Acid tests should check bytecode too about autopep8 HOT 5 CLOSED

I explored bytecode comparison when I first wrote the acid tests. I gave up on it once I found that things like has_key() and in didn't collapse into the same bytecode. But since autopep8 is now doing things a bit more complicated like shortening lines, it might be worth doing what you suggest. I suppose we would need to disable everything except whitespace, indentation, and line length changes.

By the way, the idiomatic changes I mention are basically the things 2to3 -f idioms fixes.

from autopep8.

See 8631ecb.

from autopep8.

I should probably do a recursive comparison of the code object rather than using dis. My current implementation with dis is basically just checking the top-level bytecode.

from autopep8.

For the most part, this works well. But sometimes the same code section generates different, but equivalent, bytecode representations. For example

16 LOAD_CONST             338 ((('catalogue', '0001_initial'), ('customer', '0001_initial'), ('partner', '0001_initial'), ('address', '0001_initial')))

versus

16 LOAD_CONST               1 ('catalogue')
19 LOAD_CONST               2 ('0001_initial')
22 BUILD_TUPLE              2
25 LOAD_CONST               3 ('customer')
28 LOAD_CONST               2 ('0001_initial')
31 BUILD_TUPLE              2
34 LOAD_CONST               4 ('partner')
37 LOAD_CONST               2 ('0001_initial')
40 BUILD_TUPLE              2
43 LOAD_CONST               5 ('address')
46 LOAD_CONST               2 ('0001_initial')
49 BUILD_TUPLE              2
52 BUILD_TUPLE              4

from autopep8.

Comparing the AST works better.

from autopep8.