Turn --fast-parser on by default #2734
Conversation
|
Python 3.3 has a different error for print statements that are missing parens (from |
ddfisher
force-pushed
the
fastparser-default
branch
from
a79e4ea to
f28329f
Compare
ddfisher
force-pushed
the
fastparser-default
branch
from
f28329f to
3e7834e
Compare
|
@gvanrossum: This should be ready for review now! There's a lot of test output that's been changed -- let me know if there's some way I can make the review easier. |
| [out] | ||
| file:2: error: Unknown encoding 'uft8' | ||
|
|
||
| [case testDoubleEncoding3] |
There was a problem hiding this comment.
Curious why you had to delete this test and the next?
There was a problem hiding this comment.
Because they were wrong. Neither of them causes an error when run under Python 3, and when run under mypy with the fast parser there is correctly no error as well. I'll fix the output and add them back (but in parse.test because parse-errors.test can only contain errors).
There was a problem hiding this comment.
Hmm. Adding this test to parse.test fails because we purposefully assume most test cases are ascii in several places, and we mention "portability issues" as part of the reason. I'm going to omit these test cases for now.
There was a problem hiding this comment.
I think the "portability issues" reason is a feeble excuse, UTF-8 as the default for Python 3 is universal. More likely either there's some code somewhere that doesn't encode/decode right, or it's hard to edit files with UTF-8 in them (I can agree with the latter). But it's fine.
|
|
||
| [case testNotAsBinaryOp] | ||
| x not y # E: Parse error before "y" | ||
| x not is y # E: Parse error before is |
There was a problem hiding this comment.
Technically you ought to add a separate test for x not is y
| 1 1 | ||
| [out] | ||
| main:1: error: Parse error before in | ||
| main:3: error: Parse error before numeric literal |
There was a problem hiding this comment.
Again, this test should be split into two.
There was a problem hiding this comment.
Actually, the test should probably just be deleted. It looks like it's supposed to test multiple syntax errors, which we will no longer display.
|
(See my first comment here for what's going on with Python 3.3. I think skipping that test for now is probably reasonable.) |
Since January 2017, it is now enabled by default and the only parser available. python/mypy#2734
2 participants
Resolves #1353.
Depends on #2710, #2722, and #2730. The first new commit in this PR is "Make --fast-parser the default".