Inserting a pdf into my document (again, sorry)

Questions
,
1

Using Quarto/typst

In all my reports, I have to insert separate pdf files; preferably, keeping page numbering correct, but that’s a secondary issue that I know is problematic. Since some of the other threads drifted off, I start a new one here in the hope to get a state-of-the-art solution.

Threads

I have read the relevant threads:

and the use introduction of “eval” instead of query.

Insert as image

I currently use the following workaround

#let insert-pdf(document-name, pages) = context {
  for i in range(1, pages+1) {
    move(
      image(
        document-name, 
        page: i, 
        width: page.width,
        height: page.height
      ),
      dx: -here().position().x,
      dy: -here().position().y,
    )
  }
}

but this uses images and my client does not like these non-searchable images.

Using pdftk

The alternative is to use pdftk, with a line in .ps1 like the following:

pdftk A=$inp B=$merge_with cat A1-$page B A$page1-end output $out

This works, but I have to find know $page in advance. I have a label into the document:

#label("fachinformation_pdf")

but I could not find the page of the label by program.

Finding label page

Claude suggests using

typst eval "query(<fachinformation>).first().location().page()" --in index.typ

This give a strange error message:

typst eval "query(<fachinformation>).first().location().page()" --in index.typ
error: array is empty
  ┌─ <input-expression>:1:0
  │
1 │ query(<fachinformation>).first().location().page()
  │ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

warning: unknown font family: font awesome 6 free
   ┌─ @preview/fontawesome:0.5.0/lib-impl.typ:51:10
   │
51 │     font: default_fonts, // If you see warning here, please check whether the FA font is installed
   │           ^^^^^^^^^^^^^

Note: the report builds without problems

Questions

  1. Any better alternatives to insert a pdf as of today?
  2. How to find the location of a label with eval?
  3. Secondary (maybe impossible): Get page numbering right? Note that the pdf is inserted into the middle of the main document.
1 Like
2

Inserting a PDF as an image results in selectable and searchable text – at least in my PDF viewers: VS Code (Chromium), Firefox and Okular all behave the same in this regard. Are there other PDF viewers in which the inserted PDF is not selectable/searchable? That would probably be considered a bug (either in that viewer or Typst, depending on who is right about the PDF contents).

I think the following insert-pdf is cleaner though, and doesn’t require context:

#let insert-pdf(document-name, pages) = {
  for i in range(1, pages+1) {
    page(margin: 0pt, image(
      document-name,
      page: i,
      width: 100%,
      height: 100%,
    ))
  }
}

Since your goal is to fill the whole page, calling page() explicitly works well here. If you want to potentially preserve the pages’ sizes, you could also use this instead:

#let insert-pdf(document-name, pages) = {
  for i in range(1, pages+1) {
    page(width: auto, height: auto, margin: 0pt, image(
      document-name,
      page: i,
    ))
  }
}

Querying a label does not really find labels, it finds elements tagged with labels. You may want to write #metadata(none) #label("fachinformation_pdf"), or equally #metadata(none) <fachinformation_pdf> instead, since if you attach a label to a long paragraph, the location (and page) may be earlier than you expect.

It wrote fachinformation instead of fachinformation_pdf, otherwise the query looks right :slight_smile:

2 Likes
3

It wrote fachinformation instead of fachinformation_pdf , otherwise the query looks right

Krrr… That the curse of AI, when you don’t check it. Could have happened to a human, though.

Inserting a PDF as an image results in selectable and searchable text –

Does not work for me. And the text is semi-good:

Img_00sv
Img_00sv656×347 28.6 KB

Check with pdf-Exchange. Search for “blassgelbes” does not return anything

4

hmm… maybe it has something to do with the source PDF, does it contain OCR text? If it is not confidential, it would be cool if you could share a sample of a PDF that inhibits this problem (here or in a Github issue). Embedding Typst-produced PDFs using image() definitely doesn’t have either the rendering quality or selection issues.

5

And again, mea culpa, but a thing I learned: the example really was a scan, e.g. from
AIPS - Einzelabfrage, try “Addaven” (not my example)

On my computer, I have two drivers for pdf: “Save as pdf” gives nice native pdf, while the pdfXChange produces the 2.5x larger scans. If at all, I would have expected the opposite.

I have not tried how this works out with “image”, but I am sure you are right.

Sorry for the noise. Hopefully, you “image” example will help others too.

1 Like
6

Implemented and works great. My variant below

#let insert-pdf(document-name, pages, scale: 100%, margin: 10pt) = {
  for i in range(1, pages+1) {
    page(margin: margin, image(
      document-name,
      page: i,
      width: scale,
      height: scale,
    ))
  }
}
1 Like