Python 3 preparation: Py3 has no more the special object-function "__nonzero__"

[wluebbe]

The tool 2to3 renames __nonzero__ to __bool__.
But the code has to depend on the Python version!

Note that this does not affect Cython extension types because Cython supports __nonzero__ and __bool__ independently of the Python version.

There are 25 affected modules.

This ticket is tracked as a dependency of meta-ticket ticket:16052.

CC: @embray

Component: python3

Issue created by migration from https://trac.sagemath.org/ticket/16076

[fchapoton]

comment:4

one step done (for the rings folder) in #21887

[fchapoton]

comment:5

another step in #21898

[fchapoton]

comment:9

Here is a brute-force tentative

• search and replace __nonzero__ by __bool__
• remove the former lines that have become __bool__ = __bool__

Let us see what happens..

---

New commits:

5c0ed27	substitute `__nonzero__` by __bool__
fbf00d7	cleanup : removal of former alias for __nonzero__

[fchapoton]

Commit: fbf00d7

[fchapoton]

Branch: u/chapoton/16076

[fchapoton]

comment:10

ok, this was too brutal.

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

d3182db

trac 16076 trying to get rid of the last nonzero

[sagetrac-git]

Changed commit from fbf00d7 to d3182db

[fchapoton]

comment:12

ok, still not subtle enough..

[tscrim]

comment:13

The way I've seen for this to be 2/3 compatible is to have __bool__ be an alias for __nonzero__ (or vice versa). The other way we could hack this is have in SageObject, as almost everything has this as a base class, we implement:

def __bool__(self):
    return self.__nonzero__()

[fchapoton]

comment:14

Hello Travis,

if you feel like giving it a try, please do. There is an abstract method __nonzero__ somewhere, which is the last place where __nonzero__ is not an alias (in python files).

[tscrim]

comment:15

I think we should keep things as __nonzero__ for right now (up to having an explicit alias) for right now since we are in Python2 and to avoid the extra redirection. Although I guess to test how much this would cover things for Python3, it would have to be with a Python3 build.

[fchapoton]

Changed branch from u/chapoton/16076 to none

[fchapoton]

Changed commit from d3182db to none

[fchapoton]

comment:18

why sage-pending ?

[embray]

comment:19

I might be misusing the milestone but this seems to be a very general task ticket not tied to a specific release milestone...?

[fchapoton]

comment:20

is this fixed by #24294?

[jdemeyer]

comment:21

Replying to @tscrim:

The way I've seen for this to be 2/3 compatible is to have __bool__ be an alias for __nonzero__ (or vice versa). The other way we could hack this is have in SageObject, as almost everything has this as a base class, we implement:

def __bool__(self):
    return self.__nonzero__()

I know that this reply is pointless is now, but I guess that something like that would have worked :-)

We are doing something similar for division in CategoryObject:

from __future__ import division

cdef class CategoryObject(SageObject):
    def __div__(self, other):
        return self / other

[embray]

comment:22

Replying to @fchapoton:

is this fixed by #24294?

I don't think so by itself--I believe I've found some other examples affected by this (in non-Cython modules) but I could be mistaken.