Normative: Allow [[ReferencedName]] in Reference Records to be a not-yet-resolved property key #3307

rkirsling · 2024-03-29T10:20:02Z

Resolves #3295.

bakkot · 2024-03-29T15:58:16Z

spec.html

@@ -19147,8 +19153,7 @@ <h1>
      <emu-alg>
        1. Let _propertyNameReference_ be ? Evaluation of _expression_.
        1. Let _propertyNameValue_ be ? GetValue(_propertyNameReference_).
-        1. Let _propertyKey_ be ? ToPropertyKey(_propertyNameValue_).
-        1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyKey_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.
+        1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyNameValue_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.


Suggested change

1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyNameValue_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.

1. NOTE: In most cases, ToPropertyKey(_propertyNameValue_) will be performed immediately after this step. However, in the case of `a[b] = c`, it will not be performed until after evaluation of `c`.

1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyNameValue_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.

Wonderful, thanks!

...though it seems like the linter thinks we're performing ToPropertyKey(_propertyNameValue_) instead of mentioning it in a NOTE. 🤔 Tried rephrasing to ditch the parens.

bakkot · 2024-03-29T16:07:55Z

spec.html

@@ -19147,8 +19153,7 @@ <h1>
      <emu-alg>
        1. Let _propertyNameReference_ be ? Evaluation of _expression_.
        1. Let _propertyNameValue_ be ? GetValue(_propertyNameReference_).
-        1. Let _propertyKey_ be ? ToPropertyKey(_propertyNameValue_).
-        1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyKey_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.
+        1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyNameValue_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.


Same change is necessary for super:

class C extends class {} { m() { super[{toString() { console.log('ToPropertyKey'); } }] = console.log('rhs'); } } (new C).m()

prints 'rhs' then 'ToPropertyKey'.

bakkot · 2024-03-29T16:17:45Z

Calls to NamedEvaluation where the argument is _lref_.[[ReferencedName]] (e.g.) should probably be preceded with Assert: _lref_.[[ReferencedName]] is a Property Key or private name. since it is not totally obvious.

Alternatively, these could be changed to pass StringValue of |LeftHandSideExpression| (or whichever the relevant nonterminal is) instead. I think that ends up being clearer anyway, and since StringValue is not side-effecting it's ok to just call it a second time (the first having been during Evaluation).

bakkot · 2024-03-29T16:20:49Z

Looks good other than those comments, thanks for doing this!

spec.html

linusg · 2024-04-10T14:20:35Z

This means the test linked in the referenced issue will be invalidated and needs to be updated or removed, correct?

rkirsling · 2024-04-10T14:35:14Z

That is correct. If anybody's gung-ho to do it, I won't stop 'em, but otherwise I can do so. 😉

Update S11.13.1_A7_T3.js now that consensus has been reached on tc39/ecma262#3307.

rkirsling · 2024-04-11T05:50:00Z

Corresponding test262 PR: tc39/test262#4052

Update S11.13.1_A7_T3.js now that consensus has been reached on tc39/ecma262#3307.

* Update test for o[p] = f() Update S11.13.1_A7_T3.js now that consensus has been reached on tc39/ecma262#3307. * Rename test and add an analogous one for super.

bakkot · 2024-04-17T21:57:33Z

We should update these tests and also add a test for this case before merging.

Ensure that the following cases are covered: - tc39/ecma262#3295 (comment) - tc39/ecma262#2659 (comment)

rkirsling · 2024-04-18T06:22:29Z

Done! tc39/test262#4060

Ensure that the following cases are covered: - tc39/ecma262#3295 (comment) - tc39/ecma262#2659 (comment)

ptomato · 2024-04-22T19:36:48Z

The additional tests are merged.

michaelficarra · 2024-04-22T23:37:23Z

spec.html

@@ -4264,6 +4264,9 @@ <h1>
            1. [id="step-getvalue-toobject"] Let _baseObj_ be ? ToObject(_V_.[[Base]]).
            1. If IsPrivateReference(_V_) is *true*, then
              1. Return ? PrivateGet(_baseObj_, _V_.[[ReferencedName]]).
+            1. If _V_.[[ReferencedName]] is neither a String nor a Symbol, then


Suggested change

1. If _V_.[[ReferencedName]] is neither a String nor a Symbol, then

1. If IsPropertyKey(_V_.[[ReferencedName]]) is *false*, then

Alternatively, I'm not sure why we have that AO instead of just saying "is a property key", since "property key" is a well-defined term. So we could get rid of that AO (as a separate PR) and then just use that phrasing.

Same below for PutValue, of course.

edit: Also evaluation of delete 🤦‍♂️

Sounds good; opened #3316.

syg · 2024-04-24T21:27:37Z

spec.html

-            <td>a String, a Symbol, or a Private Name</td>
-            <td>The name of the binding. Always a String if [[Base]] value is an Environment Record.</td>
+            <td>an ECMAScript language value or a Private Name</td>
+            <td>The name of the binding. Always a String if [[Base]] value is an Environment Record. May temporarily be an ECMAScript language value other than a String or a Symbol in the case of a property reference for which ToPropertyKey has yet to be performed.</td>


I think a sentence of the form "May be an ECMAScript language value [...] until yada yada" would be clearer than "May temporarily be [...]".

Sounds good; added an Otherwise, to your suggestion to clarify that this just applies to property reference cases.

spec.html

syg · 2024-05-01T21:44:28Z

spec.html

@@ -20501,7 +20507,7 @@ <h1>Runtime Semantics: Evaluation</h1>
        1. If |LeftHandSideExpression| is neither an |ObjectLiteral| nor an |ArrayLiteral|, then
          1. Let _lref_ be ? Evaluation of |LeftHandSideExpression|.
          1. If IsAnonymousFunctionDefinition(|AssignmentExpression|) and IsIdentifierRef of |LeftHandSideExpression| are both *true*, then
-            1. Let _rval_ be ? NamedEvaluation of |AssignmentExpression| with argument _lref_.[[ReferencedName]].
+            1. Let _rval_ be ? NamedEvaluation of |AssignmentExpression| with argument StringValue of |LeftHandSideExpression|.


Nit: Here and below, this reads a bit awkward given the invocation of the SDO. Perhaps pull out a separate step like Let _lhs_ be the StringValue of |LeftHandSideExpression|.

Very weak preference.

Related: we already wanted to do this for other reasons. 👍

michaelficarra · 2024-05-16T20:59:49Z

spec.html

@@ -20501,7 +20507,8 @@ <h1>Runtime Semantics: Evaluation</h1>
        1. If |LeftHandSideExpression| is neither an |ObjectLiteral| nor an |ArrayLiteral|, then
          1. Let _lref_ be ? Evaluation of |LeftHandSideExpression|.
          1. If IsAnonymousFunctionDefinition(|AssignmentExpression|) and IsIdentifierRef of |LeftHandSideExpression| are both *true*, then


I know this doesn't have to be part of this PR, but we should update this line to match the other cases below.

Suggested change

1. If IsAnonymousFunctionDefinition(|AssignmentExpression|) and IsIdentifierRef of |LeftHandSideExpression| are both *true*, then

1. If IsAnonymousFunctionDefinition(|AssignmentExpression|) is *true* and IsIdentifierRef of |LeftHandSideExpression| is *true*, then

I've gone ahead and opened a separate PR for it here: #3330

michaelficarra

LGTM

…yet-resolved property key (tc39#3307)

Missing coverage encountered while implementing <tc39/ecma262#3307> in SpiderMonkey. Ensure environment lookups are performed in the correct order: - keyed-destructuring-property-reference-target-evaluation-order-with-bindings.js Ensure `delete super[elem]` steps are correctly performed: - delete/super-property-topropertykey.js - delete/super-property-uninitialized-this.js Ensure ToPropertyKey for computed property names in object literals correctly performed: - object/computed-property-name-topropertykey-before-value-evaluation.js Ensure `GetSuperBase` is executed before `ToPropertKey`: - super/prop-expr-getsuperbase-before-topropertykey-* Ensure `GetThisBinding` is executed first: - super/prop-expr-uninitialized-this-*

bakkot reviewed Mar 29, 2024

View reviewed changes

jmdyck reviewed Mar 29, 2024

View reviewed changes

spec.html Show resolved Hide resolved

rkirsling force-pushed the loosen-referenced-name-in-reference-records branch 2 times, most recently from 82117e4 to 4d1a4b5 Compare March 30, 2024 11:02

rkirsling changed the title ~~Allow [[ReferencedName]] in Reference Records to be a not-yet-resolved property key.~~ Normative: Allow [[ReferencedName]] in Reference Records to be a not-yet-resolved property key Apr 10, 2024

rkirsling added a commit to rkirsling/test262 that referenced this pull request Apr 11, 2024

Update test for o[p] = f()

6cfdc36

Update S11.13.1_A7_T3.js now that consensus has been reached on tc39/ecma262#3307.

rkirsling mentioned this pull request Apr 11, 2024

Update test for o[p] = f() tc39/test262#4052

Merged

ptomato pushed a commit to rkirsling/test262 that referenced this pull request Apr 12, 2024

Update test for o[p] = f()

c6785d1

Update S11.13.1_A7_T3.js now that consensus has been reached on tc39/ecma262#3307.

ptomato added has test262 tests has consensus This has committee consensus. labels Apr 12, 2024

ptomato pushed a commit to tc39/test262 that referenced this pull request Apr 12, 2024

Update test for o[p] = f() (#4052)

142a6a6

* Update test for o[p] = f() Update S11.13.1_A7_T3.js now that consensus has been reached on tc39/ecma262#3307. * Rename test and add an analogous one for super.

bakkot mentioned this pull request Apr 15, 2024

Is null[{ toString: () => { console.log('hit') } }] supposed to hit the console.log? #2659

Closed

woess mentioned this pull request Apr 16, 2024

Reality and spec differ on property key resolution timing for o[p] = f() #3295

Closed

bakkot added needs test262 tests The proposal should specify how to test an implementation. Ideally via github.com/tc39/test262 and removed has test262 tests labels Apr 17, 2024

rkirsling added a commit to rkirsling/test262 that referenced this pull request Apr 18, 2024

Add additional tests for tc39/ecma262#3307.

953ad44

Ensure that the following cases are covered: - tc39/ecma262#3295 (comment) - tc39/ecma262#2659 (comment)

rkirsling added a commit to rkirsling/test262 that referenced this pull request Apr 18, 2024

Add additional tests for tc39/ecma262#3307.

6da4659

Ensure that the following cases are covered: - tc39/ecma262#3295 (comment) - tc39/ecma262#2659 (comment)

rkirsling added a commit to rkirsling/test262 that referenced this pull request Apr 18, 2024

Add additional tests for tc39/ecma262#3307.

c8421c4

Ensure that the following cases are covered: - tc39/ecma262#3295 (comment) - tc39/ecma262#2659 (comment)

ptomato pushed a commit to tc39/test262 that referenced this pull request Apr 22, 2024

Add additional tests for tc39/ecma262#3307.

007b333

Ensure that the following cases are covered: - tc39/ecma262#3295 (comment) - tc39/ecma262#2659 (comment)

ptomato added has test262 tests and removed needs test262 tests The proposal should specify how to test an implementation. Ideally via github.com/tc39/test262 labels Apr 22, 2024

ljharb added the editor call to be discussed in the next editor call label Apr 22, 2024

michaelficarra reviewed Apr 22, 2024

View reviewed changes

rkirsling mentioned this pull request Apr 23, 2024

Editorial: Do away with IsPropertyKey AO. #3316

Merged

woess mentioned this pull request Apr 23, 2024

Incorrect tests after change to evaluation order for member assignment tc39/test262#3407

Open

syg reviewed Apr 24, 2024

View reviewed changes

ljharb removed the editor call to be discussed in the next editor call label Apr 24, 2024

bakkot approved these changes May 1, 2024

View reviewed changes

syg approved these changes May 1, 2024

View reviewed changes

michaelficarra reviewed May 16, 2024

View reviewed changes

michaelficarra approved these changes May 16, 2024

View reviewed changes

jhnaldo mentioned this pull request May 20, 2024

chore(ci): Update branch version for ES-META repo tc39/test262#4089

Merged

bakkot added the ready to merge Editors believe this PR needs no further reviews, and is ready to land. label May 29, 2024

Normative: Allow [[ReferencedName]] in Reference Records to be a not-…

6e70429

…yet-resolved property key (tc39#3307)

ljharb force-pushed the loosen-referenced-name-in-reference-records branch from 7e98c0e to 6e70429 Compare May 30, 2024 23:19

ljharb merged commit 6e70429 into tc39:main May 30, 2024
7 checks passed

anba mentioned this pull request Sep 2, 2024

Improve coverage for reference type in various contexts tc39/test262#4216

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Normative: Allow [[ReferencedName]] in Reference Records to be a not-yet-resolved property key #3307

Normative: Allow [[ReferencedName]] in Reference Records to be a not-yet-resolved property key #3307

rkirsling commented Mar 29, 2024

bakkot Mar 29, 2024

rkirsling Mar 30, 2024

rkirsling Mar 30, 2024 •

edited

Loading

bakkot Mar 29, 2024

bakkot commented Mar 29, 2024

bakkot commented Mar 29, 2024

linusg commented Apr 10, 2024

rkirsling commented Apr 10, 2024

rkirsling commented Apr 11, 2024

bakkot commented Apr 17, 2024

rkirsling commented Apr 18, 2024

ptomato commented Apr 22, 2024

michaelficarra Apr 22, 2024 •

edited

Loading

rkirsling Apr 23, 2024

syg Apr 24, 2024

rkirsling Apr 25, 2024

syg May 1, 2024

michaelficarra May 1, 2024

rkirsling May 2, 2024

michaelficarra May 16, 2024

michaelficarra May 16, 2024

michaelficarra left a comment

	1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyNameValue_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.
	1. NOTE: In most cases, ToPropertyKey(_propertyNameValue_) will be performed immediately after this step. However, in the case of `a[b] = c`, it will not be performed until after evaluation of `c`.
	1. Return the Reference Record { [[Base]]: _baseValue_, [[ReferencedName]]: _propertyNameValue_, [[Strict]]: _strict_, [[ThisValue]]: ~empty~ }.

	1. If _V_.[[ReferencedName]] is neither a String nor a Symbol, then
	1. If IsPropertyKey(_V_.[[ReferencedName]]) is false, then

	1. If IsAnonymousFunctionDefinition(\|AssignmentExpression\|) and IsIdentifierRef of \|LeftHandSideExpression\| are both true, then
	1. If IsAnonymousFunctionDefinition(\|AssignmentExpression\|) is true and IsIdentifierRef of \|LeftHandSideExpression\| is true, then

Normative: Allow [[ReferencedName]] in Reference Records to be a not-yet-resolved property key #3307

Normative: Allow [[ReferencedName]] in Reference Records to be a not-yet-resolved property key #3307

Conversation

rkirsling commented Mar 29, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rkirsling Mar 30, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bakkot commented Mar 29, 2024

bakkot commented Mar 29, 2024

linusg commented Apr 10, 2024

rkirsling commented Apr 10, 2024

rkirsling commented Apr 11, 2024

bakkot commented Apr 17, 2024

rkirsling commented Apr 18, 2024

ptomato commented Apr 22, 2024

michaelficarra Apr 22, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

michaelficarra left a comment

Choose a reason for hiding this comment

rkirsling Mar 30, 2024 •

edited

Loading

michaelficarra Apr 22, 2024 •

edited

Loading