Hacker News Reader

Comment by chewxy | Hacker News Reader

I put it to you that most other language communities do not deliberate as much on generics.

Consider the document that Ian Lance Taylor put out: https://github.com/golang/proposal/blob/master/design/15292/...

More research here: https://docs.google.com/document/d/1vrAy9gMpMoS3uaVphB32uVXX...

I have also given it a bit of thinking on the tradeoffs: https://blog.chewxy.com/2017/09/11/tensor-refactor/#some-tho...

The Go community may be stuck on an analysis paralysis mode, but it's much better than ad hoc implementation that will stay forever.

View on HN

You mean the usual documents that get thrown into generics discussions, where he keeps talking about C++ and Java, while ignoring all other programming languages designed since CLU and ML in the 70's?

"In retrospect, we were biased too much by experience with C++ without concepts and Java generics. We would have been well-served to spend more time with CLU and C++ concepts earlier."

https://go.googlesource.com/proposal/+/master/design/go2draf...

This has been acknowledged and addressed. What is your proposal?

I don't know, that they actually implement them, instead of doing proposal after proposal?!?

Even the ideas that eventually die on the Web, C++, Java and .NET happen to have preview implementations available to try out at one point, instead of dead paper.

I don't think it is efficient to implement a dozen of ways to handle generics.

People who need generics will use code generation until the issue is addressed, I don't really see the need to rush.

Attention to details is something I like about go.

Just like C++ MS-DOS compilers used to do before ANSI/ISO started to work on templates, about 30 years ago, talk about progress.

C++ has concepts now. What's the excuse?

There's a proverb in these here parts that goes like "He who doesn't want to kneed dough, keeps sifting flour".

Perhaps a modern dev-friendly version would be, "He who doesn't want to code, keeps configuring their text editor".

That seems to be the case with Generics and the endless "documents" and "analysis" in Go. If you don't particularly want to do something, you keep analyzing different ways of doing it...

Most of these "analyses" especially from Ian has backing code changes.

This for example, is my experiment with trying to give Go a backing calculus that supports generics https://asciinema.org/a/166748

The runtime changes yielded over 20k lines of changes - basically about 1/4th of the Go runtime has to change. Not to mention a large amount of semantics changed too - essentially a different language

>The runtime changes yielded over 20k lines of changes - basically about 1/4th of the Go runtime has to change. Not to mention a large amount of semantics changed too

Well, isn't it a major typesystem change? Makes sense for it to apply to a big part of the runtime and semantics.

There’s another proverb that I think applies better: “measure twice, cut once”.

Having buggered up a half-lap joint the other month even though I measured things several times, I'd like to propose a change to this proverb.

"measure twice, think twice, cut once"

(or even "think twice, cut once" if you want to generalise it to other spheres)

In the Russian version of this proverb you'd need to measure seven times)

Reply on news.ycombinator.com