The clean up of syntax trees and expressions has made apparently a problem with
comparing expected and actual Rational numbers in unit tests
testConvexRewriterOnMessageBoundsBasicInnerUnconditional and
testConvexRewriterOnMessageBoundsBasicInnerConditional.
These tests involve, among others, a Rational-valued symbol with value 6/7,
which is output as 0.857142857. However, writing 6/7 as a decimal results in a
periodic decimal expansion, so this Rational does not equal an expected value
written down as 0.857142857. In fact, it will not equal *any* finite decimal
description.
I am not quite sure how this used to work, but I left the old methods around
for reproduction. It seems that it should not have worked though, and if it
did, it was by some other covering problem.
It seems that one way of solving the problem is to introduce notation for
specifying Rational-valued symbols by using fractions. Simply writing a
fraction is not enough because it will be taken as a function application of /
and still not compare as equal to a Rational symbol.
Original issue reported on code.google.com by rodrigob...@gmail.com on 23 Apr 2014 at 1:36
Original issue reported on code.google.com by
rodrigob...@gmail.comon 23 Apr 2014 at 1:36