Closed maxbrunet closed 8 months ago
Hello @maxbrunet 👋
Maybe the binding should be dropped entirely
Hello @gpakosz 👋
It is true that I don't use it so much, so I likely would not miss it. I only use it every once in while to find a new PR link for a branch pushed recently
I suppose one could configure a binding in the local config or use a plugin if they would like the feature back
Maybe if there's an Homebrew formula for urlscan
I might keep it
Or I could support both, giving preference to urlscan
A few Linux distros have a package for it: https://repology.org/project/urlscan/versions
But I cannot find it in Homebrew. On macOS, it would be Nix, MacPort, or pipx
at the moment
Can you please try the gh-695
branch?
Thank you, I tested it and this technically looks good to me, however now that I see urlscan
in action with the config, I am not convinced by the default experience, the output is very verbose and feels barely more useful than scrolling up the shell output in my opinion (pressing c
helps to get a compact view, but the /
binding closes it instead of searching) (it also matches my username@machine
prompt as an email address)
One might want to use the --compact
, --dedupe
, and --regex 'https?://.+'
flags, but not sure if everyone would agree if the config hardcodes them. I wonder if adding an option to choose the command would be possible with the default still being urlview
(that would make experimentation easy at least)
And if that's adding too much complexity, dropping the feature is still on the table imo (personally I do not depend on it heavily in my day to day)
Updated the gh-695
branch with the options you suggest
Thank you for the update, it works and looks very close to what we have with urlview
.
~--dedupe
is inefficient for me, because my right prompt contains a timestamp and that makes each URL "unique", that would probably be an issue to discuss upstream~
If you decide to ship this change, I would update the README to mention urlscan
as well:
https://github.com/gpakosz/.tmux/blob/5354501a5bff69457b37132c1959e9063b1c1714/README.md#L134
Thanks a lot for your time and responsiveness! :heart:
Edit: I actually took the example regex from the man page, but there's a formatting problem when it is rendered, it is --regex 'https?://.+\.\w'
, that way the regex stops matching when it encounters white-spaces
But that regex is not perfect either, we can do without --regex
for simplicity's sake, I think I would be ok if my prompt appears once in the result :sweat_smile:
I think I would be ok if a my prompt appears once in the result 😅
What prompt are we talking about?
$PROMPT
, urlscan
tries to be smart and guesses user@host
in the prompt as an email address even though it is not a mailto:
URL
But --regex 'https?://.+\.\w'
solves this right?
No, unfortunately this regex is too simplistic and does not match all possible URLs (e.g. http://localhost:3000
, and I tried to remove \.
or change \.\w
, but it gets truly complex to have an accurate regex)
This is the default in urlview
:
(((http|https|ftp|gopher)|mailto):(//)?[ˆ <>"\t]*|(www|ftp)[0-9]?\.[-a-z0-9.]+)[ˆ .,;\t\n\r<">\):]?[ˆ, <>"\t]*[ˆ .,;\t\n\r<">\):]
And this is urlscan
: firecat53/urlscan@1.0.1
/urlscan/urlscan.py#L248-L273
I think it is simpler to do without --regex
and use urlscan --compact --dedupe
only
Or I removed the use of --regex
Or I removed the use of
--regex
Yes, that's what I suggest :)
I meant "ok I removed ..." 😅 Damn autocarrot! 🥕
It seems
urlview
is no longer maintained^orphan. Packaging systems will start dropping it (e.g. NixOS/nixpkgs#274280) as issues will accumulate with no hope for them of being fixed upstream.Some alternatives:
I have only tried
urlscan
so far, and it looks like a good drop-in replacement.