em-dash cannot be escaped

[tusooa]

downstream: https://git.pleroma.social/pleroma/pleroma/-/issues/2810

When smartypants is on, \-\- is converted to –. This behaviour sounds problematic, because it is not two consecutive dashes.

iex(3)> Earmark.as_html("hey @yyy@xn\\-\\-i2raa\\.com")
{:ok, "<p>\nhey @yyy@xn–i2raa.com</p>\n", []}

This causes problems for our workflow, because it represents a domain name and thus should not be converted.

[RobertDober]

Ty for this report, unfortunately the smarty pants option is deprecated, which version of Earmark do you use?

[tusooa]

earmark 1.4.22 and earmark_parser 1.4.32

[RobertDober]

I see what you mean but why should we not create a dash?

In links we do not

Earmark.as_html("hey @<yyy@xn\\-\\-i2raa\\.com>")
{:ok,
 "<p>\nhey @<a href=\"mailto:yyy@xn%5C-%5C-i2raa%5C.com\">yyy@xn\\-\\-i2raa\\.com</a></p>\n",
 []}

but switching smartypants off is what you need outside

iex(1)> Earmark.as_html("hello--world")
{:ok, "<p>\nhello–world</p>\n", []}
iex(2)> Earmark.as_html("hello--world", smartypants: false)
{:ok, "<p>\nhello--world</p>\n", []}
iex(3)>

[tusooa]

I see what you mean but why should we not create a dash?

In links we do not

Earmark.as_html("hey @<yyy@xn\\-\\-i2raa\\.com>")
{:ok,
 "<p>\nhey @<a href=\"mailto:yyy@xn%5C-%5C-i2raa%5C.com\">yyy@xn\\-\\-i2raa\\.com</a></p>\n",
 []}

but switching smartypants off is what you need outside

iex(1)> Earmark.as_html("hello--world")
{:ok, "<p>\nhello–world</p>\n", []}
iex(2)> Earmark.as_html("hello--world", smartypants: false)
{:ok, "<p>\nhello--world</p>\n", []}
iex(3)>

@yyy@xn\\-\\-i2raa\\.com is a fediverse user handle.

Our workflow in Pleroma is (1) escape special characters in the user handle, (2) pass it to earmark, and (3) convert the handle to profile links.

And even if it is handled as a link it is still buggy: the text of the link should also not have the em-dash -- it should have the double ascii dashes instead.

iex(1)> Earmark.as_html("hey @<yyy@xn--i2raa.com>")      
{:ok,
 "<p>\nhey @<a href=\"mailto:yyy@xn--i2raa.com\">yyy@xn–i2raa.com</a></p>\n",
 []}

[RobertDober]

And even if it is handled as a link it is still buggy: the text of the link should also not have the em-dash -- it should have the double ascii dashes instead.

Matter of taste I would say, but if your desired behavior would solve your problem I would defenitely consider accepting a PR.

That said, is using smartypants: false not fixing your issue?

[tusooa]

Switching off smartypants fixes the issue, but as Pleroma is a piece of software that advocates for customizability, it is apparently better if this option is customizable without breaking anything.