I'm dealing with numbers, that may be too large for int as understood by Python-2.x, so my code's been casting them to long.
In Python-3.x there is no such thing, and one is simply supposed to use int. However, the code needs to work with both Python-2 and 3 (for as long RHEL8 is a thing), so I tried the following trick:
if sys.version_info[0] >= 3:
long = int
...
foo = (long)(bar)
This works in Python-3.x, but in Python-2.x it breaks with: local variable 'long' referenced before assignment. Why wouldn't it just continue to use the built-in type long -- as it did before I inserted the above two lines into the function?
Update, the actual loop, parses the output of Linux stat-utility:
if sys.version_info[0] >= 3:
long_t = int
output = output.decode(locale.getpreferredencoding())
else:
long_t = long
...
for section in [output[x:x + 5] for x in range(0, len(output), 5)]:
stats = [f.split() for f in section]
path = stats[0][1].strip('"')
result[path] = {}
for (field, type, x, y) in [
('id', str, 1, 1),
('namelen', int, 1, 3),
('type', str, 1, 5),
('blocksize', int, 2, 2),
('fblocksize', int, 2, 6),
('blockstotal', long_t, 3, 2),
('blocksfree', long_t, 3, 4),
('blocksavailable', long_t, 3, 6),
('inodestotal', long_t, 4, 2),
('inodesfree', long_t, 4, 4)
]:
result[path][field] = (type)(stats[x][y])
So, as work-around, I introduced a separate variable, long_t:
if sys.version_info[0] >= 3:
long_t = int
else:
long_t = long
...
foo = (long_t)(bar)
This works with both Python-versions, but I still don't understand, why this was necessary...