The code generator changes semantics by introducing parentheses

[lschuetze]

During m2m transformation, the generator adds parentheses to arithmetic expression that had not been in the former OpenPME expression. As a result, the semantics of the arithmetic expression is changed.

This can be seen in OpenPME example of Lennard Jones.

which has no parentheses and the resulting generated code

As you can see, the 2.0 is applied to both operands of the inner minus expression.

[svenkarol]

I found the same problem several times. Maybe it's not related to the code generator but the node tree created by the view. See the equation below:

If you now look into the node explorer you get this:

Seems that the tree was created without considering the operator precedences.

[lschuetze]

Actually, it is the opposite. The tree is created according to an equation that would use parentheses. Thus, the corresponding model is correct w.r.t. that equation that would use parentheses. If you would not create the tree according to the operator precedence the result will be different.

The problem is that the user perception is a different one as the view does not represent the resulting model if you take ordinary operator precedence and association into account.

To fix this:

• the equation must include parenthesis
• there must be a mapping from parenthesis of openpme.core to openpme.cpp
• reduction rules for binary expressions in the generator of openpme.cpp must not create parenthesised expressions

There is a bigger usability problem which should be introduced in a different issue as you (for the time being) cannot just start writing your equation from left-to-right and the model is not constructed following the precedence rules of the operators in the whole equation. The user currently has to be carefully writing the expression.