Commit-Position: refs/heads/master@{#49628}
2017-11-27 09:56:36 +00:00
{
function t() {
class X {
x = 1;
constructor() {}
}
var x = new X;
return x.x;
}
assertEquals(1, t());
}
{
let x = 'a';
class C {
a;
b = x;
c = 1;
hasOwnProperty() { return 1;}
}
let c = new C;
assertEquals(undefined, c.a);
assertEquals('a', c.b);
assertEquals(1, c.c);
assertEquals(undefined, c.d);
assertEquals(1, c.hasOwnProperty());
}
{
class C {
x = Object.freeze(this);
c = 42;
}
assertThrows(() => { new C; }, TypeError);
}
{
class C {
c = this;
d = () => this;
}
let c = new C;
assertEquals(c, c.c);
assertEquals(c, c.d());
assertEquals(undefined, C.c);
assertEquals(undefined, C.d);
}
{
class C {
c = 1;
d = this.c;
}
let c = new C;
assertEquals(1, c.c);
assertEquals(1, c.d);
assertEquals(undefined, C.c);
assertEquals(undefined, C.d);
}
{
class C {
b = 1;
c = () => this.b;
}
let c = new C;
assertEquals(1, c.b);
assertEquals(1, c.c());
assertEquals(undefined, C.c);
assertEquals(undefined, C.b);
}
{
let x = 'a';
class C {
b = 1;
c = () => this.b;
e = () => x;
}
let c = new C;
assertEquals(1, c.b);
assertEquals('a', c.e());
let a = {b : 2 };
assertEquals(1, c.c.call(a));
assertEquals(undefined, C.b);
assertEquals(undefined, C.c);
}
{
let x = 'a';
class C {
c = 1;
d = function() { return this.c; };
e = function() { return x; };
}
let c = new C;
assertEquals(1, c.c);
assertEquals(1, c.d());
assertEquals('a', c.e());
c.c = 2;
assertEquals(2, c.d());
let a = {c : 3 };
assertEquals(3, c.d.call(a));
assertThrows(c.d.bind(undefined));
assertEquals(undefined, C.c);
assertEquals(undefined, C.d);
assertEquals(undefined, C.e);
}
{
class C {
c = function() { return 1 };
}
let c = new C;
assertEquals('c', c.c.name);
}
{
let d = function() { return new.target; }
class C {
c = d;
}
let c = new C;
assertEquals(undefined, c.c());
assertEquals(new d, new c.c());
}
{
class C {
c = () => new.target;
}
let c = new C;
assertEquals(undefined, c.c());
}
{
let run = false;
class C {
c = () => {
let b;
class A {
constructor() {
b = new.target;
}
};
new A;
run = true;
assertEquals(A, b);
}
}
let c = new C;
c.c();
assertTrue(run);
}
{
class C {
c = new.target;
}
let c = new C;
assertEquals(undefined, c.c);
}
{
class B {
c = 1;
}
class C extends B {}
let c = new C;
assertEquals(1, c.c);
}
{
assertThrows(() => {
class C {
c = new C;
}
let c = new C;
});
}
(function test() {
function makeC() {
var x = 1;
return class {
a = () => () => x;
}
}
let C = makeC();
let c = new C;
let f = c.a();
assertEquals(1, f());
})()
{
let c1 = "c";
class C {
["a"] = 1;
["b"];
[c1];
}
let c = new C;
assertEquals(1, c.a);
assertEquals(undefined, c.b);
assertEquals(undefined, c[c1]);
}
{
let log = [];
function run(i) {
log.push(i);
return i;
}
class C {
[run(1)] = run(6);
[run(2)] = run(7);
[run(3)]() { run(9);}
[run(4)] = run(8);
[run(5)]() { throw new Error('should not execute');};
}
let c = new C;
c[3]();
assertEquals([1, 2, 3, 4, 5, 6, 7, 8, 9], log);
}
function x() {
// This tests lazy parsing.
return function() {
let log = [];
function run(i) {
log.push(i);
return i;
}
class C {
[run(1)] = run(6);
[run(2)] = run(7);
[run(3)]() { run(9);}
[run(4)] = run(8);
[run(5)]() { throw new Error('should not execute');};
}
let c = new C;
c[3]();
assertEquals([1, 2, 3, 4, 5, 6, 7, 8, 9], log);
}
}
x()();
{
class C {}
class D {
[C];
}
let d = new D;
assertThrows(() => { class X { [X] } let x = new X;});
assertEquals(undefined, d[C]);
}
[class] Store class fields initializer on the constructor Previously, the class fields initializer function was stored on a synthetic context allocated variable. This approach had sevaral problems: - We didn't know that class literal had fields until after we had completely parsed the class literal. This meant that we had to go back and fix up the scope of the constructor to have this synthetic variable. This resulted in mismatch between parser and preparsed scope data. - This synthetic variable could potentially resolve to an initializer of an outer class. For ex: class X extends Object { c = 1; constructor() { var t = () => { class P extends Object { constructor() { var t = () => { super(); }; t(); } } super(); } t(); } } In this the inner class P could access the outer class X's initiliazer function. We would have to maintain extra metadata to make sure this doesn't happen. Instead this new approach uses a private symbol to store the initializer function on the class constructor itself. For the base constructor case, we can simply check for a bit on the constructor function literal to see if we need to emit code that loads and calls this initializer function. Therefore, we don't pay the cost of loading this function in case there are no class fields. For the derived constructor case, there are two possiblities: (a) We are in a super() call directly in the derived constructor: In this case we can do a check similar to the base constructor check, we can check for a bit on the derived constructor and emit code for loading and calling the initializer function. This is usually the common case and we don't pay any cost for not using class fields. (b) We are in a super() call inside an arrow function in the derived constructor: In this case, we /always/ emit code to load and call the initializer function. If the function doesn't exist then we have undefined and we don't call anything. Otherwise we call the function. super() can't be called twice so even if we emit code to load and call the initializer function multiple times, it doesn't matter because it would have already been an error. Bug: v8:5367 Change-Id: I7f77cd6493ff84cf0e430a8c1039bc9ac6941a88 Reviewed-on: https://chromium-review.googlesource.com/781660 Commit-Queue: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Georg Neis <neis@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Cr-Commit-Position: refs/heads/master@{#49628}
2017-11-27 09:56:36 +00:00
{
class B {
a = 1;
}
class C extends B {
b = 2;
constructor() {
super();
}
}
let c = new C;
assertEquals(1, c.a);
assertEquals(2, c.b);
}
{
var log = [];
function addToLog(item) { log.push(item); }
class B {
a = 1;
constructor() {
addToLog("base constructor");
}
}
function initF() {
addToLog("init f");
return 1;
}
class C extends B {
f = initF();
constructor() {
addToLog("derived constructor");
var t = () => {
addToLog("t");
if (1==-1) {
super();
} else {
super();
}
}
(() => {
addToLog("anon");
t();
})();
}
}
let c = new C;
assertEquals(1, c.f);
assertEquals(1, c.a);
assertEquals(["derived constructor","anon","t","base constructor","init f"],
log);
}
{
class B {
a = 1;
returnA = () => this.a;
}
class C extends B {
c = this.a;
d = 2;
returnC = () => this.c;
returnD = () => this.d;
}
let c = new C;
assertEquals(1, c.a);
assertEquals(1, c.returnA());
assertEquals(1, c.c);
assertEquals(1, c.returnA());
assertEquals(1, c.returnC());
assertEquals(2, c.d);
assertEquals(2, c.returnD());
let c2 = new C;
assertNotEquals(c2.returnA, c.returnA);
assertNotEquals(c2.returnC, c.returnC);
assertNotEquals(c2.returnD, c.returnD);
}
{
let foo = undefined;
class B {
set d(x) {
foo = x;
}
}
class C extends B {
d = 2;
}
let c = new C;
assertEquals(undefined, foo);
assertEquals(2, c.d);
}
{
class B {}
class C extends B {
constructor() {
super();
}
c = 1;
}
let c = new C;
assertEquals(1, c.c);
}
{
class B {}
class C extends B {
constructor() {
let t = () => {
super();
}
t();
}
c = 1;
}
let c = new C;
assertEquals(1, c.c);
}
{
let log = [];
class B {}
class C extends B {
x = (log.push(1), 1);
constructor() {
let t = () => {
class D extends B {
x = (log.push(2), 2);
constructor() {
let p = () => {
super();
}
p();
}
}
let d = new D();
assertEquals(2, d.x);
super();
}
t();
}
}
let c = new C;
assertEquals(1, c.x);
assertEquals([2, 1], log);
}
{
let log = [];
class C1 extends class {} {
x = log.push(1);
constructor() {
var t = () => super();
super();
t();
}
}
assertThrows(() => new C1, ReferenceError);
assertEquals([1], log);
[class] Store class fields initializer on the constructor Previously, the class fields initializer function was stored on a synthetic context allocated variable. This approach had sevaral problems: - We didn't know that class literal had fields until after we had completely parsed the class literal. This meant that we had to go back and fix up the scope of the constructor to have this synthetic variable. This resulted in mismatch between parser and preparsed scope data. - This synthetic variable could potentially resolve to an initializer of an outer class. For ex: class X extends Object { c = 1; constructor() { var t = () => { class P extends Object { constructor() { var t = () => { super(); }; t(); } } super(); } t(); } } In this the inner class P could access the outer class X's initiliazer function. We would have to maintain extra metadata to make sure this doesn't happen. Instead this new approach uses a private symbol to store the initializer function on the class constructor itself. For the base constructor case, we can simply check for a bit on the constructor function literal to see if we need to emit code that loads and calls this initializer function. Therefore, we don't pay the cost of loading this function in case there are no class fields. For the derived constructor case, there are two possiblities: (a) We are in a super() call directly in the derived constructor: In this case we can do a check similar to the base constructor check, we can check for a bit on the derived constructor and emit code for loading and calling the initializer function. This is usually the common case and we don't pay any cost for not using class fields. (b) We are in a super() call inside an arrow function in the derived constructor: In this case, we /always/ emit code to load and call the initializer function. If the function doesn't exist then we have undefined and we don't call anything. Otherwise we call the function. super() can't be called twice so even if we emit code to load and call the initializer function multiple times, it doesn't matter because it would have already been an error. Bug: v8:5367 Change-Id: I7f77cd6493ff84cf0e430a8c1039bc9ac6941a88 Reviewed-on: https://chromium-review.googlesource.com/781660 Commit-Queue: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Georg Neis <neis@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Cr-Commit-Position: refs/heads/master@{#49628}
2017-11-27 09:56:36 +00:00
log = [];
class C2 extends class {} {
x = log.push(1);
constructor() {
var t = () => super();
t();
super();
}
}
assertThrows(() => new C2, ReferenceError);
assertEquals([1], log);
[class] Store class fields initializer on the constructor Previously, the class fields initializer function was stored on a synthetic context allocated variable. This approach had sevaral problems: - We didn't know that class literal had fields until after we had completely parsed the class literal. This meant that we had to go back and fix up the scope of the constructor to have this synthetic variable. This resulted in mismatch between parser and preparsed scope data. - This synthetic variable could potentially resolve to an initializer of an outer class. For ex: class X extends Object { c = 1; constructor() { var t = () => { class P extends Object { constructor() { var t = () => { super(); }; t(); } } super(); } t(); } } In this the inner class P could access the outer class X's initiliazer function. We would have to maintain extra metadata to make sure this doesn't happen. Instead this new approach uses a private symbol to store the initializer function on the class constructor itself. For the base constructor case, we can simply check for a bit on the constructor function literal to see if we need to emit code that loads and calls this initializer function. Therefore, we don't pay the cost of loading this function in case there are no class fields. For the derived constructor case, there are two possiblities: (a) We are in a super() call directly in the derived constructor: In this case we can do a check similar to the base constructor check, we can check for a bit on the derived constructor and emit code for loading and calling the initializer function. This is usually the common case and we don't pay any cost for not using class fields. (b) We are in a super() call inside an arrow function in the derived constructor: In this case, we /always/ emit code to load and call the initializer function. If the function doesn't exist then we have undefined and we don't call anything. Otherwise we call the function. super() can't be called twice so even if we emit code to load and call the initializer function multiple times, it doesn't matter because it would have already been an error. Bug: v8:5367 Change-Id: I7f77cd6493ff84cf0e430a8c1039bc9ac6941a88 Reviewed-on: https://chromium-review.googlesource.com/781660 Commit-Queue: Sathya Gunasekaran <gsathya@chromium.org> Reviewed-by: Georg Neis <neis@chromium.org> Reviewed-by: Mythri Alle <mythria@chromium.org> Cr-Commit-Position: refs/heads/master@{#49628}
2017-11-27 09:56:36 +00:00
}
{
class C1 extends class {} {
x = 1
constructor() {
eval("super()");
}
}
let c = new C1;
assertEquals(1, c.x);
class C2 extends class {} {
x = 1
constructor() {
var t = () => {
eval("super()");
}
t();
}
}
c = new C2;
assertEquals(1, c.x);
}
{
class C {
['x'] = 1;
['y'] = 2;
}
class C1 extends C {
['x'] = 3;
['z'] = 4;
}
let c = new C1;
assertEquals(3, c.x);
assertEquals(2, c.y);
assertEquals(4, c.z);
}
{
class X extends class {} {
c = 1;
constructor() {
let t = () => {
class P extends class {} {
constructor() {
let t = () => { super(); };
t();
}
}
let p = new P;
assertEquals(undefined, p.c);
super();
}
t();
}
}
let x = new X;
assertEquals(1, x.c);
}
{
class A {
a() { return 1; }
}
class C extends A {
b = super.a();
c = () => super.a;
d = () => super.a();
e = super.a;
f = super.b;
}
let c = new C;
assertEquals(1, c.a());
assertEquals(1, c.b);
assertEquals(1, c.c()());
assertEquals(1, c.d());
assertEquals(1, c.e());
assertFalse(Object.hasOwnProperty(c, 'a'));
assertEquals(c.a, c.e);
assertEquals(undefined, c.f);
}
{
function t() {
return class {
['x'] = 1;
}
}
let klass = t();
let obj = new klass;
assertEquals(1, obj.x);
}
{
new class {
t = 1;
constructor(t = this.t) {
assertEquals(1, t);
}
}
new class extends class {} {
t = 1;
constructor(t = (super(), this.t)) {
assertEquals(1, t);
}
}
assertThrows(() => {
new class extends class {} {
t = 1;
constructor(t = this.t) {
super();
}
}
}, ReferenceError);
}
{
class X {
p = function() { return arguments[0]; }
}
let x = new X;
assertEquals(1, x.p(1));
}
{
class X {
t = () => {
function p() { return arguments[0]; };
return p;
}
}
let x = new X;
let p = x.t();
assertEquals(1, p(1));
}
{
class X {
t = () => {
function p() { return eval("arguments[0]"); };
return p;
}
}
let x = new X;
let p = x.t();
assertEquals(1, p(1));
}
{
class X {
p = eval("(function() { return arguments[0]; })(1)");
}
let x = new X;
assertEquals(1, x.p);
}
{
let thisInInitializer, thisInConstructor, thisFromArrowFn, arrowFn;
let C = class extends class {} {
field = (thisInInitializer = this, thisFromArrowFn = arrowFn());
constructor() {
arrowFn = () => this;
super();
thisInConstructor = this;
}
};
let c = new C();
assertSame(thisInInitializer, c);
assertSame(thisFromArrowFn, c);
assertSame(thisInConstructor, c);
}