-
-
Notifications
You must be signed in to change notification settings - Fork 31.6k
bpo-11063, bpo-20519: avoid ctypes and improve import time for uuid #3796
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Please add a NEWS entry. |
The potential issue here is that the uuid headers are not always installed (for example they aren't on the Travis-CI machine, even though it has many development headers installed). |
PyMODINIT_FUNC | ||
PyInit__uuid(void) | ||
{ | ||
assert(sizeof(uuid_t) == 16); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You might use Py_BUILD_ASSERT() to run the check at the compilation.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is sizeof()
allowed in macros? Usually we have SIZEOF_*
for that.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't know. In case of doubt, keep your assert(), it's fine.
Modules/_uuidmodule.c
Outdated
|
||
|
||
static PyObject * | ||
_uuid_generate_time_safe(void) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I suggest to add a py prefix: py_uuid_generate_time_safe().
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Agreed.
Modules/_uuidmodule.c
Outdated
NULL, | ||
-1, | ||
uuid_methods, | ||
NULL, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe remove trailing "NULL,". The compiler set them to 0 / NULL anyway.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yup.
Modules/_uuidmodule.c
Outdated
"_uuid", | ||
NULL, | ||
-1, | ||
uuid_methods, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can you use C99 style here to avoid surprising/boring "NULL," members?
.m_name = "_uuid",
.m_size = -1,
.m_methods = uuid_methods,
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, great, I'll do that.
Lib/uuid.py
Outdated
else: | ||
def _generate_time_safe(): | ||
return _uuid.generate_time_safe() | ||
_has_uuid_generate_time_safe = True |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why not just _generate_time_safe = _uuid.generate_time_safe?
I would prefer an explicit "return" here.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Hmm, ok.
|
||
@unittest.skipUnless(c_uuid, 'requires the C _uuid module') | ||
class TestInternalsWithExtModule(BaseTestInternals, unittest.TestCase): | ||
uuid = c_uuid |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't see the point of testing all uuid.*getnode() functions with and without _uuid. Basically, it's the same test run twice, no?
Only getnode() and uuid1() should be tested with and without _uuid, no?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Only getnode() and uuid1() should be tested with and without _uuid, no?
Probably. I didn't want to modify the tests too much but I'll take a look.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've changed my mind: I think all tests should be run both with and without the C _uuid
extension module. It doesn't cost much to do so, and it's more future-proof.
A Python core developer has requested some changes be made to your pull request before we can consider merging it. If you could please address their requests along with any other requests in other reviews from core developers that would be appreciated. Once you have made the requested changes, please leave a comment on this pull request containing the phrase |
@pitrou: "The potential issue here is that the uuid headers are not always installed (for example they aren't on the Travis-CI machine, even though it has many development headers installed)." That's fine. We have buildbots for further tests. |
@Haypo, I think I've addressed your comments. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah, the new PR now looks better! I hope that it's my last round of minor comments :-)
PyModuleDef_HEAD_INIT, | ||
.m_name = "_uuid", | ||
.m_size = -1, | ||
.m_methods = uuid_methods, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, I really like this syntax. We should use it more often!
setup.py
Outdated
@@ -1668,6 +1668,20 @@ class db_found(Exception): pass | |||
if '_tkinter' not in [e.name for e in self.extensions]: | |||
missing.append('_tkinter') | |||
|
|||
# Build the _uuid module if possible | |||
build_uuid = False | |||
if find_file("uuid.h", inc_dirs, ["/usr/include/uuid"]): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Since you add /usr/include/uuid to inc_dirs, I would maybe store find_file() result into uuid_incdirs and then pass it to Extension(): inc_dirs=uuid_incdirs? Similar code is used for ssl_incs.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok.
setup.py
Outdated
# Build the _uuid module if possible | ||
build_uuid = False | ||
if find_file("uuid.h", inc_dirs, ["/usr/include/uuid"]): | ||
if self.compiler.find_library_file(lib_dirs, 'uuid'): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You may also store the result of find_library_file() into uuid_libdirs to pass it to Extension library_dirs. I'm not sure about this one.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let me try...
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No, it shouldn't be necessary. Other uses of find_library_file()
don't do it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ok, fine.
Lib/test/test_uuid.py
Outdated
@@ -14,7 +17,8 @@ def importable(name): | |||
except: | |||
return False | |||
|
|||
class TestUUID(unittest.TestCase): | |||
class BaseTestUUID: | |||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add "uuid = None" here.
uuid = c_uuid | ||
|
||
|
||
class BaseTestInternals: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add "uuid = None" here.
Lib/uuid.py
Outdated
continue | ||
# Try to find the safe variety first. | ||
if hasattr(lib, 'uuid_generate_time_safe'): | ||
_uuid_generate_time = lib.uuid_generate_time_safe |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Would you mind to rename "_uuid_generate_time" to "_uuid_generate_time_safe"? I was confused while reading the code.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, sure.
if _has_uuid_generate_time_safe is not None: | ||
return | ||
|
||
_has_uuid_generate_time_safe = False |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm not sure of the value of _has_uuid_generate_time_safe. The exact value (True or False) is not use. Maybe rename the flag to _system_functions_initialized or something like that, and only use False/True values?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's used in the tests.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, I didn't notice. In this case, it's fine. I was just surprised to see 3 states where only 2 states are used in uuid.
I hacked the code to disable _uuid, uuid_generate_time_safe and uuid_generate_time in uuid.py. With these changes, tests fail:
Python 3.6 test contains:
|
You're right, I'm gonna reinstate the |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM as soon as you fix the test failure I reported in my latest comment.
Is it possible to get this back-ported into (i guess, at least) 3.6? |
No, sorry, it's not possible. The change is too big to be qualified as a
bugfix. For me, it's a new feature and so cannot be backported. The risk of
regression is too high.
The best you can do is to write a 3rd party extension on PyPI to backport
the module to 3.6 and maybe older (2.7??).
|
https://bugs.python.org/issue11063
https://bugs.python.org/issue20519