Feature/initial v2

[whaeck]

A preliminary prototype implementation of NDItk for reading, writing and modifying the multigroup neutron and photon NDI tables.

The capability is limited but it illustrates how most of the toolkit will work. We cover the basic cases that we should encounter while reading records:

• reading metadata
• reading simple records consisting of lists of reals and integers
• reading more complex records that need to be divided into subrecords
• basic verification of the data to check consistency (e.g. are the number of groups consistent across the table)

The python bindings are currently missing, but I will add these shortly.

[tgsaller]

I'll look at this a lot more next week, but here are some scattered comments for now.

[tgsaller]

I think I've changed my mind. Instead of structure or group structure, energyBounds is probably the most accurate name for this.

[whaeck]

I have made changes to these, check the latest version to see if these changes are OK for you. Basically, every reocrd like this will have a values() that returns the data while the table has a corresponding primaryGroupStructure() or similar function. As a result, the interface will look like:

table.primaryGroupStructure().values()
table.outgoingGroupStructure( 0 ).values()

[tgsaller]

outgoingGroupStructure only applies for secondary particles? And is ( 0 ) index 0, or saying particle type 0 (gamma)?

[whaeck]

Particle id, so 0 for gamma and 1001 for protons. I distinguish primary and outgoing structures. The primary structure is for the primary particle, aka the projectile and I assume the structure for the outgoing projectile type is the same as the primary structure (hence the scattering matrix being square while the production ones aren;t necessarily square)

[tgsaller]

I'm unfamiliar with "|" in C++. What does this do?

[whaeck]

This is a C++20 addition. While we use C++17, this was added in the C++20 extension from njoy::tools (it replaces range-v3). The | operator (or pipe operator) is shorthand for:

return njoy::tools::std20::views::all_view( this->values_ );

While in this case it offers little advantage, there are cases where it adds readability:

auto f = [] ( auto&& value ) { /* do something */ };
return this->values_ | njoy::tools::std20::views::drop( 5 ) 
                     | njoy::tools::std20::views::transform( f );

which means: "drop the first 5 elements from this->values_ and then apply the transformation f to it". This is equivalent to:

auto f = [] ( auto&& value ) { /* do something */ };
return njoy::tools::std20::views::transform_view( njoy::tools::std20::views::drop_view( this->values_, 5 ), f );

[tgsaller]

My new concern is that some machines won't have C++20 things. Maybe they do, it's 2024. Do you know if Rocinante can handle that?

[whaeck]

While this is indeed a c++20 extension, it was implemented in the tools library using c++17 standard programming. The interface is in line with the c++20 specs. I have tested this even on intel-classic compilers and it works. You should not need a C++20 compatible compiler, you need a c++17 compatible one. Once we can actually switch standards, we can do away with all this.

One of the things we have to do soon is extend the testing capabilities we have to test more OS types and compilers. I plan on talking to the MCNP team about that.

[tgsaller]

Can you add some cross sections to one of these reactions? So it's like 2 groups for one, and 0 groups for the other.

[joewpeterson]

Is there any chance that in the future we will want to write out more or fewer than 5 entries per line? If yes, it might be worth our while now to set some #cols constant value we can change in one place. Write an internal loop within while( full -- ) to loop over #cols

[whaeck]

Not sure, but magic numbers are to be avoided anyway so why not change it? Other than global consts I don't see an easy way to do this though.

The same would go for the output precision too.

[joewpeterson]

purely a taste thing here, but to me
auto partial = this->data().size() % 5;
is clearer?

[whaeck]

Yes, you are correct. I'll change it.

[tgsaller]

Looks good!