bpo-32571: Avoid raising unneeded AttributeError and silencing it in C code (alt). #5222
Conversation
…ode. Added _PyObject_GetAttrIdWithoutError() and used in appropriate cases.
…bject_LookupAttrId().
| Return -1 and set *result == NULL if an error other than AttributeError | ||
| is raised. | ||
| */ | ||
| PyAPI_FUNC(int) _PyObject_LookupAttr(PyObject *, PyObject *, PyObject **); |
There was a problem hiding this comment.
I would rename the method to PyObject_LookupAttr (no leading underscore). I'd use it in my libraries and the leading underscore would make me think that I shouldn't.
I also like Inada-san's suggestion to simplify the return value even further:
-1, if an unrelated error occurred during attribute lookup0if no errors have occurred;*resultwill be set toNULLif the attribute was not found, and to a non-NULLvalue if it was found.
There was a problem hiding this comment.
I think this API is experimental and should be kept private for some time. We are still discussing it, and after making some design decision we may change it in future.
There was a problem hiding this comment.
I agree with @serhiy-storchaka.
While I feel this API is nice, I'm not sure we should set
*result=NULL when returning -1 or not.
There was a problem hiding this comment.
Don't we care C API methods with a leading underscore the same as without? I've never seen us breaking the _ prefixed API to be honest. So leading underscore or not it's set in stone.
There was a problem hiding this comment.
@1st1 Actually speaking, I broke _PyObject_GenericGetAttrWithDict(). I added one argument to suppress AttributeError.
|
@serhiy-storchaka Would you merge this? |
|
I guess we can merge this ourselves if Serhiy isn't around. I'd like us to have this in 3.7. |
Add
_PyObject_LookupAttr()and_PyObject_LookupAttrId()and use them in appropriate cases instead ofPyObject_GetAttr(),_PyObject_GetAttrId()and_PyObject_GetAttrIdWithoutError().https://bugs.python.org/issue32571