ScopeProxy doesn't update when c++ symbol updates

[wrongbad]

Here's a complete reproducible example:

>>> import cppyy
>>> cppyy.include("iostream")
True
>>> cppyy.cppexec("int x = 1; std::cout << x << std::endl;")
1
True
>>> cppyy.gbl.x
1
>>> cppyy.cppexec("int x = 2; std::cout << x << std::endl;")
2
True
>>> cppyy.gbl.x
1
>>> del cppyy.gbl.x
>>> cppyy.cppexec("int x = 3; std::cout << x << std::endl;")
3
True
>>> cppyy.gbl.x
3

I would expect the 2nd eval of cppyy.gbl.x to return 2 but it seems to have cached the value 1 until I explicitly delete the cached value.

[wlav]

Yes, on the Cling side, the redeclaration simply masks the first declaration. Ie., there are 2 global variables named x in the system; it's not an update of the original variable. Since lookups are expensive and there is no callback from Cling to indicate that a variable has been redeclared, there's no easy way to propagate such an update other than forcing a fresh lookup by using an explicit del.

Note that the same is true in Cling:

>>> import cppyy
>>> cppyy.cppexec("int x = 1;")
True
>>> cppyy.cppdef("int f() { return x; }")
True
>>> cppyy.gbl.f()
1
>>> cppyy.cppexec("int x = 2;")
True
>>> cppyy.gbl.f()
1
>>> cppyy.cppdef("int g() { return x; }")
True
>>> cppyy.gbl.g()
2
>>> 

Again, this is because a redeclaration isn't a replacement.

[wrongbad]

I see. Thanks for the explanation.

And to be clear, you mean symbol lookups are slow, but looking up new values at the same address is fine to do each time right?

[wlav]

Yes. Lookups are slow b/c on the Python side, all that is known is that name x, not what it is and thus where to find it. I.e., there's a whole range of lookups happening (is this a class? a function? an enum? a ...). Furthermore, the following is also legal in Cling:

>>> import cppyy
>>> cppyy.cppexec("int x = 1;")
True
>>> cppyy.gbl.x
1
>>> del cppyy.gbl.x
>>> cppyy.cppexec('class x {};')
True
>>> cppyy.gbl.x
<class cppyy.gbl.x at 0x149704dc0>
>>> 

Thus storing "x is a global variable of type int", although sufficient for the example above, is also not a general solution. I.e., an actual complete solution would require a complete lookup on every use, which will give a couple orders of magnitude performance reduction in regular use. (Note that erasing a class doesn't work the same as erasing an int global variable, so the opposite case doesn't actually work.)