Add option to warn about absolute internal URLs ?

[lurch]

I dunno if this is a feature that html-proofer already has (?) but if not, I think this would be a nice option to add.

In raspberrypi/documentation#669 a user found a link in https://www.raspberrypi.org/documentation/configuration/config-txt/video.md which didn't work, but this wasn't picked up by html-proofer.

In the source markdown (scroll down to the dpi_group, dpi_mode, dpi_output_format section), the link looks like [here](/hardware/raspberrypi/dpi/README.md) - note that it starts with a leading slash. This link works fine when viewing the page on Github (which links it to https://github.com/raspberrypi/documentation/blob/master/hardware/raspberrypi/dpi/README.md) and would also work fine if the HTML generated from this was being deployed to the root of a domain; but in our particular case the HTML generated from this is deployed to a subdirectory https://www.raspberrypi.org/documentation/ which means the absolute-link ends up linking to https://www.raspberrypi.org/hardware/raspberrypi/dpi/README.md which 404s. (the correct link would be https://www.raspberrypi.org/documentation/hardware/raspberrypi/dpi/README.md)

To catch situations like this (where the generated URL isn't being deployed to the root of a domain) it would be nice if html-proofer had an option (which should probably be disabled by default?) to warn if an internal URL starts with a leading slash.

[lurch]

raspberrypi/documentation#670 fixes the broken link. But it would still be nice if html-proofer could have caught it :-)

[fulldecent]

Thanks for sharing. In general, relative URLs that start with a slash are neither discouraged nor documented as deviation from best practice.

It is possible for html-proofer to make such a check. But adding such an option would not be consistent with our project scope.

We encourage you to implement this, if you are interested. Please see our documentation at https://github.com/gjtorikian/html-proofer/wiki/How-to-make-a-custom-class and we will be happy to list it on our list of extensions at https://github.com/gjtorikian/html-proofer/wiki/Extensions-(custom-classes). It's really simple. You can copy the code from https://github.com/fulldecent/html-proofer-mailto_awesome as a starting point and the main code you want is probably one line of code in this file: https://github.com/fulldecent/html-proofer-mailto_awesome/blob/master/lib/mail_to_awesome.rb

For now I will close this issue as out of scope for this project.

[gjtorikian]

As well, for something like this, you may be able to use the url_swap option, which, if I'm understanding correctly, might look like this:

url_swap: { "^/hardware/": "/documentation/hardware/" }

That is: any link starting with /hardware should be prepended with /documentation first.

[lurch]

In general, relative URLs that start with a slash are neither discouraged nor documented as deviation from best practice.

Yeah, that's why I suggested it should be an option that's disabled by default, unless explicitly enabled ;-)

But adding such an option would not be consistent with our project scope.

Oh, I would have thought it comes under the "if your internal links are working" section? :-/
For situations where the pages aren't being hosted in the root of a domain, I'd consider checking for not-absolute-internal-links as more important than checking-internal-hash-references *shrug*

We encourage you to implement this, if you are interested.

I'm afraid I've never written any Ruby, so I wouldn't even know where to start :-(
Rather than closing this, perhaps it could be left open as a "feature request" in case somebody who does know Ruby feels like implementing it?

That is: any link starting with /hardware should be prepended with /documentation first.

Nah, that wouldn't work in this scenario. For the repo in question we need all links to be relative, because the documentation can be browsed from both https://github.com/raspberrypi/documentation and from https://www.raspberrypi.org/documentation/

[fulldecent]

Does this solve the problem?

url_swap: { "^/": "http://NONEXISTANTDOMAIN.COM/" }