Pymupdf: Question / Comment: Open PDF files as streams with fitz.open()

[liamsuma]

I am working on a project that takes PDFs as streams downloaded from Azure Blob Storage. I need PyMuPDF to open the stream and read content just like a normal file. Is there a way to achieve this? For example, I have a test case here using PyPDF2:

from io import BytesIO
stream_buffer = BytesIO()
with stream_buffer as download_stream:
    blob_client.download_blob(max_concurrency=3).readinto(download_stream) # just download PDFs from Azure as in-memory streams
    fileReader = PyPDF2.PdfFileReader(stream_buffer) # working with stream
    print(fileReader.numPages) # this will give me number of pages from each PDF 

I also took a look at the official doc here about opening a doc which states the input file is the actual file. Is there a parameter that I could pass to recognize it as a stream in PyMuPDF? or is this something I need to do before feeding docs to PyMuPDF i.e. getbuffer() or getvalue() from io library. I also found an old version of PyMuPDF back in 2016 with similar issue here and wonder if anything has changed since then.

Thanks for your valuable input and suggestions.

[JorjMcKie]

The documentation says this:
Overview of possible forms (using the open synonym of Document):

>>> # from a file
>>> doc = fitz.open("some.pdf")
>>> doc = fitz.open("some.file", None, "pdf") # copes with wrong extension
>>> doc = fitz.open("some.file", filetype="pdf") # copes with wrong extension
>>>
>>> # from memory
>>> doc = fitz.open("pdf", mem_area)
>>> doc = fitz.open(None, mem_area, "pdf")
>>> doc = fitz.open(stream=mem_area, filetype="pdf")
>>>
>>> # new empty PDF
>>> doc = fitz.open()
>>>

"From memory" seems what you are look for. mem_area may be one of the Python types bytearray, bytes, io.BytesIO.

[liamsuma]

Thanks for such a quick reply. Please allow me to test it and get back to you ASAP.

[liamsuma]

It worked and I should've read more carefully about docs.

stream (bytes,bytearray,BytesIO) –
A memory area containing a supported document. Its type must be specified by either filename or filetype.
(Changed in version 1.14.13) io.BytesIO is now also supported.

I am closing this issue now and thanks again for your quick help.