Convert PDF page to image with pyPDF2 and BytesIO

Question

I have a function that gets a page from a PDF file via pyPdf2 and should convert the first page to a png (or jpg) with Pillow (PIL Fork)

from PyPDF2 import PdfFileWriter, PdfFileReader
import os
from PIL import Image
import io

# Open PDF Source #
app_path = os.path.dirname(__file__)
src_pdf= PdfFileReader(open(os.path.join(app_path, "../../../uploads/%s" % filename), "rb"))

# Get the first page of the PDF #
dst_pdf = PdfFileWriter()
dst_pdf.addPage(src_pdf.getPage(0))

# Create BytesIO #
pdf_bytes = io.BytesIO()
dst_pdf.write(pdf_bytes)
pdf_bytes.seek(0)

file_name = "../../../uploads/%s_p%s.png" % (name, pagenum)
img = Image.open(pdf_bytes)
img.save(file_name, 'PNG')
pdf_bytes.flush()

That results in an error:

OSError: cannot identify image file <_io.BytesIO object at 0x0000023440F3A8E0>

I found some threads with a similar issue, (PIL open() method not working with BytesIO) but I cannot see where I am wrong here, as I have pdf_bytes.seek(0) already added.

Any hints appreciated

Solution

Per document:

write(stream) Writes the collection of pages added to this object out as a PDF file.

Parameters: stream – An object to write the file to. The object must support the write method and the tell method, similar to a file object.

So the object pdf_bytes contains a PDF file, not an image file.

The reason why there are codes like above work is: sometimes, the pdf file just contains a jpeg file as its content. If your pdf is just a normal pdf file, you can't just read the bytes and parse it as an image.

And refer to as a more robust implementation: https://stackoverflow.com/a/34116472/334999