[Python-3000] Removing repr (original) (raw)

Greg Ewing greg.ewing at canterbury.ac.nz
Fri Apr 7 02:23:42 CEST 2006

Michael P. Soulier wrote:

I thought that the distinction was as follows.

str() should return a human-readable representation of something. repr() should return a machine-readable representation of something, suitable for an exec command.

As Michael pointed out, this is totally impractical in the majority of cases.

The guideline I use is quite different: str() is for normal program output, repr() is for debugging output.

To expand a bit more: repr() should return something that unambiguously indicates the type of the object. If something about the value can be indicated as well without being too verbose, that's a bonus, but it's not the most important thing.

I suggest that the passage in the docs about repr() output being eval-able be deleted, since it misleads more than it helps, and replaced with something to the above effect.

BTW, these days I normally don't bother giving my classes a repr at all, and just make do with the default <someclass instance at 0xXXXXX>. If I want to know anything more, I insert prints at the appropriate place to probe the object further.

Anyway, -1 on combining them. In my view they do serve two very clearly distinct purposes -- just not the ones the docs say!

-- Greg

More information about the Python-3000 mailing list