Scope wiki linkification characters using POSIX words

[hlfan]

Fixes the Rinku-independent part of #6611

[pablobm]

Hm, I wish we tested those too. I understand we present them as a configuration detail, but as they get more complex their logic is going to be forgotten.

[tomhughes]

There's no reason we can't test the behaviour of the default though?

We do have a load of tests in test/lib/rich_text.rb though most of them override the actual linkification rules though I don't really know why as testing with the default seems at least as useful and probably more so.

[pablobm]

Haven't tried, but by looking at the code I think if we don't call with_settings (the test helper) then the defaults will apply, making them testable 👍

[hlfan]

The default test settings are a bit different from the default settings:

openstreetmap-website/test/lib/rich_text_test.rb

Line 398 in a30c8d0

with_settings(:server_url => "www.openstreetmap.org", :server_protocol => "https") do

That's what I needed to override to paste in the cases from #5780.

[pablobm]

@hlfan - I think we shouldn't try to fix #6616 (mentions with spaces) with a regexp or similar. I think we should implement a selector of the sort you see in GitHub or Slack, where you get a dropdown the moment you enter an "@" character, and then this gets stored as @[user_id:$USER_ID] (or something to that effect) that is interpreted and replaced as appropriate when renderding. This would also fix the issue of users changing usernames and leaving old usernames behind in comments, sometimes maliciously.

[hlfan]

Do you wanna make a separate issue for the user dropdown?

[pablobm]

Do you wanna make a separate issue for the user dropdown?

Probably best, yes. Thank you.

[hlfan]

@tomhughes \p{Word} is causing timeouts... is there a more performant option?

[pablobm]

Not sure it's enough for a timeout, but I can confirm that in the key=value patterns, the \p{Word} ones are running about 10x slower than similar \w ones (just as a comparison, I know they are not the same). These are some numbers, taken from running test/models/diary_entry_test.rb in my computer:

 11us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\w]+)=(?<value>[-+,.:;'%()\/\w]+)/i
116us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\p{Word}]+)=(?<value>[-+,.:;'%()\/\p{Word}]+)/i

 11us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\w]+)=(?<value>[-+,.:;'%()\/\w]+)/i
116us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\p{Word}]+)=(?<value>[-+,.:;'%()\/\p{Word}]+)/i

 24us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\w]+)=(?<value>[-+,.:;'%()\/\w]+)/i
236us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\p{Word}]+)=(?<value>[-+,.:;'%()\/\p{Word}]+)/i

 32us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\w]+)=(?<value>[-+,.:;'%()\/\w]+)/i
302us /(?<=^|[^\w!#$%&'*+,.\/:;=?@_~^\-])(?<key>[-+,.:;'%()\p{Word}]+)=(?<value>[-+,.:;'%()\/\p{Word}]+)/i

[pablobm]

Do you wanna make a separate issue for the user dropdown?

@hlfan - Sorry, for a moment I thought you were saying you were going to create the issue yourself. Now I re-read the comment and realised you hadn't. Done now! #6635

[tomhughes]

The default timeout is 1s which seems a world away from the times @pablobm is reporting unless the github actions runners are really slow...