Bug in gaussian parsing polarizability

In my .log file the gaussian parser chokes trying to parse polarizabilities:

Attempting to read foo.log
[Gaussian foo.log ERROR] Encountered error when parsing.
[Gaussian foo.log ERROR] Last line read:   Exact polarizability:  -20751.174      -0.389    1217.309   -6307.929       0.037    -646.713

Traceback (most recent call last):
  File "/home/oliver/.local/bin/ccget", line 8, in <module>
    sys.exit(ccget())
  File "/home/oliver/.local/lib/python3.8/site-packages/cclib/scripts/ccget.py", line 176, in ccget
    data = ccread(filename, **kwargs)
  File "/home/oliver/.local/lib/python3.8/site-packages/cclib/io/ccio.py", line 169, in ccread
    return log.parse()
  File "/home/oliver/.local/lib/python3.8/site-packages/cclib/parser/logfileparser.py", line 315, in parse
    self.extract(inputfile, line)
  File "/home/oliver/.local/lib/python3.8/site-packages/cclib/parser/gaussianparser.py", line 2110, in extract
    polarizability[indices] = [utils.float(x) for x in
  File "/home/oliver/.local/lib/python3.8/site-packages/cclib/parser/gaussianparser.py", line 2110, in <listcomp>
    polarizability[indices] = [utils.float(x) for x in
  File "/home/oliver/.local/lib/python3.8/site-packages/cclib/parser/utils.py", line 74, in float
    return _BUILTIN_FLOAT(number.replace("D", "E"))
ValueError: could not convert string to float: '.309   -'

This happens because the parsing is based on exact line positions, which does not apply in this case at least:

polarizability[indices] = [utils.float(x) for x in
                                           [line[23:31], line[31:39], line[39:47], line[47:55], line[55:63], line[63:71]]]

It can presumably be fixed by using line.split() or similar instead. I can submit a PR but not the .log file sadly as it is private.

[shivupa]

@ghutchis I was thinking that the older format could possibly work but put nonsense in each of the slices of the line, since the older line is shorter than the new format. If we try the new format first, we would get an out of range error triggering the except.

[ghutchis]

@shivupa - would it be better to try the old one by default and use the except as the newer version? I’m guessing there are far more files with the older format.

@oliver-s-lee - would you be able to run the testdata file with C01 for a regression test?