Lambda the Ultimate

I wish the language was a little more C++-like and less ML, to ease us C++ programmers. It reminds me of Single Assignment C (SAC), a functional language based on C syntax. I think this C/C++ approach (similar syntax and native module support) will be helpful for industry adoption. Java has been very successful as "C++ minus warts plus GC".

I'll always remember Vyper because it started me down the road of functional programming. I was programming primarily in Python, and stumbled across Vyper, which was supposed to be a more expressive version of Python. Vyper itself never became useful to me, and the author stopped working on it shortly after I discovered it, but it led to other things. Vyper is implemented in OCaml, so that was my next stop. That really introduced me to the functional paradigm, so I continued on to Dylan, Haskell, and others.

I didn't notice this before. I am very much in favour of this distinction (think Ada). I reccall that others here didn't share my enthusiam.

I was one of the others :-) Note that in Felix functions are prohibited from side-effects (except a few selected ones), while procedures are intended for operations with side-effects (otherwise, what is their point ;-) This is a distinctions that languages like Ada don't enforce (disclaimer: I'm no Ada expert, so I could be wrong on this).

You might argue that even the indication to someone reading the code is justification enough, but on that I disagree. Making the distinction between functions and procedures does loose you some functionality or at least flexibility, e.g. related to generics. Code self-documentation just doesn't provide enough compensation for me.

True. The language doesn't enforce it, it just encourages it (for example functions can only have "in" mode parameters).

I know it's not perfect, but I much prefer that procedure not appear in expressiosn. For me that's a more improtant factor then wether the language enforces the no side-effect rule.

With C/C++ 'procedures' are implemented as functions returning a void 'value'. Given the restrictions on expressions with a void type, many C/C++ programmers don't think of them as expressions at all. Indeed, in my experience few C/C++ programmers use these 'procedures' in expressions, at least not explicitly.

The fact that void is a type which can participate in an expression does allow intermixing of functions / procedures in C++ generics. Whether or not this is a good thing is obviously a matter of personal taste.

Also, as I recall, the discussions were about whether x.sort() should sort x or return a sorted copy. And if x.sort() does sort in place what value, if any, should be returned.

As I recall the problem was Python returning null for that operation. For a ST PL this is almost a non-issue.