cyclus_large constant

[bennibbelink]

Many places throughout the codebase 1e299 is used to represent "a very large number". This is not best practice and I suspect it could lead to overflow errors

[bennibbelink]

Turns out we already have constants for this in cyc_limits.h:

/// maximum (+ or -) value for an integer variable
static const int kIntBoundLimit = std::numeric_limits<int>::max();

/// maximum (+ or -) value for a linear variable
static const double kLinBoundLimit = std::numeric_limits<double>::max();

[gonuke]

I found those recently, too, but they seemed to be named in a way that was intended for some specific purpose. Perhaps we can revisit their naming so that it is more clear how they can/should be used.

[bennibbelink]

They're not used anywhere in cyclus/cyclus or cyclus/cycamore. Based on their names it sounds like they might be intended for this specific purpose (specifying an upper/lower bound) but I agree that we could use more intuitive names. What do you think about changing them to something more generic like:

• cy_max_int
• cy_min_int
• cy_max_double
• cy_min_double

Looking at the rest of this file:

• There are also no instances of kConstraintEps anywhere else in the code, and I'm unsure of the use case for it vs. cy_eps.
• kModifierLimit is used only in commodity_producer.h (and a test). It is describe as maximum value for a function modifier (i.e., a_i for variable x_i) in a comment, but its not clear to me whether this is how it's being used in CommodityProducer

[bennibbelink]

Trying to clean up places in cyclus and cycamore where we use 1e299 and running into overflow issues in the Enrichment class when calculating enrichment parameters. It appears that by using std::numeric_limits::max() instead of 1e299 there are some calculations that push us over the edge. @gonuke do you know if 1e299 was picked arbitrarily or is it meant to be an intentional upper bound on some parameters?

Many of the places it is used in cycamore can be found here

[gonuke]

I don't recall how we chose 1e299. Any idea how different it is from the IEEE max double?

[bennibbelink]

The IEEE max double is the same as std::numeric_limits<double>::max() - roughly 1.8e+308. I will continue exploring...