Is it still feasible to direct an internal link to a page that has not been created yet, without specifying the page number ?

[nsimonovici]

Hi there,

As I switch from fpdf to fpdf2, I notice a change in behaviour in my internal links.

I was using the "add_link" to set up internal links, making the cell clickable, while saving the created link (result out of "add_link") in a list.

As my layout can move depending of pictures or texts sizes, I did not want to hardcode the page to where links are directing.
That's why I would trigger the "set_link" method of desired link in the list at the moment of my choice when building my PDF file, directing automatically the priorly created link in cell to a page created afterwards.

But that was before. Now I can read from the latest release that :

And my "trick" does not work anymore, sadly.

Therefore is there a retro compatibility way to achieve what was working perfectly in fpdf ? Maybe downgrading my package might by a solution. EDIT : it is not (tried version 2.6.0)

Thanks for any kind of assistance

[Lucas-C]

Hi!

Thank you for reaching out @nsimonovici and welcome to the fpdf2 community 😊

What you describe is very strange and may be a bug...
FPDF.add_link() in the latest releases should be fully backward compatible with the earlier versions of fpdf2 and even PyFPDF.

We only added some extra features to it in v2.6.1.
Calling set_link() is simply now optional: whereas you previously had to call link = pdf.add_link() and pdf.set_link(link, page=2), you can now just make a single call link = pdf.add_link(page=2), if you know the target page number upfront.
But calling set_link() is still valid.

We have a set of unit tests regarding add_link() that ensure the same features still work as they used to:
https://github.com/PyFPDF/fpdf2/blob/master/test/test_links.py
And those tests are all passing.

Could you please provide some minimal Python code reproducing your issue?
If you don't know how to build a minimal reproducible example, check this tutorial: https://stackoverflow.com/help/minimal-reproducible-example

[Lucas-C]

A fix is coming in this PR: #730

You can test it using pip:

pip install git+https://github.com/PyFPDF/fpdf2.git@master

[nsimonovici]

I can confirm it is now working as expected ! Kudos !

Any idea when would that be published in an official version ?

[Lucas-C]

Any idea when would that be published in an official version ?

I'm planning on releasing a new version next week.

[Lucas-C]

A fix has been released in v2.7.0: https://github.com/PyFPDF/fpdf2/releases/tag/2.7.0

[nsimonovici]

Thanks Lucas you're amazing, that was handled smoothly.

All the best !

[osresearch]

I don't know if this is a regression or expected behaviour, but I ran into a problem upgrading to fpdf2 with the same error. Pre-building all of the links by calling add_link before any calls to add_page results in a failure. The test case in https://github.com/PyFPDF/fpdf2/blob/master/test/test_links.py adds the page before generating the forward link to section 1, so it doesn't trigger it this behaviour, but this minor tweak to the test fails:

from fpdf import FPDF
pdf = FPDF()
pdf.set_font("helvetica")

link_to_section1 = pdf.add_link()  # *before* add_page

pdf.add_page()  # page 1
pdf.cell(text="Section 1", link=link_to_section1)

pdf.add_page()  # page 2
pdf.set_link(link_to_section1, page=pdf.page)
pdf.cell(text="Section 1: Bla bla bla")

results in:

ValueError: Cannot insert link 1 with no page number assigned

[Lucas-C]

Hi @nsimonovici 🙂

Given that there is an explicit error explaining the problem, yes this is the expected behaviour.

If you have a specific use case in mind where you would like to create links BEFORE calling .add_page(),
and you can describe it to us and explain why calling .add_page() before is not possible,
we could consider supporting this use case in fpdf2 🙂

[andersonhc]

add_link() creates an initial object pointing to the current page - 0 in your case - and using the link in cell() without a valid page is raising the exception.

To make it work you can:

• Create the links after the first call to add_page()
• Create all the links pointing to the page 1 with link_to_section1 = pdf.add_link(page=1) - It will initially point to page 1 but update when you call set_link()

I will wait for @Lucas-C input on how we can improve this behavior on fpdf2. I don't think only removing the validation on cell() is a good idea because we can end up with an invalid document if you never call set_link() to point to a valid page.
I believe the best approach is to change add_link() and default to 1 if self.page = 0 since a page 1 is mandatory and will be automatically added at output() if it hasn't been done before.

[osresearch]

In the pdfcal code it was not a problem to move the link pre-generation for everyday of the year until the title page was generated, although it used to work in fpdf 1, which is why I was wondering if it was a regression.

Perhaps a way to make the error clearer is to have pdf.add_link() throw an error if it is called too soon without an explicit page number; having the error occur when the link is referenced happens quite a bit later rather than on the line that actually needs to be fixed.

[Lucas-C]

I believe the best approach is to change add_link() and default to 1 if self.page = 0 since a page 1 is mandatory and will be automatically added at output() if it hasn't been done before.

If we change the default value, we won't be able to distinguish the case where an explicit value is provided, and when we should use the current page number. Or did I miss something?

Perhaps a way to make the error clearer is to have pdf.add_link() throw an error if it is called too soon without an explicit page number; having the error occur when the link is referenced happens quite a bit later rather than on the line that actually needs to be fixed.

Sounds like a good idea 🙂
Would you like to submit a Pull Request?

[osresearch]

I agree with @andersonhc suggestion of setting the default to page 1 (iff no pages have been added yet) rather than throwing an error. I've made the change in #1337