This one is amusing … the case of the mysterious syntax error … turns out that repeated keyword arguments (kwargs) were illegal in python 2.4, ignored in python 2.5 and illegal again in python 2.6. This means that if some have crept into your codebase, you’ve now got a handful of syntax errors!
For example, here’s test.py:
def foo(**kwargs):
for k, v in kwargs.iteritems():
print "%s: %s" % (k, v)
foo(a = 1, b = 2, a = 3)
And here’s what happens when you run it under 2.4, 2.5 and 2.6:
nick@pluto:~/tmp$ python2.4 test.py
File "test.py", line 5
foo(a = 1, b = 2, a = 3)
SyntaxError: duplicate keyword argument
nick@pluto:~/tmp$ python2.5 test.py
a: 3
b: 2
nick@pluto:~/tmp$ python2.6 test.py
File "test.py", line 5
foo(a = 1, b = 2, a = 3)
SyntaxError: keyword argument repeated
How could this creep in? Most likely through revision control mergers. Function calls with lots of kwargs are often laid out like:
foo(
kwarg_with_a_long_name=1,
another_self_documenting_flag=False,
)
Alice adds a kwarg for yadda="yadda yadda", at the start on the kwargs list, before kwarg_with_a_long_name. So does Bob, but Bob adds it at the end. The revision control system will happily merge these two changes without conflicts as:
foo(
yadda="yadda yadda",
kwarg_with_a_long_name=1,
another_self_documenting_flag=False,
yadda="yadda yadda",
)
And under python 2.5, neither Alice or Bob are likely to notice … after all, yadda=”yadda yadda” just as they intended.
But when the underlying system is upgraded to python 2.6, a mysterious “SyntaxError: keyword argument repeated” appears.
