Stages of learning Go, with code examples
sourcegraph.com
sourcegraph.com
Edit: Somebody is heavily downvoting me on every reply, I would love to know why. I like Go and want to get into it since a few months but the missing package manager held me off. So what is wrong about my question?
In Go, code is packaged via source control repos. If you want a different version, you need a different repo. There are some tools to mitigate this, so that you can get away with a different branch per version (see http://gopkg.in), which can make it a little nicer for the maintainer to deal with multiple branches. To the consumer, it still looks like a different repo (i.e. every different version is a different URL to a repo).
Note that because Go is a compiled language, the only time versioning comes into play is a development time. As long as your developers are using the right version, you never have to worry about versions during deployment.
Now, if you're concerned that your project depends on code repos from external sources, then there are several tools that can insulate you from third parties doing stupid stuff. The most complete and best one is called godep, which basically has two modes - reliable and paranoid. The reliable mode insulates you from changes made to a repo you depend on by recording the exact VCS revision your project is using and always checks out that commit. That way, if an upstream repo checks in a breaking change, it won't affect you (and you control when you update to newer commits in that repo). Paranoid mode actually copies all the code from your dependencies into your own repo. This insulates you from upstream repos getting deleted. And of course, this gives you the ability to test out upstream changes before integrating them into your project. The tool makes it easy to copy all of your dependencies at once with a single command, so it's not a huge hassle to do (which it would be if you had to do it by hand).
"You really don't need X for Go" is becoming the Go community catchphrase. And they're right: I won't need anything for Go because I wouldn't use a language whose community actively resists improvement.
I could respect that Go doesn't have a package manager, generics, etc., if the Go community took more of a "if you want it, develop it" approach. But instead, the core development team has made it unclear whether they would even accept some of these features. I get that compilation performance is a high priority, but they've missed the big picture, that developer time also matters.
The following exemplifies the problem:
> Note that because Go is a compiled language, the only time versioning comes into play is a development time.
Oh, that won't affect me at all; development time is only like 90% of the life of a project.
The community definitely does resist people who aren't even users loudly declaring exactly what it is that Go needs. I fear the community may have instinctively built up certain callouses that I'm not a big fan of, such as the near inability to have a sensible discussion about generics within the community due to the near inability to have a sensible discussion about generics outside of the community. It isn't all necessarily hunky-dory.
But if you haven't used Go, which is what it sounds like to me, you don't understand the current situation's cost/benefit analysis and are unlikely to be able to "just" "solve" the community's "problems" without such an understanding.
Step 1 of any engineering effort is identify the problem, and it can never be skipped, no matter how sure you are you can this time. If you're not a Go user, you've skipped this step.
Of course you're free to A: not be a Go user for any reason you choose and B: say whatever you like about Go from the outside. And I really mean that, quite deeply, not merely as a rhetorical flourish. There's too many languages to become expert in them all and we all must choose carefully according to our own knowledge and desires. My only point here is only that you shouldn't expect the community to be enthralled with the resulting commentary.
So what is it that you think people who want Go to have generics don't understand? You say that people should acquire some experience before criticizing Go, but you're only accepting a certain kind of experience. I've written tens of thousands of lines of code in similar languages without generics, and tens of thousands of lines of code in similar languages with generics, and there is no question which I would rather write. I have written a number of compilers and am well aware of the concerns and how to implement a performant compiler. Why is that experience not something you value?
You're basically creating a chicken and egg problem: people who like Go will use it, and people who don't like Go won't, and what you're saying is that the latter won't be allowed to criticize Go. But people who don't like Go are the only ones who are going to ever criticize Go. And criticism is what anything needs to not suck.
Don't talk to me about "acquiring experience" as an excuse to discard any experience that comes from outside your community. My core complaint about Go: the authors simply have ignored decades of compiler research. By deciding that only users of Go can criticize Go, you're deciding to only take feedback from people who would use Go, and the only people who would use Go are people who aren't aware of the last few decades of compiler research: people who think coroutines are a new idea and aren't sure whether generics are useful.
As an outsider to the community, I wouldn't care if Go is crappy, except that I work in this industry, and if Go becomes popular enough, a situation will arise where I have to use Go or a language that imitates Go, or turn down a job because of it. Go being terrible and popular means that it affects the whole software industry negatively.
Also, arguing that I need more Go experience to criticize Go is somewhat ridiculous in response to an article which says, "Being an advocate is not really a separate stage, but a role you can fill at any of the other stages." So you need Go experience to criticize Go, but you don't need Go experience to advocate for it? This is why I assert that Go's popularity is mostly hype.
This describes my feeling when going through the thread and reading that significant weaknesses of Go are not significant.
You can choose who to believe - someone who has tried the language, or someone who merely thinks they know what it'll be like. I know whose opinion I'd trust.
First of all, I have written Go code, I just wouldn't call myself an expert. And lack of generics becomes a pain in the ass rather quickly. I decided to come back when they had added generics, but it turns out that isn't happening.
But even if I hadn't ever written a line of Go, I don't think that would invalidate my opinion, because I've used other languages that makes the same mistakes. How many languages do I have to try out before my opinion on language features would be worth considering to you?
Your logic is like that of a child who wants to touch a hot stove. "Have you ever touched this hot stove? No? Then how do you know it will burn?" I've touched enough hot stoves in my life, thanks.
You also haven't responded to anything I said in my previous post, you're just reiterating what you said previously.
> the only people who would use Go are people who aren't aware of the last few decades of compiler research: people who think coroutines are a new idea and aren't sure whether generics are useful
This is simply untrue. Most developers I know that like Go have written tens of thousands of lines of code in other languages that do have generics and more advanced features. You seem to assume that because people disagree with you, that they must just be ignorant. How about if people disagree with you because we have different tastes in languages? Some people love LISPs, they drive me crazy. Some people love Python, I think it's merely ok. ... I love Go, and you hate it. That's ok. I don't think everyone has to like Go. But I don't tell the LISP people they should drop all this paren BS. I don't tell the Python people that they should add braces and static typing to their language.
> what is it that you think people who want Go to have generics don't understand?
...that generics are not needed in many of the cases where they're used. That YAGNI holds true a lot more often than they think. I had actually started coming to this conclusion (about generics) before I even started writing Go - and I'd already started using them less. Not having foo<bar<baz>> in my code simplified it a lot, and given that most of the time I only ever implemented one version... it just wasn't worth defining a whole bunch of generic logic, when really I only had one concrete implementation that I'd ever use.
> My core complaint about Go: the authors simply have ignored decades of compiler research.
Not ignored, so much as "considered and discarded".
You say you've written Go and the lack of generics was a problem rather quickly. I have been writing Go for almost 2 years full time and 9 months on the side before that. Generics have almost never been a problem for me.
I see two possible reasons why we've had different experiences with Go:
Either you happened to be working on a project that really benefits from generics (something that makes heavy and repeated use of trees or vector/matrix math, etc)... or you were trying to write Go code the way you'd write in some language that does have generics.
If the former, well, that's truly unfortunate. Go doesn't work well for all projects. I wish it were better at that sort of thing, and maybe at some point we'll get some features that'll help out with those pain points. Maybe those features will be something like Generics... but I, for one, hope that it's not just generics like everyone else has generics, because I've been there, and I don't like it. There's a hell of a lot of code that doesn't need matrix math and left leaning red black trees and such. Go is good for all those projects.
If the latter, well, then you've really just shot yourself in the foot. This is like taking a minivan offroading in mud and then getting mad that it handles poorly. You have to use the language in the way it was designed. Yes, this means you probably have to write a min function every once in a while. Yes, this means you have to write a loop to test if a string is in a slice of strings. Just like you can't take the shortcut through the mud in the minivan. Just like you have to write type classes in haskell, or getters and setters in Java. Every language has its quirks.
All that being said, it's also completely valid fot you to say you just don't like Go. That's fine. I don't like a lot of languages, even though many other people do. But telling other people they're wrong or dumb for liking Go is not a position you can really justify. It's a matter of opinion, and opinions can't be wrong.
So is it simpler to have all these little functions doing basically the same thing with the chance of bugs or 1 generic implementation that works for all of them?
Looking at things in this light bug potential in Go is O(n) whereas a language with generics is O(1).
The point being that without generics you end up implementing lots of these small similar functions and making lots of potential sources of bugs.
Idiomatic Go encourages writing easy and simple to fix code, which is good. However what I get for all this is go allows me to write the simple and easy to fix code that could potentially have bugs everywhere.
Hope this makes sense, haven't used voice typing for anything this large yet.
I do understand where you're coming from. More code means more code that can have bugs, I totally agree.
However, more complicated code also means more bugs, and generic code is definitely more complicated and non-generic code (in all but the most simplistic cases). So, I don't think it's necessarily as straight forward as O(n) vs O(1).... and also for small values of n, O(n) is pretty close to O(1)... and my contention is that n is almost always small.
So, I don't think you're wrong, per se, I just think that in real life, it's not so black and white, and in my experience, the kind of code I have to duplicate is the dumb stuff, like min, which is nearly impossible to screw up, and if you do, it's immediately obvious in any kind of basic test of the surrounding code.
In Haskell at least this is usually not correct. A sorting algorithm with type
Ord a => [a] -> [a]
[Int] -> [Int]
Perhaps you meant something different by "generic" though. There are a few subtly different uses of that word. One of them means a very neat and powerful generality as seen is Haskell's parametric polymophism. Another is a convoluted run-time tag-based dispatch mechanism which is often under-specified. I can imagine that programmers who have only ever seen "generic" code in the Python object oriented style (to take one example) think this style of "generic" code is more messy than "specific" code.
I had to correct a few things, but I was pretty impressed by the accuracy :D
> I do understand where you're coming from. More code means more code that can have bugs, I totally agree.
Right, I find it easier to use very generic functions and compose them together. Go achieves some of this with interfaces, but my rub with that is that they aren't type safe IIRC. Which my experience has taught me is quite valuable.
> However, more complicated code also means more bugs,
No arguments here!
> and generic code is definitely more complicated and non-generic code (in all but the most simplistic cases).
Can we test this right now? Have any examples in mind so we could figure this out in code?
> So, I don't think it's necessarily as straight forward as O(n) vs O(1)....
I agree, I didn't mean for my statements to be taken as absolutes or perfectly accurate but more as a lens to look through.
> and also for small values of n, O(n) is pretty close to O(1)... and my contention is that n is almost always small.
I agree here, and this will lead us to the source of our (and many others disagreements).
I feel that when as much code is as generic as possible, I can make more generic code and lessen the potential points of failure.
Forgive the projection, but I feel your thoughts are more along the lines of "Really generic code isn't simple and simple code is the key to fast, easy to maintain, and well... simple code".
I was a big fan of Go in the past and thought along the same lines. I gave Haskell a try and things being very generic didn't seem to make complexity go up much to me.
This means it feels like a lot of Go proponents just dismiss Ocaml, Haskell, Scala, Clojure, etc as something convoluted and overly complex from the outside, but having spent some time in getting to know the language it seems just as simple or more simple than Go (although VERY different) in many areas (not deployment (Go does very good here) and I suspect not reasoning about performance but my use cases haven't necessitated enough profiling Haskell code to say for sure).
Just a random question, what do you and (if you don't mind guessing a little) other Gophers think of Nim? If I were to pick a dead simple imperative language along the lines of Go that didn't "ignore programming language research" I would pick Nim over Go unless there was a huge deficit in libraries for my use case.
If people disagree with me on things that are pretty solidly proven by my experience, then yes.
> How about if people disagree with you because we have different tastes in languages?
I mean, if you're argument basically comes down to, "That's just, like, your opinion, man", then I'd counter that its pretty objectively true that people are more productive in some language than others.
> ...that generics are not needed in many of the cases where they're used. That YAGNI holds true a lot more often than they think. I had actually started coming to this conclusion (about generics) before I even started writing Go - and I'd already started using them less. Not having foo<bar<baz>> in my code simplified it a lot, and given that most of the time I only ever implemented one version... it just wasn't worth defining a whole bunch of generic logic, when really I only had one concrete implementation that I'd ever use.
So your position is that we should take away features that people overuse?
There are tons of places where you just need a few functions, not a class with a bunch of methods on it. How far are you going to take this logic?
If you want to be babysat by your language, write Java.
> You say you've written Go and the lack of generics was a problem rather quickly. I have been writing Go for almost 2 years full time and 9 months on the side before that. Generics have almost never been a problem for me.
Well, there are a few explanations for that:
1. You've worked in a very narrow domain.
2. You're willfully pretending the problem doesn't exist so you can feel good about your chosen language.
3. You're completely not seeing the numerous cases where your code has duplication which could be removed using generics.
> You have to use the language in the way it was designed.
On the contrary, if the language is designed poorly I do not have to use it at all.
> It's a matter of opinion, and opinions can't be wrong.
So, lacking justification, you just segment off what you're saying into the realm of opinion, where if you believe hard enough anything you want can be true!
EDIT: A funny thing is that the Go core project itself is apparently not one of those projects you claim exist that don't benefit from generics. See here[1] about halfway down the page:
> A weakness of Go is that any generic-type operations must be provided by the run-time. Some day that may change, but for now, to make working with slices easier, Go provides a built-in generic append function. It works the same as our int slice version, but for any slice type.
But sure, that will never come in handy in most projects.
> If people disagree with me on things that are pretty solidly proven by my experience, then yes.
I guess I just am aware that others may have had different experiences than I have, and unless their opinion is way out in left field (like, "programming in assembly is just as productive as python")... then I assume they're not ignorant. People think Ruby is fantastic and I think it's pretty horrible, but I don't think they're ignorant of other languages, they've just developed different tastes.
> So your position is that we should take away features that people overuse?
Some of them, yes. Go doesn't have the ternary operator. I consider that a good thing. I've seen critical security bugs that had been in production code for years because someone misused the ternary operator in a way that never would have happened with a boring old if/else.
> So, lacking justification, you just segment off what you're saying into the realm of opinion, where if you believe hard enough anything you want can be true!
I'm saying that what makes a language good or bad is generally subjective. I don't think you can objectively say one language is better than any other except in extreme cases. Is Haskell better than Python? Is C# better than OCaml? It's like asking if minvans are better than sports cars... it depends on what you want to do with them and what you enjoy. I used to like to floor it getting on the highway and hearing the engine rev as I was pushed back into the seat of my v6-powered sedan... now I take great satisfaction in throwing a new couch in the bed of my truck and not having to wait or pay for delivery. Is my truck better than my sedan was? Would my truck be better if it had the same 0-60 time as my old car? Sure... it would also cost like $15k more. There are always tradeoffs.
You say the Go community is being blind to the benefits of generics, and we say you're being blind to the costs of generics. And I don't mean compiler complexity or speed. I don't give a fig about the compiler. I care about the complexity of the code. I care about error messages that make you open up 4 different files to try to figure out what combination of code is actually being run in a single place.
I don't think we're going to agree here, and that's fine.
And then it proceeds to ignore these improvements anyway because it has a dogmatic view on certain features and the credentials of the person suggesting the improvements are absolutely irrelevant.
There's no language you couldn't construct a similar argument for, is there? Except maybe Perl. There's always going to be some capability favored by some loud person who's going to publicly disqualify some language for not having that feature.
Weird that in this case it's an equivalent of npm or Rubygems, which are nice tools for beginners but a complete pain in the ass for serious projects; essentially, you're disqualifying Golang for not reinventing DLL Hell like everyone else.
Generics, I get. Package management?
I'd be interested in hearing more of your perspective here, because mine is the exact opposite. Newbie writing some small scripts? Don't worry about RubyGems. Writing a real project without Bundler? I think I'll pass...
> Weird that in this case it's an equivalent of npm or Rubygems, which are nice tools for beginners but a complete pain in the ass for serious projects; essentially, you're disqualifying Golang for not reinventing DLL Hell like everyone else.
npm and RubyGems have two basic problems:
1. Curation of package sources. 2. People installing packages they shouldn't have.
Curation of package sources: notice that people don't complain nearly as much about dpkg/apt packages: they're curated much more carefully than npm/RubyGems packages (which is to say, they are curated).
People installing packages they shouldn't have: often the problems people run into with package management are self-inflicted. The adage is old: "Libraries from CPAN should not be treated as if they were part of the core." "Everyone else", as you say, is not experiencing DLL hell. I have zero difficulty with package management, because I use mature libraries and specify versions in my package configurations. There's no excuse to be experiencing DLL hell these days. The people who experience DLL hell are the ones who don't check in package configurations and then are surprised when the 0.1 alpha package they installed makes a breaking change and they don't know which version their code works with.
This is a problem in the Ruby community in particular because part of the draw of Ruby is "there's a package for everything". Newbie devs get caught up in the ecstatic frenzy of solving all their problems with gems without researching or testing those gems at all. But that's not a problem solved by Go's lack of a package manager, it's a problem solved by looking more carefully at what dependencies you introduce before introducing them.
I think the bit about source versus binary deployment is a bit of a red herring. It's not like any true Scotsman is deploying by running pip update in production either. As a developer you want to have code that compiles at all times. Having a binary you can't reproduce or change isn't going to be much comfort.
If you're going to use third party libraries there is no reliable way to judge the stability of either the libraries you use directly or all your transitive dependencies. Pull in enough dependencies and the chance of a conflict approaches one. Some of your dependencies will get updated and some of them won't, and soon enough you'll have a hard time finding any combination of versions that will compile.
Dependency management is not optional, but unfortunately Go makes it easy to hurt yourself by providing climbing holds without a safety harness. It isn't obvious to new programmers that they should use godep before they come back to an older project and it won't compile anymore.
I'm not sure what you mean by "conflict". That's the nice thing about the way go packages work. There's no v1 of a package and v2 of a package. All packages are different. If one dependency uses foo.v1 and one uses foo.v2 ... they don't conflict. They're as different to go as foo.v1 and bar.v1.
>Some of your dependencies will get updated and some of them won't, and soon enough you'll have a hard time finding any combination of versions that will compile.
Again, this just doesn't happen. A dependency has one canonical location. If the code changes at that location, it changes for anyone that uses that code. Yes, it's possible this could break only part of your code, if only part of your code was relying on the old behavior/API... but that should be rare (and if it happens, it means you can't trust that third party to stay stable).
I think most people who complain about Go lacking versioning of packages are still stuck in the mindset of pip, npm, etc, where versions are fungible, and your code doesn't know exactly what version it might be working against. This is just not the way Go works. There's one and only one version of a dependency that your project sees. Anyone using that dependency uses the same version. If the dependency's code compiles on its own, it'll compile on your machine, too.
There is nothing in Go that prevents people from making breaking changes to a package - and therefore breaking changes happen. I'll trust my real-world observations of broken dependencies over your assertion that they can't happen.
My point is if you're not tracking the version of your dependencies, it can be hard to even know what versions to go back to to get something that compiles. It may not be enough to just roll back the package with the breaking change, since you may have other dependencies that have been updated to depend on the new version.
The only way to reliably be able to compile your program in the future is to record the exact versions of all your dependencies.
GOROOT/
pkg
src/
lib1
app1
app2
Also, on whole this page https://golang.org/cmd/go/ I can't find ANY info, about how to fetch exact version from repo, not the master branch. In Go world all master branches are for stable code only, not for development? And all tools have only one version?
There's no way to use "go get" to get anything other than the head of the master branch... at least, not by default. If you use a service like http://gopkg.in then you can encode a version in the url/import path which can then redirect to a branch in github, so for example import "gopkg.in/natefinch/lumberjack.v2" will get the v2 branch of github.com/natefinch/lumberjack (gopkg.in does this with some git magic)
I don't agree with this part (or we are talking about different things), but thanks for the second part of reply. At least something :)
If your build process downloads anything from the Internet, your build environment can change on you. Networks can always fail. At best you can detect this occurring and have your build process scream and die, but it can't be fixed.
Note the complete lack of the word "Go" in the previous paragraph. If other languages provide package managers that make it easy to download code from the Internet during your build process, they are tempting you to do wrong things. You should be in personal possession of everything you need to do a full build, whatever that may be (source or otherwise). And, again, note the complete lack of the word "Go" in that sentence.
This has nothing to do with Go. If you are depending on package managers to download your code from external sites during a build, stop reading this right now and GO FIX THAT. Seriously. Someday it will bite you, hard.
And may Turing & Knuth help you if you're downloading packages with unpinned versions that can just upgrade to whatever they want, whenever they want.
Now I will actually use "Go" in a sentence. The fact that Go expects you to vendor your source I do not consider a problem with the language. The fact that it fails to ship with a blessed tool for doing so is a potentially legitimate criticism, but the fact that it fails to ship with a tool that begs you to do the easy, wrong thing is a feature, not a bug. I wouldn't be surprised that they eventually bless something out of the community. Godep seems to be the current leader. Personally I can't recommend or anti-recommend any of them, I've just been vendoring by hand since I'm only using a handful of libraries, and the cost/benefit hasn't worked out in favor of learning something to save me mere minutes of effort. I know where to find solutions if that ratio ever changes.
Should every library author vendor their dependencies separately too?
"Should every library author vendor their dependencies separately too?"
Whatever it takes to produce your end-artifact should be under the control of the relevant organization. Again, let me emphasize, this has nothing to do with Go, and you violate it at your sole and singular peril. When the $PACKAGE site goes down due to DOS, security penetration, or sheer time-based neglect, being angry at me won't help you any.
So, all you're really automating is me going to look at the package's homepage and seeing if there's a new version. I'd still have to read the changelist to see if the bug that concerns me is fixed, what other changes are included, and run all my tests (and possibly write new ones). This is automating the easiest part of upgrading a dependency.
When you pick up the maintenance of an old application you have to reinstall the dependencies from scratch if you aren't using vendoring or Bundler. Each method has their advantages and disadvantages, but both are much better than praying that the library author hasn't introduced a breaking change in a release since the library was last used.
Although it isn't the same when you are deploying Go applications, development has exactly the same issues as Ruby without locking the dependencies.
Or look at [cabal](https://www.haskell.org/cabal/) for Haskell. It's a package manager that doubles as build system.
It helps when you use external open source projects and want to update it to a new version. With go you have to keep track of you external dependencies and their versions on your own.
dub for D, http://code.dlang.org/
cargo for Rust, https://crates.io/
biicode for C and C++, https://www.biicode.com/
NuGet for C and C++, http://docs.nuget.org/consume/support-for-native-projects
OPAM for OCaml, https://opam.ocaml.org/
Not sure if I would agree with that. A full-fledged package manager automates the process of installing, upgrading, configuring and removing software packages in a consistent manner. Dependency management and semantic versioning is an important part to provide this consistency. Wether you deploy source code or binaries does not make any difference why a package manager is always good to have.
Specially in commercial software.
That sounds like deb(dpgk)/rpm's job. In golang-land I just need to use vendoring+godeps for repeatable builds.
I've written up a tutorial about it using git-subtree:
We're having lots of fun with this with Prometheus, quite a bit of work is ongoing to get it into Debian due to this for example.
It does completely sidestep version dependencies hell and encourages you to think carefully about which dependencies you package. It does that by making you manually responsible for keeping dependencies up to date if you distribute other people's code with your own.
So, in answer to your original question, no go doesn't finally have a package manager with proper versioning (at least no official one). In practice, this is not as big an issue as you might think if you come from a language with such a package manager. In practice, I haven't really found it as much of a burden as I thought I might.
For example, when I released v1 of my package "lumberjack" you imported it with
import "github.com/natefinch/lumberjack"
import "gopkg.in/natefinch/lumberjack.v2"
There are multiple ways to solve these problems (vendoring is one, your solution another, godep another path), but the go get tool doesn't explicitly allow for versioning in the way the parent requested and there is no official solution to this.
The official solution to versioning is separate URLs. If you mean something more complicated like automatic resolution of different version specifications (e.g. ver >= 1.2), then no, nothing exists like that. But generally all you need is "project A uses version foo, and project B uses version bar", and Go supports that just fine.
That goes for most code at some point in time; as it evolves the behaviour changes in subtle ways. I do think the go team's insistence that packages should never break their API and remain effectively versionless is somewhat utopian.
The official solution to versioning is separate URLs.
I remember reading an off-hand comment from someone on the go team that you could just rename your package, however I don't think that's a very good idea, and this doesn't play nicely with many popular hosts like github. They have support for version tags on repos in go get, but only for golang versions so it is unused, which is a shame. So there's no official tooling for it. At present the majority of go code I've seen simply imports dependencies from github or similar without versioning and hopes for the best. Unless the other pkg author wants to version, that's all you can do unless you want to vendor (for large dependencies like net/html that can be impractical).
If you mean something more complicated like automatic resolution of different version specifications (e.g. ver >= 1.2), then no, nothing exists like that.
That's what the parent was asking for I think, and more than that something which applies to all pkgs. Personally I've chosen to minimise external dependencies, and vendor them in version control. It's simple, and it pus me in control of updating and versioning for code I use, not the package authors.
Vendoring your deps is currently the only certain way to get stable versioning.
It's not really a package manager, though you can get some of the same results. It's more pulling in exact build dependencies.
Even the Fedora packagers for golang apps are asking to relax the requirement on unbundling vendored things for golang as there is no sane way to handle versions other than git commit hashes.
PS: I'm no go developer either. But the way it pulls in requirements was a nice idea in my eyes.
Idiomatic Java was the FactoryProxySingleton EE-wank fest from Apache, SUN, Spring and co.
It was by changing those idioms around (e.g. how the Play framework wrote Java), that saner alternatives emerged.
The de-facto idiom of PHP before 2010 was messy code.
It started getting more coherent after that, but the preffered idiom now (e.g. for Zend, Synmphony, Laravel) is still a tedious, verbose, copy of the Java circa 2006 style (large class hierarchies, singletons, "Facades", DI, etc).
Idiomatic JS pre-Cockford was ad-hoc BS. After Cockford it got a lot better, but then it changed again (can't say for the worse) with all sort of functional tricks and patterns that weren't in vogue before.
In all of those cases, if they have stuck to the same "idiomatic" way from the start, it would have been worse.
Within a single project, sure, enforce some idiomatic way of doing things.
But within a whole language people should be able to experiment and not be called upon for being "unidiomatic" all the time. This is exceptially disturbing and often in Go, where there's some cargo cult in many advocates that they have found the be-all end-all way to code.
It seems to me this push for "idiomatic" is also related to the Blub paradox. Blub programmers know a couple of ways of doing things (the "idioms" of their language) and cannot understand why someone might want to program with higher (or just different but convenient) concepts he learned in another language.
Let me be the realist here : it isn't.
Go is just a very young language that, for the moment, only has developers that are curious and therefore either good or willing to learn. This has the additional consequence that for the moment the mailinglist is very helpful. History teaches us that both effects have nothing to do with the language and are temporary (unless the language never sees any "real" acceptance, like Haskell).
You can look messages from some years back showing the same sentiment for Delphi, Visual Basic, Java, Python (especially Python had lots of people making this argument), ... And when those messages were posted, they were true. It didn't last long for any of them.
> But people would struggle without Visual Studio debugging and a big run button.
Sadly this is being used as an excuse for making language with substandard (to put it mildly) debugging facilities and editors.
And of course, like for every other argument, the Go team itself is just being extremely arrogant about this. "Why do you want this ?" type responses, or just dead silence, or outright declaring the people asking the question to be substandard developers or some such.
That doesn't mean that idiomatic practices are always virtuous, but non-idiomatic code is only virtuous because it overcomes an inherent mismatch between what the user wants and what the language provides. With many languages (certainly js, java, c++) we see non-idiomatic practices guide changes to the language and then be replaced. Java gained functions to avoid the abuse of anonymous inner classes. ES6 is stealing from transpiled javascript languages left and right.
Making idiomatic code virtuous is the goal of creators and maintainers. Writing idiomatic code means that you have chosen languages, frameworks, and libraries that are closely aligned with the code you need to write and the style you want to write it in.
I can't say for sure, but I'm guessing it probably came from years of experience working on the sorts of projects that Go was designed to address. Go is really geared toward large distributed projects with a lot of contributors. In that context, it makes sense to value some sort of conformity. It may not always result in the most efficient code, but it makes the code much more maintainable and accessible to newcomers. Once you're familiar with idioms you don't need to spend so much time figuring out what a new piece of code is trying to accomplish.
You seem to argue that since old PHP and javascript code had terrible coding practices, therefore Go shouldn't try to enforce any particular type of practice. I don't see that your conclusion follows from that premise. Now if you're arguing that good coding practices arise from the community, that's well and good. Idioms can change and I expect that idiomatic Go code in 2020 will be different from idiomatic Go code today. That doesn't mean there isn't value to writing idiomatic Go code today.
I haven't actually seen any cargo cultists explicitly say that no one should be writing in other programming languages for any project. That would be a pretty stupid sentiment, although I shouldn't be too surprised if some people felt that way. Go's strength is really on server side code where storage is cheap and network and I/O latency is a bigger issue than CPU or memory utilization. If you're developing for that kind of environment, there's a lot to be said for using Go. And while you're at it, you'll save yourself and anyone else who's working with your code a lot of hassle if you follow best practices and write idiomatic code most of the time. If you're in a specific situation where you can get better performance or whatever by deviating from the standards, then go for it, but leave some comments explaining your choices to anyone else who has to maintain your code.
If I'm going to write a one-off script to process a million lines of text, I'm probably going to write it in Perl, not Go (and yeah I know it's 2015). My Perl code will follow TMTOWTDI whatever-the-heck-works practices. If I'm already writing Go for something, and others might want to peek at my code, then I will make an effort to make my code idiomatic.
Because being able to communicate is a virtue. That is all idioms are, a way to improve communication. By being well understood in a given community or sub-set of a community, they create a lot of value and a type of short-hand for understanding code.
Idioms are not static unchanging things -- they are flexible, adaptable, constantly changing tools that assist in understanding language (both spoken and programmed).
> Idiomatic Java was the FactoryProxySingleton EE-wank fest from Apache, SUN, Spring and co. It was by changing those idioms around (e.g. how the Play framework wrote Java), that saner alternatives emerged.
Seems like your problem is with bad idioms, you seem to be a fan of the Play style -- both are just idioms, a simple way to improve communication.
> The de-facto idiom of PHP before 2010 was messy code.
That isn't even an idiom. An idiom is a well understood part of a language and community. "Messy code" isn't an idiom, it is just messy code. There is no value gained from it, it has no short-hand value and it is not well understood across a community.
> It started getting more coherent after that, but the preffered idiom now (e.g. for Zend, Synmphony, Laravel) is still a tedious, verbose...
But well understood by those in the space, a Laravel person can see another Laravel persons code and "get it" very quickly because they speak in the same idioms. It might be ugly to you, but it is useful to the community. Rails has its idioms, Django as well...
> Idiomatic JS pre-Cockford was ad-hoc BS. After Cockford it got a lot better, but then it changed again (can't say for the worse) with all sort of functional tricks and patterns that weren't in vogue before.
Idioms are about communities more than languages, JS toolkits vary rather wildly in what they consider idiomatic.
> In all of those cases, if they have stuck to the same "idiomatic" way from the start, it would have been worse.
Again, idioms are about what those native in the language or toolkit understand and can expect others to understand, they have never been (nor are intended to be) static, they are a reflection of what a community understands. They are shorthand.
> But within a whole language people should be able to experiment and not be called upon for being "unidiomatic" all the time. This is exceptially disturbing and often in Go, where there's some cargo cult in many advocates that they have found the be-all end-all way to code.
Go has a very strong core set of idioms that Go developers understand and expect. Breaking them has a significant cost to all future readers of the code and should be done only if it adds value in some ways. But a lot of questions in Go are far from settled, error returns as interfaces, concrete types or strings. I would be interested in explicit examples rather than vague cargo cult claims.
> It seems to me this push for "idiomatic" is also related to the Blub paradox. Blub programmers know a couple of ways of doing things (the "idioms" of their language) and cannot understand why someone might want to program with higher (or just different but convenient) concepts he learned in another language.
"different but convenient" can to be looked at in two different ways. The first is for you, the original developer. It is more convenient for you... but what about the next 5 (or 50) developers who are unfamiliar with the idioms in play, it will radically slow them down as they lack the shorthand to quickly understand it. Go is about programming at scale (in terms of people) and something that benefits one while slowing down all others is going to be looked down on. Of course, there are valid reasons, it is faster, it is lighter, it is X. But when you diverge from the norm, you better be willing to explain it in a comment, both the what and the why.
No, my problem, is with the pre-supposed and agreed upon set of idioms ("idiomatic Go", "idiomatic Java", "idiomatic Python").
I'm not even against what those idioms contain per se, as individual items.
>"different but convenient" can to be looked at in two different ways. The first is for you, the original developer. It is more convenient for you... but what about the next 5 (or 50) developers who are unfamiliar with the idioms in play, it will radically slow them down as they lack the shorthand to quickly understand it.
Well, the reverse can be true.
Java programmers familiar with the circa-2004 idiomatic BS over-reliance of GoF patterns and deep class hierarchies created code that was "idiomatc Java EE" but difficult to undertand, overengineered, and inflexible.
Someone coding unidiomatic Java (e.g. Play) and giving the project to them, they'd instantly understand what its going on -- even better than someone giving them an "idiomatic Java EE" monstrocity.
>Go is about programming at scale (in terms of people) and something that benefits one while slowing down all others is going to be looked down on
People keep saying that, but most Go projects I've seen are made by only few of people, maybe even a couple. Plus, people complain about even single-developer Go projects being "unidiomatic".
Programming at scale, like big 3D games, Photoshop, the Linux Kernel, Webkit etc, is still done in C/C++, with no sign of this changing.
Idioms have to be agreed upon (by the community) else they have no value at all. Shorthand only works if both people in the conversation understand what it means.
It seems like you are upset that the idioms came from the standard libraries (which is where I believe the initial idioms for Go, Java, and Python stemmed from), but you also have to give time.
Play was released in 2007 -- Java was released 1995. So it took a number of years for that change to take place. During the between time, idioms helped move java forward, make it understandable, make developers able to communicate via code.
Go was released in 2012 (v1), so of course a lot of the idioms are going to stem from the standard library -- we don't have two decades of Gophers moving the ball forward and agreeing as a community what the idioms are, they gotta start somewhere.
> Java programmers familiar with the circa-2004 idiomatic BS over-reliance of GoF patterns and deep class hierarchies created code that was "idiomatc Java EE" but difficult to undertand, overengineered, and inflexible.
Idioms are not static. Please stop using that as a strawman to tear down. Idioms rot. If they are not what the community currently uses, they are not idiomatic.
> People keep saying that, but most Go projects I've seen are made by only few of people, maybe even a couple. Plus, people complain about even single-developer Go projects being "unidiomatic".
Current use and design goals are not the same. Google developed Go to be used at (many-people) scale, even if it currently isn't (which I don't agree with: https://code.google.com/p/go-wiki/wiki/GoUsers). It isn't like Go can stop individuals from using it.
I'm not annoyed by idioms because they don't ever change -- but because while they exist in a certain form, they inhibit experimenting with different approaches.
Or to be more presize, I'm not annoyed by idioms themselves (as I also wrote above), as individual practices. I'm annoyed by people insisting on them, and criticizing everything diverting (unidiomatic) as something that's bad.
Nothing is static, except change. But for a specific period of time (which can be like aeons in the software industry) idioms are for all intends and purposes static.
E.g Java EE land, from say, 2000-ish to 2007. The idioms were fixed and agreed upon, and most people and groups, from Apache Struts to Java engineers and IBM played by them, but they neither helped the community "communicate via code" nor helped "make programs understandable".
Instead, they hindered both communication and understanding. Almost everybody agrees now that the adopted idioms at the time were bad, and inhibited progress and even slowed down projects.
It's not even like "it was good for it's time, but we eventually came up with something better". We always had the option to do the better thing (it doesn't take unique new research to not use a 20-levels deep hierarchy or to not think GoF patterns should be applied everywhere). And since we moved on, everybody seems to agree it wasn't "good for the time" either.
I see you all the time in these go threads, by the way. You would better spend your time forgetting about go and doing something else since you so clearly detest it for not having generics (obviously less convenient than a functional language, but clearly not required to do programming).
In what way is Go code composable?
There is countless proof about Go's composability through functions and interface, but I keep coming back to that one document.
I can't tell if you're serious or sarcastic yourself. The article proposes this snippet, presumably copy/pasted for every option:
// Verbosity sets Foo's verbosity level to v.
func Verbosity(v int) option {
return func(f *Foo) option {
previous := f.verbosity
f.verbosity = v
return Verbosity(previous)
}
}
I direct your attention to sort: https://code.google.com/p/go/source/browse/src/sort/sort.go#... This defines sort on []int and []float64. If you want to sort a []int32, you have to manually define Int32Slice and copy paste lines 232-237. Copy paste the same lines again for any other slice of built-ins you wish to sort, e.g. uint32, uint64, int64, float32, etc.
They removed the entire set of vector containers now, but for the canonical way to implement IntVector, see https://code.google.com/p/go/source/browse/src/pkg/container... -- and compare that to stringvector.go in the same directory. It's almost entirely copy/paste with string substitution to replace int with string.
To be fair, there is "go generate" now that can generate functions that would otherwise require tedious copy/pasting. If you are familiar with the preprocessor in C++ expanding template definitions, this tool can provide similar functionality but is more generalized.
I think embedding [0] is the real big composability win in Go.
Comparing to C++ templates is a rather generous comparison. C++ templates provide a rich type system that allows type checking and prevents errors, and integrates with the rest of C++'s type system.
Go generate is a lot closer to C/C++ macros, the only improvement being that the generation is done in a turing-complete language instead of by text substitution.
In C-like languages this problem is typically solved with generics, but the go team has decided to ignore 20+ years of language development and try the same thing that didn't work so many times that it motivated the creation of generics.
"But," you say, "Generics slow down compilation!" To which I respond that the first step to compiling code is creating code that does something worth compiling.
Edited for clarity.
Ugh. Sounds like Go is turning into a language that I'll want to avoid in the future. There is no excuse for code generators. Friends don't let friends generate code.
I think embedding [0] is the real big composability win in Go.
This looks indeed more sensible.
AKA macros.
I wish they were at least intellectually honest and just call them "macros" and start nice supporting hygienic macros in the language.
Most languages with generics don't provide for user-controlled code generation via generics (if they do generics-based codegen at all, IIRC GHC uses erased generics for instance). They tend to use macro systems or preprocessors instead.
I don't know that sarcastic is the correct word, but I certainly wasn't completely serious. Maybe "playfully antagonistic".
I suppose the problem I have here is, if I was designing a language, and composability was a stated goal, then I wouldn't have come up with Go.
> a brilliant article from Rob Pike about using and composing functions to create flexible APIs
I'm sorry, but that API is not something to be proud of.
I agree: not only does using it require a lot of boilerplate, but Option takes several arguments and returns a value that only restores the previous value of the last option! That sounds very error prone, I'm sure someone will think the value return by Option undoes all of the option setting.
Technically, it's composability, just not the best kind.
Type structure is a syntactic discipline for
enforcing levels of abstraction.
I don't have a bulletproof argument here, but essentially I'd like to say that greater modularity allows you to forget inessential details and subsequently increase the surface area with which you can compose your modules.
So finally, we come to generics as a powerful tool for using types to erase inessential details. Generics essentially operate via providing a new logical structure to types, the ability to introduce and operate sensibly over type variables. So I would argue that a type language without generics is much like a programming language without variables [1]. You are perhaps not completely crippled, but pretty close.
So, my argument is something like "without generics your type language is crippled so as to make control of modularity very weak which will prevent you (syntactically) from ensuring composability".
The "(syntactically)" bit is important since, as we all know, you can achieve modularity and composability "by hand" through code generation, copy and pasting, creating commenting policies... but then we've stumbled back into the "static types versus dynamic types" argument, so let's not rehash it.
[0] https://wiki.mpi-sws.org/star/paramore?action=AttachFile&do=...
[1] These kind of variables, not "mutable cell" style variables: https://existentialtype.wordpress.com/2012/02/01/words-matte...
That's the key - generics are one way to accomplish things, but they're not the only way.
compose :: (b -> c) -> (a -> b) -> a -> c
? compose :: (b -> c) -> (a -> b) -> a -> c
You think right, that'll teach me (once again) not to typecheck my comments. Thanks, and fixed.
I gave talk that conveys similar sentiment. Slide: https://speakerdeck.com/nexneo/joy-of-single-purpose-service...
If you want to fix the most broken parts of Go -- its ecosystem management tooling -- just replace your `go` command.
I have, and I love working on a project with a proper top-level src folder, with vendor-able packages discoverable in a top-level lib folder pinned to known-good commits.
There are literally dozens of go replacements, but the winning combination for me is using Goop [1], Makefiles, and symlinks to subvert `go get` and the naive way it parses import declarations [2].
However, your code will be unusable by anyone who follows the normal Go way of coding, so you'll be missing out on working with the rest of the community. If you're writing proprietary software, that might not matter too much. But if you're writing open source code, it will be a huge hindrance to getting contributions from outsiders.
Well, that's not really true given the major projects using tools like Godeps [1], so while that argument certainly sounds good, it doesn't actually seem to be a problem in practice.
> the simplicity it presents to regular contributors and users used to go get make it worth the additional effort
You can do anything!
or not...
The standard libraries are not perfect. They do have very useful building blocks, but they are far from complete or perfect.
Take the http library, for example. There's no built-in context:
type HandlerFunc func(ResponseWriter, Request)
type Handler interface { ServeHTTP(ResponseWriter, Request) }
This resulted in a complete mess when it comes to http middleware because everybody is solving the context problem their own way (including google itself).
...The author spends his time away from the libraries, covering the core of the language. By the nature of this approach the matter is dense, but it also gives the author ample time to discuss best practices, for instance:
"...[i]ncidentally it is a Go convention that functions that have channel parameters have the destination channels first, followed by the source channels [in the function definition]..."
Many of these best practices stem from conventions found in Go's own source code, which happily, the author is well acquainted with. Learning a language by isolating its syntax and analysing idioms and conventions is effective for those willing to persevere, and "Programming in Go" abounds in such detail.
Ideally, "Programming in Go" should be paired with a textbook which focuses on use-cases. Balbaert's "The Way to Go" is such a book, but sadly, is not nearly so well written as Summerfield's. In any case, "Programming in Go" is the book for those searching for a solid understanding of Go's fundamentals.
Deleted comment
I'm pretty sure Go was never designed to be the latter, and the former is a matter of personal preference.