New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Correct Regular Expressions Behavior Related to Annex B #58320
base: main
Are you sure you want to change the base?
Correct Regular Expressions Behavior Related to Annex B #58320
Conversation
This PR doesn't have any linked issues. Please open an issue that references this PR. From there we can discuss and prioritise. |
src/compiler/scanner.ts
Outdated
@@ -3390,7 +3400,7 @@ export function createScanner(languageVersion: ScriptTarget, skipTrivia: boolean | |||
error(Diagnostics.Unicode_property_value_expressions_are_only_available_when_the_Unicode_u_flag_or_the_Unicode_Sets_v_flag_is_set, start, pos - start); | |||
} | |||
} | |||
else if (unicodeMode) { | |||
else if (!annexB) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
In Annex B, braces after p
actually should not be parsed at all, but it does provide helpful errors like #58275 (comment)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not really outdated
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The outdated label is just because the comment is on an old revision of the PR and GitHub can't figure out where the comment goes after.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Agreed. I could’ve removed the review and re-comment but finally chose not to.
/\q\u\i\c\k\_\f\o\x\-\j\u\m\p\s/, | ||
|
||
!!! error TS1125: Hexadecimal digit expected. | ||
~~ | ||
!!! error TS1510: '\k' must be followed by a capturing group name enclosed in angle brackets. | ||
|
||
!!! error TS1125: Hexadecimal digit expected. | ||
|
||
!!! error TS1125: Hexadecimal digit expected. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
These are all valid in Annex B. Even /\u{1/
is. In Annex B essentially all weird things are valid, you know. Currently (after #58295) /[\1]/
is valid but /[\8]/
isn’t. I don’t think it’s ideal.
This will probably need to be rebased. |
@jakebailey is correct. #58339 also made some changes to this code, so a rebase or merge from |
I know, I just haven’t got the time to do so. It would be faster if you could do that for me (I will be back soon). |
src/compiler/scanner.ts
Outdated
error(Diagnostics.Numbers_out_of_order_in_quantifier, digitsStart, pos - digitsStart); | ||
} | ||
} | ||
else if (!min) { | ||
if (unicodeMode) { | ||
if (!annexB) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Though it may be redundant, I think it might be better to still indicate unicodeMode
here so that someone editing this code in the future doesn't mistakenly think this only applies to non-Annex B code. It may be better to use unicodeMode || !annexB
and remove the if (unicodeMode) { annexB = false; }
at the top of scanRegularExpressionWorker
.
The same would go for other uses of annexB
as well.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Are you sure you are really fine with a dozen of occurrences of unicodeMode || !annexB
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No, but
const anyUnicodeModeOrNonAnnexB = unicode Mode || !annexB;
would work.
src/compiler/scanner.ts
Outdated
@@ -2801,7 +2811,10 @@ export function createScanner(languageVersion: ScriptTarget, skipTrivia: boolean | |||
scanGroupName(/*isReference*/ true); | |||
scanExpectedChar(CharacterCodes.greaterThan); | |||
} | |||
else if (unicodeMode) { | |||
else { | |||
// This is actually allowed in Annex B if there are no named capturing groups in the regex, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could we keep track of whether we encountered a (?<
during reScanSlashToken
and add an entry to the RegularExpressionFlags
enum? The spec passes NamedCaptureGroups
as a production parameter just as it does for UnicodeMode
and UnicodeSetsMode
, but only ever passes it as ~NamedCaptureGroups
in Annex B.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, didn’t think of this clever but dirty solution.
I can implement this but… the point is still about how much Annex B things we are going to respect. If we allow \k
then should we allow \u
and \x
(and also \8
and \9
inside character classes) too? (#58320 (comment))
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it's better to error on \u
, \x
, \8
, \9
because in those cases you are more likely to have actually meant something different. Writing \k
when there are no named capture groups is far less ambiguous.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If then I would personally lean towards linting also \1
~\7
inside character classes to prevent this kind of mistakes. Or perhaps the opposite, only outside character classes. I don’t know. I understand that you don’t want the syntax checking to be too breaky, but someone must find this useful.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Every decimal escape in a character class I've ever seen has been a bug, so it makes sense to error for that case.
@graphemecluster do you have the time to make changes or respond to comments, or would you like me to take over the final changes to get this merged? |
2868bd6
to
2e62d25
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A few more minor comments to address and this will be ready to merge.
@@ -1556,9 +1556,7 @@ export function createScanner(languageVersion: ScriptTarget, skipTrivia: boolean | |||
tokenFlags |= TokenFlags.ContainsInvalidEscape; | |||
if (isRegularExpression || shouldEmitInvalidEscapeError) { | |||
const code = parseInt(text.substring(start + 1, pos), 8); | |||
if (isRegularExpression !== "annex-b") { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since this was the only place we check for "annex-b"
, we can make isRegularExpression
a boolean
again.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I thought the same too, but there is a isRegularExpression === true
at the bottom
// Annex B treats any unicode mode as the strict syntax. | ||
annexB = false; | ||
} | ||
// Annex B treats any unicode mode as the strict syntax. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The comment explained why we set annexB
to false, but doesn't apply to this whole declaration. I would change it to be more explanatory, i.e.:
// Annex B treats any unicode mode as the strict syntax. | |
// Regular expressions are checked more strictly when either in 'u' or 'v' mode, or | |
// when not using the looser interpretation of the syntax from ECMA-262 Annex B. |
@@ -2887,7 +2903,7 @@ export function createScanner(languageVersion: ScriptTarget, skipTrivia: boolean | |||
return "\\"; | |||
} | |||
pos--; | |||
return scanEscapeSequence(/*shouldEmitInvalidEscapeError*/ unicodeMode, /*isRegularExpression*/ annexB ? "annex-b" : true); | |||
return scanEscapeSequence(/*shouldEmitInvalidEscapeError*/ unicodeMode, /*isRegularExpression*/ anyUnicodeModeOrNonAnnexB || "annex-b"); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since "annexB"
is no longer used in scanEscapeSequence
, this can just be true
:
return scanEscapeSequence(/*shouldEmitInvalidEscapeError*/ unicodeMode, /*isRegularExpression*/ anyUnicodeModeOrNonAnnexB || "annex-b"); | |
return scanEscapeSequence(/*shouldEmitInvalidEscapeError*/ unicodeMode, /*isRegularExpression*/ true); |
&& charCodeUnchecked(p + 1) === CharacterCodes.question | ||
&& charCodeUnchecked(p + 2) === CharacterCodes.lessThan | ||
&& charCodeUnchecked(p + 3) !== CharacterCodes.equals | ||
&& charCodeUnchecked(p + 3) !== CharacterCodes.exclamation |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I missed this in my last review. These need to be range checked:
&& charCodeUnchecked(p + 1) === CharacterCodes.question | |
&& charCodeUnchecked(p + 2) === CharacterCodes.lessThan | |
&& charCodeUnchecked(p + 3) !== CharacterCodes.equals | |
&& charCodeUnchecked(p + 3) !== CharacterCodes.exclamation | |
&& charCodeChecked(p + 1) === CharacterCodes.question | |
&& charCodeChecked(p + 2) === CharacterCodes.lessThan | |
&& charCodeChecked(p + 3) !== CharacterCodes.equals | |
&& charCodeChecked(p + 3) !== CharacterCodes.exclamation |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
One more I missed.
isPreviousTermQuantifiable = true; | ||
break; | ||
} | ||
} | ||
if (max && Number.parseInt(min) > Number.parseInt(max)) { | ||
else if (max && Number.parseInt(min) > Number.parseInt(max) && (anyUnicodeModeOrNonAnnexB || text.charCodeAt(pos) === CharacterCodes.closeBrace)) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
else if (max && Number.parseInt(min) > Number.parseInt(max) && (anyUnicodeModeOrNonAnnexB || text.charCodeAt(pos) === CharacterCodes.closeBrace)) { | |
else if (max && Number.parseInt(min) > Number.parseInt(max) && (anyUnicodeModeOrNonAnnexB || charCodeChecked(pos) === CharacterCodes.closeBrace)) { |
Follow-up of #58295
This fixes issues like the
'}' expected
cases given at #58275 (comment).