PDF watermark from Reader

[vsenko]

Looks like currently it's impossible to create PDF watermarks via Reader (or byte slice, etc.), only from file:

pdfcpu/pkg/pdfcpu/stamp.go

Line 1192 in e89570a

otherCtx, err := ReadFile(wm.FileName, NewDefaultConfiguration())

It would be really handy to add such a feature. Furthermore it's already there for Image watermarks:

pdfcpu/pkg/pdfcpu/stamp.go

Line 1224 in e89570a

wm.image, wm.width, wm.height, err = CreateImageResource(ctx.XRefTable, wm.Image, false, false)

[hhrutter]

Hello!

I am not quite sure what you are looking for.
Please explain your specific usecase.

In case you are looking for something similar to:
api.ImageWatermarkForReader(r io.Reader, desc string, onTop, update bool, u pdfcpu.DisplayUnit)

maybe:
api.PDFWatermarkForReader(r io.Reader, desc string, onTop, update bool, u pdfcpu.DisplayUnit)

then that can't be done.

A PDF resource for stamping/watermarking is a combination of some content stream bytes, its resources and a bounding box whereas an Image resource can be represented by a simple byte slice.

On top of that pdfcpu supports multi-stamps where each page of the input PDF gets stamped with a different page of the PDF stamp source file. See more here: https://pdfcpu.io/core/stamp

[vsenko]

Excuse me for being that brief, consider the following usecase: the watermark is generated on-the-fly for each incoming PDF from some metadata, thus it's inconvenient to save the watermark beforehand to the file. The code that generates the watermark is on another process/host.

Thank you for identifying the problem with io.Reader approach, but maybe we can go with io.ReadSeeker:

api.PDFWatermarkForReader(r io.ReadSeeker, desc string, onTop, update bool, u pdfcpu.DisplayUnit)

[vsenko]

Another option (to keep the interface more consistent) is to copy data from io.Reader into io.ReadSeeker under the hood, but it will require twice more memory during the copy.

[hhrutter]

Generating a pdfcpu PDF watermark somewhere in order to use it later on for stamping is a valid use case but can't be done easily.

Like already expained above, a PDF watermark is a non trivial object consisting of some content bytes, its resources and some metadata. Sure, we could write out this object/blob as a file, but this is not practical because of its required resources, especially when some embedded user fonts are involved. A much better alternative is to to render the PDF watermark to an image but that's not an option right now.

Switching from io.Reader to an io.ReadSeeker doesn't change anything AFAIC.

I need to think more about this because I like the use case.

[hhrutter]

This is part of https://github.com/pdfcpu/pdfcpu/releases/tag/v0.6.0