According to the documentation repr() and str() are different when it comes to number formatting.  A test with a 100 million random and selected IEEE 64-bit values returned no differences

Would you mind linking to the relevant part of documentation?

This discrepancy was supposed to be eliminated in 3.2; see Issue 9337. So assuming you are looking at the right version, it is the documentation that is at fault.

> A test with a 100 million random and selected IEEE 64-bit values 
> returned no differences

The float type's tp_str and tp_repr both call float_repr in Objects/floatobject.c. See the 3.5.1 PyFloat_Type definition [1].

Perhaps you were reading something in reference to Python 2, which has separate float_str and float_repr functions. See the 2.7.11 PyFloat_Type definition [2] and the header file where PyFloat_STR_PRECISION is defined to be 12 digits [3].

[1]: https://hg.python.org/cpython/file/v3.5.1/Objects/floatobject.c#l1839
[2]: https://hg.python.org/cpython/file/v2.7.11/Objects/floatobject.c#l2118
[3]: https://hg.python.org/cpython/file/v2.7.11/Include/floatobject.h#l24

Closing: as Martin pointed out, the behaviour is intentional, and without a reference to a particular part of the documentation this issue isn't useful. (Anders: if you find a part of the 3.5 documentation that does state that str and repr are different for floats, please feel free to re-open and link to that section of the documentation so that it can be fixed.)

Apparently the docs have changed since 2.7:
https://docs.python.org/3.5/tutorial/floatingpoint.html

However, the documentation still "sort of" mentions repr() as the most accurate form which isn't entirely correct since it nowadays is identical to str() for floats.

No big deal, I just thought I was doing something wrong :-)

related: http://bugs.python.org/issue26229

stage: resolved

assignee: docs@python
components: + Documentation, - Interpreter Core