The modified grammar in B.1.4 actually changes the behavior in Unicode mode

[nhahtdh]

According to B.1.4:

This alternative pattern grammar and semantics only changes the syntax and semantics of BMP patterns.

Let's say we have a pattern in Unicode mode, according to the grammar in B.1.4, with the parsing rule:

When parsing using the following grammar, each alternative is considered only if previous production alternatives do not match.

starting from Term[U], we ignore [~U] ExtendedTerm, then try in order [+U] Assertion[U], [+U] Atom[U], [+U] Atom[U] Quantifier. Going into [+U] Atom[U], we encounter Atom[U] :: PatternCharacter, in which PatternCharacter allows free { and } to be treated as literal characters.

However, according to the grammar in 21.2.1, lone { and } without escape can never be parsed as an Atom, in both Unicode and BMP modes.

As we can see here, the statement above seems to contradict the grammar. The grammar in B.1.4 is actually extended to allow { and } to be parsed as Atom in both BMP and Unicode mode.

Which one is supposed to be correct? The statement or the grammar?

[littledan]

Where do you see the interpretation that { is extended in Unicode mode? I see ExtendedPatternCharacter (which adds {) added only from ExtendedAtom, which is included in Term only if the [U] flag is off.

[littledan]

Where do you see the interpretation that { is extended in Unicode mode? I see ExtendedPatternCharacter (which adds {) added only from ExtendedAtom, which is included in Term only if the [U] flag is off.

[nhahtdh]

@littledan If you follow my link above, the following production leads to my conclusion:

Set +U. Go to Atom[U] in this rule (we need to try Assertion[U] first, but we know that it will fail)

Term[U] ::
[~U] ExtendedTerm
[+U] Assertion[U]
[+U] Atom[U]
[+U] Atom[U] Quantifier

In Atom[U], enter PatternCharacter

Atom[U] ::
PatternCharacter
.
\ AtomEscape[?U]
CharacterClass[?U]
( Disjunction[?U] )
( ? : Disjunction[?U] )

PatternCharacter actually allows braces:

PatternCharacter ::
SourceCharacter but not one of
^ $ \ . * + ? ( ) [ ] |

Anyway, this issue only affects EMCA 6.0 specs. I don't see the issue in 7.0 specs.

[nhahtdh]

@littledan If you follow my link above, the following production leads to my conclusion:

Set +U. Go to Atom[U] in this rule (we need to try Assertion[U] first, but we know that it will fail)

Term[U] ::
[~U] ExtendedTerm
[+U] Assertion[U]
[+U] Atom[U]
[+U] Atom[U] Quantifier

In Atom[U], enter PatternCharacter

Atom[U] ::
PatternCharacter
.
\ AtomEscape[?U]
CharacterClass[?U]
( Disjunction[?U] )
( ? : Disjunction[?U] )

PatternCharacter actually allows braces:

PatternCharacter ::
SourceCharacter but not one of
^ $ \ . * + ? ( ) [ ] |

Anyway, this issue only affects EMCA 6.0 specs. I don't see the issue in 7.0 specs.

[littledan]

If you're filing issues, please refer to the current draft spec, which you can find at https://tc39.github.io/ecma262/ . We don't issue errata for older, historical releases, so you're likely to find fixed bugs there.

[littledan]

If you're filing issues, please refer to the current draft spec, which you can find at https://tc39.github.io/ecma262/ . We don't issue errata for older, historical releases, so you're likely to find fixed bugs there.

[nhahtdh]

I guess this issue is no longer relevant then. Closed.

[nhahtdh]

I guess this issue is no longer relevant then. Closed.