I really want sum types, strict semantics, and untyped IO, and a good ecosystem.
I mean, I don't necessarily condone it, but if you're into "recipe programming" you can just use the Turtle package/modules and get pretty far. The tutorial is pretty amazing. I'm not saying it's a bad thing, necessarily, but I'm one of those crazy people who thinks that a little bit of cogitation should be (and is!) rewarded. Maybe I'm wrong.
EDIT: Oh, wait, you said "untyped IO". What does that mean, I wonder?
I kind of want Ocaml with a Go sized ecosystem.
(I appreciate that people may have different requirements; mine were and are mostly around my own productivity -- not so much real-time/low-latency/predictable performance/etc.).
I'd like to see a language that compiles to Go, but also to Javascript, so that code can run on both client and server without changes.
That language should not equal Go, because, honestly, some aspects of this language are quite strange: https://golang.org/doc/faq#nil_error
Yet it feels like magics. Why just maps? why not sets, with useful set methods ? (union...) . It doesn't feel consistant in a statically typed language. It also feel like a mistrust of the developer. Go doesn't trust me enough so I can implement my own typesafe containers ? or does Go asks me to do the compiler's job by writing code generation tools ?
And there is no maybe[T] or we wouldn't need multiple return values when dealing with errors. If there was a maybe[T], we could chain functions and only deal with errors in the last return statement.
Having programmed in Go for production a while now, in my experience it's rarely necessary. Sometimes a set is needed, but it's very simple to just use a map[T]struct{}. And if something like union or intersect is needed it's usually less than 20 lines to implement all those operations. Usually that operation is only needed in one instance, and it's very quick to write.
Would it be faster to have a builtin generic set? Absolutely, but it's such a small inconvenience I honestly don't care.
Maybe I'm missing something, but? https://golang.org/src/encoding/xml/example_test.go
https://github.com/golang/protobuf uses this as well.
[0]: http://www.frege-lang.org/
[1]: https://github.com/Frege/frege/wiki/Differences-between-Freg...
It's story of building and dependency management is one of its major areas of growth in the last 2 releases because the developers agree.
But no... I mean... I can get a lot of editors to do a better job of formatting. Other languages I use offer true interactive development. And go build requires a Zigguraut Of Filesystem around it to function properly. Go get provides no realistic versioning other than "this is what I grabbed RIGHT NOW and I hope you don't need bugfixes because good luck with figuring out the hash I got."
The lack of reproducible builds and better tooling for versioning is a pretty damning stroke against the ecosystem, because I nearly lost a project when my HD crashed. I had my repo checked in AND go deps set up and it still took me a solid day of wrangling to recover a project.
For all my kvetching about bower and maven, when I set them up they tend to stay set up and so the right thing. Go projects decay.
Add that to the language's total punt on error handling and sync semantics so crazy that the ecosystem got a basic race detector before a basic interactive execution tool...
I just don't get why this ecosystem has adherents.
Docker does not make anything crossplatform; it just enables you to package your dependencies such that you can easily run on any other linux-amd64 machine. Go applications usually don't have many dependencies so Docker usually doesn't buy you much if anything at all.
>But no... I mean... I can get a lot of editors to do a better job of formatting.
I've never used a language with an ecosystem that so adamantly abided by one standard format. PEP8 comes somewhat close, but there are still fights over different linters etc...
>Go get provides no realistic versioning other than "this is what I grabbed RIGHT NOW
The vendor directory introduced in Go 1.6 (and usable in Go 1.5) has addressed the versioning issue with `go get`. I agree creating/updating vendored dependencies can be tricky depending on the tool you use and having a Go workspace per project is slightly cumbersome (though no less cumbersome than say virtualenv), but even before Go "officially" supported vendoring, going from zero to development is literally just `git clone` and `godep restore`. Go projects do not decay if you've vendored your dependencies.
>so crazy that the ecosystem got a basic race detector before a basic interactive execution tool...
Am I crazy that I prefer having a race detector over a REPL? There were plenty of REPL projects that never gained traction because you can develop pretty easily without them. I usually jump into the playground[0] if I need to test something. This has the added advantage of being sharable after I write it, too.
Which IS the primary use case for nearly all of Go's interesting use cases. The exceptions are CLIs for API access? But you're going to have some trickiness with testing in these cases that warrant a non-trivial build-test cycle anyways. If that works well for you that's wonderful.
A good packaging of Go with some helper shell commands is also quite nice, for other reasons.
> I've never used a language with an ecosystem that so adamantly abided by one standard format. PEP8 comes somewhat close, but there are still fights over different linters etc...
You say this like it's a good thing. I see only the downsides. I'm not much of a fan of Python's contemptuous ecosystem either.
> The vendor directory introduced in Go 1.6 (and usable in Go 1.5) has addressed the versioning issue with `go get`.
It most assuredly has not addressed this issue. All it's done is decreased the confusing filesystem zigguraut so that a single repo checkin doesn't look patently ridiculous. Pretty important when GitHub is the primary code distribution mechanism.
But so long as the maintainers of Go pretend that everyone will always keep master up to date and never make breaking changes, Go projects are still at a disadvantage compared to every other civilized combiled build+deps tool.
I had high hopes for Go vendoring, but examinations in 1.5 really reinforced to me that the maintainers and designers of Go are interested in the (less compliant) real world issues everyone runs into, don't care about shipping reusable code to a wide audience over a long period of time.
Languages with contempt for programmers as a central tenant often have problems like these.
> Go projects do not decay if you've vendored your dependencies.
Hope you don't have a library with bugs. I sure found many in the libraries I pulled and if I put down a server for a month to work on something else, then tried to grab a security update? It was never not a headache.
> Am I crazy that I prefer having a race detector over a REPL?
I only refer to situations and inanimate objects as crazy. Anything else is disrespectful. You awknolwedge interactive development is necessary but don't want first class support for it. I just don't understand why you wouldn't want to actually have something that could integrate into a real workflow. Sharability is something we already can get with trivial pastebin/gist integration.
There are plenty of good third-party options available for dependency management (nice double standard btw since you apparently have no problem with third-party stuff for fmt/vet/lint concerns) and there is also this https://golang.org/cmd/go/#hdr-Vendor_Directories so a better world is coming don't you fret.
go race is pretty much a port of Thread Sanitizer, which was a C tool. Before that, there was Helgrind, also a C tool. All of these tools have a long list of forerunners in academia, largely for Java (Eraser, FastTrack, etc.)
Eraser and FastTrack are un-googleable, if you have any links they would be appreciated.
Vendoring makes life hard for people who have to package software. Traditionally, you have build dependencies which are the libraries your software uses. In Go (because nobody follows semantic versioning when changing the API of their libraries, and the Go tooling doesn't allow you to specify versions of a package -- WHICH IS ACTUALLY CRITICAL) you just have to dump all of your dependencies into the same place as your source and be done with it. I don't see why people see vendoring as the solution. Defining version dependencies in your source is the solution if you want to go with "our source code defines what we need to pull in". Don't get me started on the static linking issue. Oh, and build dependencies have to be packaged as source code (packaging them in their compiled form is quite hard to do).
The 'go' tool is not the compiler. The compiler is 'gc' and does not know anything about dependency management.
It's story of building and dependency management is one of its major areas of growth in the last 2 releases because the developers agree.