How to fix non-utf8 names when loading old blend files?

Question

When loading old files, it's possible the names aren't utf-8 compliant. Meaning Python won't be able to modify them. See: T23197

Besides using a hex editor, is there a way to fix names from very old blend files so they are utf-8 compliant?

Answer 1

It's possible to get strings in Python as bytes, which allows to re-encode them into utf-8:

Eg:

import bpy
for m in bpy.data.materials:
    name_as_bytes = m.path_resolve("name", False).as_bytes()
    name_as_utf8 = name_as_bytes.decode('utf8', 'surrogateescape')
    m.name = name_as_utf8

To run this on your entire file:

import bpy

def id_iter():
    type_iter = type(bpy.data.objects)

    for attr in dir(bpy.data):
        data_iter = getattr(bpy.data, attr, None)
        if type(data_iter) == type_iter:
            for id_data in data_iter:
                if id_data.library is None:
                    yield id_data

def wash_name(id_data)
    name_as_bytes = id_data.path_resolve("name", False).as_bytes()
    name_as_utf8 = name_as_bytes.decode('utf8', 'surrogateescape')
    id_data.name = name_as_utf8

for id_data in id_iter():
    wash_name(id_data)

---

Ideally loading old files will always work. In this case it's not done because its a big change for a relatively rare problem.