Strange xml generation for an intension constraint with distance expression

[wjsuijlenh]

Dear PyCSP3 developers,

I came across a strange XML generation behaviour regarding an intension expression of the form z == abs(x[y] - x[i]) for i in range(n), where x, y, z are variables and n is a constant. The generated expression in XML becomes eq(z,abs(neg(sub(aux_gb[0],%0)))).

I did not expect to see a neg in here.

If I swap the places places of x[y] and x[i], I get eq(z,dist(aux_gb[0], %0)) , which is perfect!

My guess is that you intended that the first case also had a similar conversion.

Here is a small self-contained Python code that demonstrates the behaviour

import pycsp3
from pycsp3 import *

n = 2
m = 3
x = VarArray(size = n, dom = range(m))
y = Var(dom = range(m))
z = Var(dom = range(m))
i = Var(dom = range(n))

satisfy(
  (y == abs(x[j] - x[i]) for j in range(n)),
  (z == abs(x[i] - x[j]) for j in range(n))
)

On today's master (9af123e1) this generates the following XML

<instance format="XCSP3" type="CSP">
  <variables>
    <array id="x" size="[2]"> 0..2 </array>
    <var id="y"> 0..2 </var>
    <var id="z"> 0..2 </var>
    <var id="i"> 0 1 </var>
    <array id="aux_gb" note="aux_gb[i] is the ith auxiliary variable having been automatically introduced" size="[1]"> 0..2 </array>
  </variables>
  <constraints>
    <group>
      <intension> eq(y,abs(neg(sub(aux_gb[0],%0)))) </intension>
      <args> x[0] </args>
      <args> x[1] </args>
    </group>
    <group>
      <intension> eq(z,dist(aux_gb[0],%0)) </intension>
      <args> x[0] </args>
      <args> x[1] </args>
    </group>
    <element>
      <list> x[0] x[1] </list>
      <index> i </index>
      <value> aux_gb[0] </value>
    </element>
  </constraints>
</instance>

To summarize, I would have expected that in the XML above, the first generated intension constraint would

1. not contain a neg operation
2. be similar to the second generated intension expression

[xcsp3team]

I didn't attempt to fix the problem at the source of the process generating the expression. I now impose some reasonable rules simplifying expressions, as e.g., abs(neg(.)) => abs(.) I have pushed the code, and now, regarding your example, you should get the same expressions. (we plan to make an update on Pypi by the end of August)

best.

[wjsuijlenh]

Nice! Thanks. I tried this morning's master branch (67ff8626) and it works now as expected.