repr of floating points

[Noam Raphael]

Hello,

repr(0.1) is '0.10000000000000001', in my Python 2.3.
From reading the article "Floating Point Arithmetic: Issues and
Limitations", I understand that it's the closest 17-digit decimal
representation of the binary floating-point number actually stored.

However, I don't understand why this should be what repr of floats
returns. repr should, if possible, return a string that evaluating it
will be == to the original object. But eval("0.1") obviously returns the
same floating point as 0.1 does! So why not make repr of floats return
the *shortest* decimal representation that will evaluate to the stored
binary number? Specifically, why not make repr(0.1) be "0.1"?

Thanks,
Noam Raphael

[Josiah Carlson]

> repr(0.1) is '0.10000000000000001', in my Python 2.3.

From reading the article "Floating Point Arithmetic: Issues and
Limitations", I understand that it's the closest 17-digit decimal
representation of the binary floating-point number actually stored.

However, I don't understand why this should be what repr of floats
returns. repr should, if possible, return a string that evaluating it
will be == to the original object. But eval("0.1") obviously returns the
same floating point as 0.1 does! So why not make repr of floats return
the *shortest* decimal representation that will evaluate to the stored
binary number? Specifically, why not make repr(0.1) be "0.1"?

Search the archives of python-dev. The answer lies within. Search
before asking next time.

- Josiah