Improve decoding of HTML entities #5064
Conversation
src/html.cr
Outdated
|
|
||
| # see https://html.spec.whatwg.org/multipage/parsing.html#numeric-character-reference-end-state | ||
| private def self.decode_codepoint(codepoint) | ||
| if 0x80 <= codepoint <= 0x9F |
There was a problem hiding this comment.
case switch might look neater here.
src/html.cr
Outdated
| "\uFFFD" | ||
| elsif 0xFDD0 <= codepoint <= 0xFDEF || # unicode noncharacters | ||
| codepoint & 0xFFFF >= 0xFFFE || # last two of each plane (nonchars) disallowed | ||
| (codepoint < 0x0020 && codepoint != 0x0009 && codepoint != 0x000A && codepoint != 0x000C) || # unicode control characters |
There was a problem hiding this comment.
codepoint != 0x0009 && codepoint != 0x000A && codepoint != 0x000C->
!{0x0009, 0x000A, 0x000C}.includes?(codepoint)
spec/std/html_spec.cr
Outdated
| end | ||
|
|
||
| it "unescapes characters above Char::MAX_CODEPOINT" do | ||
| str = HTML.unescape("limit �") |
There was a problem hiding this comment.
I don't get it, if 10FFFF is not a character, then why should something right after it be a character
There was a problem hiding this comment.
That's a typo. It doesn't decode � but replaces it with replacement character \uFFFD.
There was a problem hiding this comment.
Question still stands. -> but � -> \uFFFD - why?
There was a problem hiding this comment.
 is a valid codepoint, yet not supported as a character reference because it's a unicode noncharacter. Therefore it is not decoded.
� is an invalid codepoint and therefore replaced by replacement character.
See Numeric character reference end state for details.
straight-shoota
force-pushed
the
jm-html-entities
branch
from
961dcd9
to
39244b4
Compare
This PR adds a few fixes for several edge cases that were broken from #5055 or before
0x80and0x9Fare replaced with compatibility replacements for old numeric entities from Windows-1252The code is based on the HTML5 tokenizing guidelines and inspired by the implementations for Go and PHP.