Parsing fails for single-line OFX v2.0.2 file

[kbauer]

My bank emits OFX files as XML files without any line-breaks, e.g. (after anonymizing the contents):

test.ofx.txt

Parsing fails with a traceback:

>>> python3 -c "import ofxtools; parser = ofxtools.Parser.OFXTree(); parser.parse('test.ofx.txt')"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
    import ofxtools; parser = ofxtools.Parser.OFXTree(); parser.parse('test.ofx.txt')
                                                         ~~~~~~~~~~~~^^^^^^^^^^^^^^^^
  File "C:\Users\...\site-packages\ofxtools\Parser.py", line 82, in parse
    self.header, message = self._read(source)
                           ~~~~~~~~~~^^^^^^^^
  File "C:\Users\...\site-packages\ofxtools\Parser.py", line 115, in _read
    header, message = parse_header(source)
                      ~~~~~~~~~~~~^^^^^^^^
  File "C:\Users\...\site-packages\ofxtools\header.py", line 268, in parse_header
    line = source.readline().decode("ascii")
UnicodeDecodeError: 'ascii' codec can't decode byte 0xc3 in position 920: ordinal not in range(128)

Note that the decoding is actualy unnecessary here, as this is in the loop for skipping empty lines preceding the OFX header, if any.

The bug could be solved by one of these:

• Clean: Simply do not decode. Modern Python3 versions support text operations like .strip() also on bytes. However, I don't know if this would raise the minimum required Python version.
• Safe: Decode to an 8-bit encoding like "latin1" instead of "ascii". Unlike "ascii", 8-bit encodings can decode any byte stream. I'm not sure what would happen on a null character though.

[kbauer]

I checked the two concerns:

• Decoding null-bytes does not do anything special, just gives a null character.
• bytes.strip exists at least since Python 3.4.

So both options would work.

[kbauer]

Version 0.9.5, via pip. Reproducible in master, but the fix is less clear there.

[kbauer]

Safest probably:

    line = source.readline().decode("ascii", errors="replace")