Support BOM in xml feeds

[Alek050]

Suggested fix for BOM's in xml feeds.

closes #539

[probberechts]

Thanks for looking into this! Would it be possible to use the utf-8-sig encoding instead? If that works, I’d prefer it over the current workaround, as it feels a bit hacky.

[Alek050]

@probberechts I tried that first, but it raised the same error:

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xef in position 0: unexpected end of data
decoding with 'utf-8-sig' codec failed

[probberechts]

That's odd. It seems to work fine for me. Here is a minimal example:

import io
bom = b"\xef\xbb\xbf"
xml_content = bom + b'<?xml version="1.0"?><root/>'
feed = io.BytesIO(xml_content)
first_char = feed.read(4).decode("utf-8-sig")
assert first_char == '<'

Am I missing something?

[Alek050]

Hahaha no, my bad.

I stupidly tested feed.read(1).decode("utf-8-sig")instead of feed.read(4). I changed the code.

The only assumption is that the 4th byte is now the first of the xml string. This holds for utf-8 encoding, however for utf-16 and utf32 it is 2 and 4 bytes respectively. But I guess we can keep it like this because the .decode("utf-8-sig") will not work on the other encoding types anyway (?).