Skip to content

markdown parser fixes - #105

Open
nobodywasishere wants to merge 18 commits into
icyleaf:masterfrom
nobodywasishere:nobody/markdown-parser-fixes
Open

nobodywasishere wants to merge 18 commits into
icyleaf:masterfrom
nobodywasishere:nobody/markdown-parser-fixes

Conversation

@nobodywasishere

Copy link
Copy Markdown
Collaborator

No description provided.

Treat CRLF as one line ending and a standalone CR as a separate line
ending so mixed input preserves blank lines.

Assisted-by: GPT-6.2 Sol
Use the parser's actual line counter for source positions, including
input that ends in a newline.

Assisted-by: GPT-6.2 Sol
Keep an unmatched link or image opener from claiming text in a later
paragraph.

Assisted-by: GPT-6.2 Sol
Match the entity syntax before decoding so a stray ampersand cannot
raise an exception or consume later inline markup.

Assisted-by: GPT-6.2 Sol
Use U+10FFFF as the upper bound so valid supplementary code points
render and invalid code points still become the replacement character.

Assisted-by: GPT-6.2 Sol
Keep escapes such as %2F intact so a link destination retains its
meaning. Render parentheses as required by the existing double-encoding
fixture and continue to decode Markdown entities once.

Assisted-by: GPT-6.2 Sol
A full reference link can use a label containing a single character.

Assisted-by: GPT-6.2 Sol
Collapse runs of spaces and tabs in references and accept tabs where
reference parsing expects horizontal whitespace.

Assisted-by: GPT-6.2 Sol
Check whitespace on the correct side of each delimiter and classify
Unicode separator and symbol characters according to CommonMark.

Assisted-by: GPT-6.2 Sol
Allow the heading rule to consume an underline inside the container
while keeping an unquoted thematic break outside the quote.

Assisted-by: GPT-6.2 Sol
A pipe in a valid heading must not cause the heading to be consumed
as another table row.

Assisted-by: GPT-6.2 Sol
Advance offsets by bytes without leaving the cursor inside a UTF-8
character before a regex match.

Assisted-by: GPT-6.2 Sol
Rewind over trimmed punctuation instead of copying the remaining
paragraph into the current node. Keep scheme detection and URL matching
on the same byte offsets.

Assisted-by: GPT-6.2 Sol
Match the cmark-gfm fixture for a single-label hostname in a URL
with an explicit scheme.

Assisted-by: GPT-6.2 Sol
Leave a prefix embedded in a word as literal text while allowing the
following address, including one-character local parts, to autolink.

Assisted-by: GPT-6.2 Sol
A pending example can still carry extension tags such as autolink
and footnotes. Parse the full tag list.

Assisted-by: GPT-6.2 Sol
Restore the original CR CR LF input for the Windows line-ending
regression and pass those bytes unchanged to the parser.

Assisted-by: GPT-6.2 Sol
Give explicit links precedence over extended autolinks in their labels.
Keep autolinks in ordinary bracketed text and leave closing brackets for
the inline parser. An angle-bracket autolink still disables enclosing
link openers.

Assisted-by: GPT-6.2 Sol
@nobodywasishere
nobodywasishere force-pushed the nobody/markdown-parser-fixes branch from 6412d5c to 3db59c3 Compare November 19, 2026 10:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant