In some environments (cloud) its sometimes not optimal to save/load files from disk and instead load them in remotely. In such cases you can get an io Buffer object (like io.BytesIO or io.BufferedReader) which behaves exactly like a file without being a file on disk.

The library in its current state theoretically supports using such objects when using MODE_FD but in practice it doesn't allow this because L86 in reader.py tries to access the name attribute of the buffer which does not exist because it doesn't wrap a file. Python also doesn't allow you to set that attribute on a buffer yourself by simply assigning it.

(

Considering the filename defined in L86 is only used in error messages, is it possible to add a fix that defaults to some value if the attribute does not exist for the buffer?

POC for the issue:

import io
import maxminddb

f = open(<your-mmdb-name-here>, "rb")
buf = io.BytesIO(f.read())

reader = maxminddb.open_database(buf, MODE_FD) # this fails '_io.BytesIO' object has no attribute 'name'

# similarly you can wrap into a BufferedReader

f = open(<your-mmdb-name-here>, "rb")
buf = io.BytesIO(f.read())
stream = io.BufferedReader(buf)

reader = maxminddb.open_database(stream, MODE_FD) # this again fails for missing attribute

Note that just defaulting to something in L86 fixes this issue and allows the reader to work correctly.

I would be more than happy to contribute a fix for this if you want, it should be a very small change.