SNHA algorithm implementation (WIP)

[mittelmark]

Please feel free to have a look at our ongoing implementation of a Bioinformatics algorithm using Fusion:

https://github.com/mittelmark/snha4any

Currently there are 4 classes with support for statistical methods, random numbers matrices and graph generation. More will follow.

Best,
Detlef

[pfusik]

Thanks for sharing! It's great to see what people are doing in Fusion.

I spotted an issue in your C implementation of FileRead and some minor simplifications, and submitted PRs.

[ariankordi]

The CSV parser and filesystem classes are great inclusions here. I've implemented CSV in my own project as well to read unit test fixtures, but yours looks more "generic" and useful. Nice.

If I implemented FileSystem.fu myself, I may have chosen to include the entire implementation for a specific language in one #if block each, rather than having #if blocks in every method.
(I also wish that fut defined those language-specific defines without manually passing the -D definition)

[mittelmark]

Thank you for the hint of just single large #if blocks for each language. That makes total sense. If I add an other function to the implementation I will add this to an existing block rather than opening new ones. Initially I thought that fut has these language defines automatically but then find out that I have to provide -D C etc on my own.

[pfusik]

I recognize an issue that fut doesn't see the return inside the native block. How about:

native return {
    ...
    return foo;
}

or even:

native (c, cpp) static void Foo() {
    // all C/C++ code
}

But I don't like having more and more native code inside Fusion source code. This has further issues, such as no syntax highlighting. fut is 28 KLOC with no single native block. My second biggest Fusion project, RECOIL, is 23 KLOC and again no native block. Everything is handled by calling Fusion code from native code, or subclassing Fusion code.

Operations such as reading/writing whole files could be added to Fusion API.

[mittelmark]

Sure that looks more comfortable than my code in FileSystem.fu. Having more functionality in Fusion itself avoids writing for such sometimes painful native code blocks. But still Fusion's core should stay small. Having reading and writing whole files certainly is something worth considering to be added. Currently the trade-off is exceptionally good, just a few source files and a small single file executable ... BTW: I can configure my editor to highilight in C++ etc mode within these fusion blocks, however currently the short code fragments are not worth the configuration and they should in principle not be written with too much comfort - as native is always the last option ... ;)