I'm generally for generics, but there are tradeoffs. As soon as they debut, expect a flood of overly abstracted libraries. I'm actually pretty apprehensive because I could easily see the costs outweighing the gains.
Agreed that “excess abstraction entering the Go ecosystem” is probably the main danger, but even still, I’m not too worried. The Go community is all about simplicity, and I don’t see that changing. It’s not going to suddenly become the Enterprise Java community. Libraries with crazy abstract APIs won’t become popular.
I expect it’ll be similar to Python, which supports generics, very powerful reflection, many programming paradigms, etc., but the popular libraries tend to have simple, easy to understand APIs, because that’s just the culture of the community.
This last point isn’t true at all. Lots of popular libraries do weird things. Lots of popular library functions do completely different things based on the types and combinations of provided parameters (Pandas being one of the most egregious offenders, but the stdlib does this all over too) or they overload operators to make a clever, magical DSL (e.g., SQLAlchemy). Or they use metaclasses to programmatically generate APIs (e.g., AWS sdk, SQLAlchemy, etc). I take notice because I use Python professionally and spend so much more time pouring over docs and source code trying to grok APIs than I do with Go.
Yeah, the scientific Python world is a different story, but I’d say most Python libraries are simple to use, including the AWS SDK. Agreed that SQLAlchemy is much tougher to use if you opt in to their DSL, but to be fair that’s true of every SQL lib with a DSL, in every language I’ve used.
Basically, I find that though it’s possible to do all sorts of crazy shit in Python, most popular libraries have simple APIs, because simplicity is part of Python’s culture. There are certainly exceptions though.
Agreed that the scientific community are the worst offenders, but I don't think it's just them. There are examples all over, but I don't think this conversation will benefit from their enumeration. I will point out that you're letting SQLAlchemy off the hook because their DSL is "opt in" and it's par for the course for DSLs; however, their DSL is not really "opt in", it's the standard way for building queries. Further, "DSLs will be DSLs" is kind of missing the point, which is that the Go community rejects these kinds of hacky DSLs while the Python community embraces them.
At the end of the day, I notice a significant difference in wanton magic between the Go and Python communities, and it's big enough to make me prefer Go over Python when given the choice (despite that I've used Python far longer and more regularly).
