Sage silently accepts symbolic expressions with two comparison operators

[sagetrac-tmonteil]

See the following example:

sage: 0 <= x <= 1/2
0 <= x

This should raise an error. Note that it can create confusion when used in assumptions:

sage: assume(0 <= x <= 1/2)
sage: bool(x <= 1/2)
False

This has been reported on this ask question.

CC: @rwst @slel

Component: symbolics

Issue created by migration from https://trac.sagemath.org/ticket/24726

[jdemeyer]

comment:1

This is not a solvable problem. The problem is that Python treats

0 <= x <= 1/2

as

(0 <= x) and (x <= 1/2)

Since bool(0 <= x) is False, this returns (0 <= x).

There is proposal at PEP 535 to fix this, but it's not clear whether it will be accepted.

[sagetrac-tmonteil]

comment:3

I got your point, but it is not clear for me that this implies that there is nothing to do on the symbolic ring to mitigate the issue.

[jdemeyer]

comment:4

Replying to @sagetrac-tmonteil:

I got your point, but it is not clear for me that this implies that there is nothing to do on the symbolic ring to mitigate the issue.

As long as bool(0 <= x) returns False, it is not possible to fix this.

[slel]

comment:5

We could let the Sage preparser turn multiple inequalities in assume statements into multiple assumptions.

[rwst]

comment:6

That's an interesting idea. And it could raise an error outside assumptions. Let's at least wait for that PEP before closing this.

[jdemeyer]

comment:8

Replying to @slel:

We could let the Sage preparser turn multiple inequalities in assume statements into multiple assumptions.

Why special-case assume? The problem already occurs before calling assume:

sage: 0 <= x <= 1/2
0 <= x

[jdemeyer]

comment:9

And if you do want to hack the preparser for that, I would do it with the PEP 535 semantics.

[jdemeyer]

comment:10

Since PEP 535 is not particularly likely to come soon, I suggest to close this as "wontfix".