Support relative file URIs in hyperlinks when using CLI

[noverby]

For example when converting the following html to a pdf, it should be possible to make the hyperlink refer to the relative URI and not the absolute URI:

<a href="external-file.pdf">File</a>

What I get: "file://absolute-path/external-file.pdf"
What I want: "external-file.pdf"

I have tried to change the --base-url argument, but it does not remove the absolute path if I set it to ""(empty).

So I looked it up in the code, and it is fairly easy to change, but I just wanted to make sure that there is interest in including this feature. If there are indeed interest I can make a pull request.

Relative file URIs are also what Prince outputs by default.

[liZe]

Hmmm… Really looks like #437, I'll take a look.

[liZe]

Unfortunately, what you want is currently impossible using the CLI:

• HTML finds base_url from the document when using a file,
• the CLI uses '.' instead of None as default base_url value when using stdin, and
• base_url is transformed into an absolute path when it's different from None (even if it's '').

Even if these 3 choices may look strange when looked at carefully, I think that they are actually really good when they're used without thinking (and that's what most CLI users do). I don't want to change this behavior.

So you currently have to use HTML(string='<a href="external-file.pdf">File</a>') to get what you want.

A solution would be to transform --base-url="" into base_url = None, and keep base_url = '.' when --base-url is not set. You would then be able to get relative URIs using weasyprint --base-url="" - output.pdf < input.html. What do you think about that?

[noverby]

@liZe That would be nice. Thank you 👍

[W4RH4WK]

It seems like this does not work anymore.

I reference another PDF via <a href="books/anarchy.pdf"> and even when calling weasyprint with --base-url="" it still writes an absolute path to the PDF.

[liZe]

It seems like this does not work anymore.

It seems to work for me, when using both --base-url="", and - output.pdf < input.html to give no filename hint.

[W4RH4WK]

I see, providing the input via stdin is relevant for this.

With this, the filepath indeed remains relative. However, style sheets are no longer found:

ERROR: Relative URI reference without a base URI: <link href="normalize.min.css">
ERROR: Relative URI reference without a base URI: <link href="style.css">

[liZe]

However, style sheets are no longer found:

Of course, they’re not found because there’s no base-url.

Handling your use case is not as easy as it looks. Even if we add a new option (something like keep-local-anchors), I think that it will be really hard to find a behavior that satisfies everyone.

If you want to get this feature, you can open a new issue so that we can discuss your use case and find a way to solve it in a quite flexible way. I can’t promise that it will be implemented, as the current behavior is already quite complex (even if it’s easy to use in easy cases, as explained above).

[W4RH4WK]

Alright. This is currently not high on my priority list. When I come around to this, I might checkout the code and see whether I can come up with something that is viable. I'll open an issue / PR as necessary.

Thank you for the information.