automatic title #86

New Issue

gogoigo · 2020-03-11T07:17:06Z

gogoigo commented

2020-03-11 07:17:06 +00:00

Feature: After add a link, it could check the website and autocomplete title input. User could confirm or edit this text.
I could develop this feature.

Feature: After add a link, it could check the website and autocomplete title input. User could confirm or edit this text. I could develop this feature.

cmccabe commented

2020-03-11 10:56:12 +00:00

I like this idea. Do you think this would require pulling in a large amount of library code to handle, or could we write it ourselves easily?

There is beautifulsoup, of course, but that's a pretty hefty lift if we're just grabbing page title.

Here's a SO answer about that, for posterity reference: https://stackoverflow.com/questions/35956045/extract-title-with-beautifulsoup

I like this idea. Do you think this would require pulling in a large amount of library code to handle, or could we write it ourselves easily? There is beautifulsoup, of course, but that's a pretty hefty lift if we're just grabbing page title. Here's a SO answer about that, for posterity reference: https://stackoverflow.com/questions/35956045/extract-title-with-beautifulsoup

sloum commented

2020-03-11 17:32:05 +00:00

This could be error prone. A lot of users in spaces where linkulator is deployed use gopher/gemini/etc. Those protocols do not use html and do not have a concept of a <title> in the html sense. So that means we now need to parse a url, add branching, and then extract a title where applicable. Then we need to use readline to inject that title (or an empty string) into an editable field.

I do believe the majority of links are indeed http links, so maybe it is not worth prioritizing the other protocols... but I know they were a consideration at other stages of the project.

I can only speak for myself here, but I have never named my link the name of the article. I have not wanted to, as I have found my descriptioins to be ebtter for the purposes of linkulation. I would definitely want the ability to NOT have the link named after the article (which was included in the above recommendation) and I would prefer to not perform an HTTP request when adding a link (which was not part of the above recommendation).

Part of the pubnix vibe (and gopher as well) is that no secret requests are occuring. I think this would need to be handled in a very transparent way so that users know they are sending out a request... maybe as an opt-in that users can add in their settings? Maybe as a flag?

This could be error prone. A lot of users in spaces where linkulator is deployed use gopher/gemini/etc. Those protocols do not use html and do not have a concept of a `<title>` in the html sense. So that means we now need to parse a url, add branching, and then extract a title where applicable. Then we need to use readline to inject that title (or an empty string) into an editable field. I do believe the majority of links are indeed http links, so maybe it is not worth prioritizing the other protocols... but I know they were a consideration at other stages of the project. I can only speak for myself here, but I have never named my link the name of the article. I have not wanted to, as I have found my descriptioins to be ebtter for the purposes of linkulation. I would definitely want the ability to NOT have the link named after the article (which was included in the above recommendation) and I would prefer to not perform an HTTP request when adding a link (which was not part of the above recommendation). Part of the pubnix vibe (and gopher as well) is that no secret requests are occuring. I think this would need to be handled in a very transparent way so that users know they are sending out a request... maybe as an opt-in that users can add in their settings? Maybe as a flag?

sloum added the

enhancement

label 2020-03-11 17:55:45 +00:00

asdf commented

2020-03-12 02:49:02 +00:00

Thanks for the suggestion @gogoigo :)

@sloum makes some really good points. I think the proposed function could be useful to some people, but should not do automatic lookups, only running when specifically requested. This would mean further complexity, which might be challenging to balance against the gain in functionality.

It would be a good exercise for fun and learning though, and would not want to discourage you if you are keen to try it out. If you have more details to share about the implementation, or even a prototype, it would be useful to see how these different needs could be balanced.

Thanks for the suggestion @gogoigo :) @sloum makes some really good points. I think the proposed function could be useful to some people, but should not do automatic lookups, only running when specifically requested. This would mean further complexity, which might be challenging to balance against the gain in functionality. It would be a good exercise for fun and learning though, and would not want to discourage you if you are keen to try it out. If you have more details to share about the implementation, or even a prototype, it would be useful to see how these different needs could be balanced.

asdf added the

help wanted

label 2021-08-19 10:18:54 +00:00

Sign in to join this conversation.