Menu ▾ ▴

#87 CSS3 parsing of :not(X) , :nth-child(X) arguments incomplete

0.9.28
open
nobody
None
1
2019-06-30
2019-06-29
Jeremie Miserez
No

Parsing ":not(FOO)" or ":nth-child(FOO)" results in a PseudoClassConditionImpl with value "not(FOO)" or "nth-child(FOO)", i.e. the argument itself is not parsed any further. This is in SACParserCSS3.

This makes it impossible to differentiate between ":not(#foo)" and ":not(.foo)" without putting the argument string through second parser, i.e. the tree is incomplete. I found this issue when calculating CSS selector specificity using the the current parsers: ":not(#foo)" and ":not(.foo)" return the same single SAC_PSEUDO_CLASS_CONDITION/PseudoClassConditionImpl, although their specifity is different (1-0-0 vs 0-1-0) and there could be more selectors inside.

I guess the issue only is that CSS3 parsing/grammar is incomplete. I'll see if I can attach a patch, but it may take a while.

See here for more:
- https://www.w3.org/TR/selectors-3/#negation
- https://www.w3.org/TR/selectors-3/#nth-child-pseudo

Discussion

  • Jeremie Miserez

    Jeremie Miserez - 2019-06-30
    • Description has changed:

    Diff:

    --- old
+++ new
@@ -1,9 +1,9 @@
-Parsing ":not(FOO)" results in a PseudoClassElementImpl with value "not(FOO)", i.e. the argument itself is not parsed any further. This is both in SACParserCSS21 and SACParserCSS3.
+Parsing ":not(FOO)" or ":nth-child(FOO)" results in a PseudoClassConditionImpl with value "not(FOO)" or "nth-child(FOO)", i.e. the argument itself is not parsed any further. This is in SACParserCSS3.

-This makes it impossible to differentiate between ":not(#foo)" and ":not(.foo)" without putting the argument string through second parser, i.e. the tree is incomplete.
+This makes it impossible to differentiate between ":not(#foo)" and ":not(.foo)" without putting the argument string through second parser, i.e. the tree is incomplete. I found this issue when calculating CSS selector specificity using the the current parsers: ":not(#foo)" and ":not(.foo)" return the same single SAC_PSEUDO_CLASS_CONDITION/PseudoClassConditionImpl, although their specifity is different (1-0-0 vs 0-1-0) and there could be more selectors inside.

-I found this issue when calculating CSS selector specificity using the the current parsers: ":not(#foo)" and ":not(.foo)" return the same single SAC_PSEUDO_CLASS_CONDITION/PseudoClassElementImpl, although their specifity is different (1-0-0 vs 0-1-0) and there could be more selectors inside.
+I guess the issue only is that CSS3 parsing/grammar is incomplete. I'll see if I can attach a patch, but it may take a while.

-See here for more about the negation pseudo-class: https://www.w3.org/TR/selectors-3/#negation
-
-This isn't critical, but the parser is still incorrect.
+See here for more:
+- https://www.w3.org/TR/selectors-3/#negation
+- https://www.w3.org/TR/selectors-3/#nth-child-pseudo
     

  • Jeremie Miserez

    Jeremie Miserez - 2019-06-30
    • summary: Parsing of :not(X) argument incomplete --> CSS3 parsing of :not(X) , :nth-child(X) arguments incomplete
     

  • Jeremie Miserez

    Jeremie Miserez - 2019-06-30

    Updated the issue to better describe the problem.

     

Anonymous
Anonymous

Add attachments
Cancel