Reading the PDF metadata on large files

[drusellers]

Is there a way to get the PDF attributes like Creation Date without loading the whole file into memory? We have large multi-gigabyte files and are trying to process them on a fleet of smaller ECS instances. The streaming features of pypdf have been great, but would love to get the same metadata without loading the whole file into memory. Happy to read the whole file in chunks.

[pubpub-zz]

Have you tried to pass a buffered file stream as stream argument ?

[drusellers]

I don't think we have. TY I'll give it a shot. We are reading from an HTTP stream - does that make any difference?

[stefan6419846]

If having a stream would solve this issue, then it should not matter what type of stream/which stream source we have. Nevertheless, I am not completely sure how much memory pypdf would actually consume for such a task with such large files - regardless of whether we are talking of full blobs or streams.