Evolutionary Programming!

[Jason Jeffory via Digitalmars-d]

On Tuesday, 5 January 2016 at 20:11:10 UTC, Ola Fosheim Grøstad 
wrote:
> On Tuesday, 5 January 2016 at 16:10:21 UTC, Jason Jeffory wrote:
>> It seems that many programmers get dissatisfied with the state 
>> of something and try and branch off and create something that 
>> suits them. The evolution of programming is the evolution of 
>> life in some way.
>>
>> Is it possible that one could develop or modify an existing 
>> programming language that can adapt in such a way to provide 
>> maximum unity between programmers?
>
> In a way that is the ideal for language design. Design a 
> minimal core that can be extended. A few languages are a bit 
> like that: Lisp, Beta, Self, Smalltalk?
>
> But I think the most powerful concept now is _software 
> synthesis_. The basic idea being that you specify the 
> constraints and let the computer write the actual code, or some 
> variation over that.
>
>> A language itself, that can adapt grammatically(since people 
>> like different grammars), be both powerful and simple, and 
>> provide all that is needed(more or less)?
>
> Ok, on the grammar level you have some languages that try. Like 
> Pure, which is based on term rewriting. But the _key_ problem 
> with this in the general case and also with very powerful and 
> flexible type systems is that situations may arise where you 
> either have bugs related to (perceived) ambiguities or that 
> computations become very expensive and takes forever to 
> complete (if at all). Of course, one can solve that with 
> time-out errors and smart heuristics, but there is a design 
> complexity cost here.
>
>> I know many will immediately write off this statement, but I 
>> have to ask: Is it possible for a single language to capture 
>> it all?
>
> If you mean "capture it in all ways that a human will perceive 
> as significantly", then yes, probably. It might be impossible 
> to master, like C++, but such a language could probably be 
> constructed.
>
>> System level,  application level? Why not both?
>
> Ok, but I would rather have a deductive language that can 
> easily express efficient imperative code... I don't know what 
> it would look like, but I think that would be my ideal.
>
>> People will claim that such a language is infeasible because 
>> it is too complex!
>
> I don't think it is infeasible, but I think that the amount of 
> effort it takes to design it should be spent on creating AI 
> that can do boilerplate coding. And let humans do high level 
> application design instead.
>
> My point is this: We all spend a disproportional amount of time 
> writing braindead code. With bugs. Because the human brain 
> isn't very good at pinpoint level accuracy. We are really not 
> very good at explicit imperative programming.
>
> We are much better at modelling approximately with clay, 
> gradually refining the model until it is good enough.
>
> That's the major flaw of computer languages. None of them 
> reflect human creative processes.
>
>> Any more thoughts?
>>
>> Think of all the wasted man hours spent on people working 
>> unfocused on the issue.
>
> Yes, this is horrible. We should rather write more generic 
> language agnostic code, and drop down to specific code where 
> efficiency matter.
>
> Take a look at Github and see how many source code files have 
> functions that implement almost the same function. That's 
> insanely inefficient.
>
> Just take a look at D repositories and we see the same, lots of 
> implementations of essentially the same functionality.
>
> What we need is not libraries but maps of "abstract function 
> implementations". The closest we get to this is StackOverflow. 
> I think that is just the beginning. We could have a language 
> agnostic representation and automatic translation to Java, C#, 
> JavaScript, C, C++, D...
>

Yeah, I'm not sure I understand the StackOverflow thing but it 
would be cool if people didn't waste so much time having to do 
the same boilerplate code. I'm not sure if AI is far enough 
along. Proper AI learning algorithms could potentially help(they 
might suck at first but this is what training/practice is for. 
Eventually they will be better at it than us)


>> But a top down approach to programming(and computers) is 
>> required.
>
> You need both. Top down and bottom up, then software synthesis 
> (let the computer fill in the details).

Yes, I meant that basically programming compilers and designing 
languages seems to be approached current from the bottom up(much 
like it was in the past). We are starting to see top down 
approaches with DSL's, intermediate languages, etc...

>
>> Such a programming language would truly be evolutionary(or at 
>> least revolutionary), right?
>
> Software synthesis is an active research topic. Maybe we'll see 
> it in 30 years? Real world languages lag 20-30 years behind the 
> computer science frontier...

I basically agree with what you said. I think it is a difficult 
problem and I guess it will take time for humanity to figure it 
all out... we are on track. It just feels like there it is very 
inefficient... but like most things, it tends to be exponential.

I suppose the average programmer can't comprehend the 
complexities require for the initial adaptation and it takes time 
for them to "get with the program". Think of when C added ++... 
most programmers were not good at oop... now it seems everyone 
has a pretty solid understanding of it.

Thanks