Ask HN: Why is Google in love with purposely-impaired languages like Go, Dart?

37 points by nnq ↗ HN
Google seems to invest a lot in R&D for programming languages and tooling (probably only surpassed by Microsoft, and percentually by dev-tooling-focused companies like Jetbrains), yet nevertheless, all they seem to produce seems to be "mediocre by design" in a way, or at least designed to be restrictive and/or to impair developer flexibility and productivity.

On the languages side Go and Dart seem like the epitome of "Blub languages" in PG's essay lingo. Even modern Java seems flexible and featurefull in comparison with them. Compared to Go, Rust looks like Common Lisp looked like to a Java programmer of old. Compared to Dart, Kotlin and Swift look like Smalltalk probably looked to a 90's C++ programmer...

Where is this attitude coming from? Is Google's philosophy that most programmers, including their own, are either stupid and/or irresponsible and shouldn't be trusted with powerful tools that can increase their productivity when used right? I mean, software is their business, they're not churning ecommerce infrastructure CRUDhorrhea... they should be capable of leveraging powerful programming language features to massively increase developer productivity, right?

40 comments

[ 6.5 ms ] story [ 73.3 ms ] thread
I have very little experience in compiled languages. Coming to Go after Python was not only a joy, but I was able to easily implement concurrency and improve the performance of my game of life simulation side project by something like 25x.

So to me it feels like they did something right.

I guess maybe there just sometimes being a way to do something right only one way, makes searching for solutions, easier. I contrast that with another language I love, Ruby, where it can sometimes feel like there are so many ways to do things that the language isn't opinionated at all.

>Coming to Go after Python was not only a joy, but I was able to easily implement concurrency and improve the performance of my game of life simulation side project by something like 25x.

Python is notoriously bad at both these things though, so it's not a very high bar. I think a more meaningful comparison would be to languages like Clojure, Erlang/Elixir, Rust, or any other that has concurrency as one of the top priories, but makes different tradeoffs.

I know this, but I also know I struggled with C and Go felt delightfully straightforward.
"The key point here is our programmers are Googlers, they’re not researchers. They’re typically, fairly young, fresh out of school, probably learned Java, maybe learned C or C++, probably learned Python. They’re not capable of understanding a brilliant language but we want to use them to build good software. So, the language that we give them has to be easy for them to understand and easy to adopt." – Rob Pike
> the language that we give them has to be easy for them to understand and easy to adopt

All the respect for Rob Pike, but that makes no sense: learning ANY new language should be even easier for someone fresh out of school, and I imagine they set a high enough hiring bar to be sure that new engineers can adapt to whatever they are thrown at. All programming languages are easy to learn...

The harder part is figuring out process that are repeatable and result in high velocity and code quality, but I'd imagine that this is what an organization like Google is good at!

The hardest path is finding someone else (hiring) to read and understand the intentions behind your code.
> figuring out process that are repeatable and result in high velocity and code quality, but I'd imagine that this is what an organization like Google is good at!

Seems like Go is manifestation of such process. Because something like Rust is not for a repeatable process of _producing a lot of software_ by big teams with variation in experience and headcount. Rust is probably about _fewer people_ carefully designing _less_ of software, so it can be done once and stay forever.

Rust is for architecting bridges, Go is for paving streets.

well, when they convert the bulk of their stuff to go, if the plan works, and the theory is correct, then they can hire something more like a laborer instead of an engineer.

The main thing is that they are so huge, they can afford such experiments. I don't put much faith in that approach outside of a huge organization, and the jury is still out internally.

And given their shift to no brainteasers and no degree, and diversity and other non skill based focus, and sheer number of employees, the bar may not be set where you think it is these days. Hell it might not be much harder than stating your political beliefs (assuming you have the right political beliefs) based on recent cultural evidence.

"They’re not capable of understanding a brilliant language"

Wow, way to put down your engineers. What is the famously high hiring bar at Google for?

"It must be familiar, roughly C-like. Programmers working at Google are early in their careers and are most familiar with procedural languages, particularly from the C family. The need to get programmers productive quickly in a new language means that the language cannot be too radical." – Rob Pike
> they should be capable of leveraging powerful programming language features to massively increase developer productivity, right?

Here you are claiming that leveraging powerful language features can massively increase developer productivity, but you offer no evidence to back up that claim.

- powerful features -> less code to do something -> less bugs because less code

- powerful features -> more general abstractions -> more code reuse -> less new code written -> less bugs

It's not that simple, but it's good to start with the core assumption that less code -> less bugs and what more powerful languages offer you is ability to solve problems with less code while at the same time keeping that more compact code readable.

Or maybe big software-focused companies like G don't see "# of LOC" as a liability and don't try to reduce it while at the same time as trying to do more? Is this the core difference between a large corporation and a small startup that tries to do more with less?

> core assumption that less code -> less bugs

This assumption is false. Either the flows are written in code, are you have to have them in your brain every time you are reading some code which can handle 10 flows.

> solve problems with less code while at the same time keeping that more compact code readable

Less code means less readable.

> This assumption is false. Either the flows are written in code, are you have to have them in your brain every time you are reading some code which can handle 10 flows.

No, it's not. Your interpretation of it is wrong, though. No code means no bugs, this is true. Now, write a new feature, unless the code is flawless, the code-base will certainly have more bugs than before, this is true. Now, write a linked-list, unless the code is flawless, the code-base will have more bugs than if using the one in the language std-lib. About language features, writing 5 implementations of a linked-list for each type instead of a generic one that it's type checked will certainly contain more bugs as it's more error prone to write the same things 5 times and keep it updated.

> Less code means less readable.

This is ridiculous. Otherwise, we'd all be using assembly. About Go, the lack of generics and simple error propagation is detrimental to being readable. Every function can fail, you don't say!. You're manually propagating the error instead of handling it here, you don't say!. That's what I think every time I go through some Go code-base.

The obvious counter-example to the claim that less code means fewer bugs is kdb/q. It is an extremely powerful and terse language but you can write an astonishing density of bugs per line of code.
really? I thought ppl use APL-family languages bc they allow to express things in very little code so there's more chance of getting it right?...
I don't know about if it works by chance, but I've heard of someone working for two years to get two short lines of APL 'right'.
I think this is true in the case of the individual developer who has some level of understanding of and ownership over their entire codebase. This is no longer true when you're working in medium to large teams, like google is.
New languages help promote economic velocity and prevent towers of babel from rising up. The features of the languages themselves are kind of a red herring that distract folks from the cause of their disenfranchisement. You can build a veritable empire on a 100 million lines of PHP scripts if you've got the resources.
> promote economic velocity and prevent towers of babel from rising up

...could you elaborate what you mean by that?

Not that reserved keywords are the only measure of a well designed language, but it is interesting to note that Golang has 25 (which most likely contributes to the feeling of fluency earlier than many languages). Java has 50 and C++ has 82 keywords for comparison.
The human brain has no problem handling a 70k+ vocabulary of a natural language, just saying... until a language goes over a few thousand reserved words I'd say there's no adverse effect to them as long as they are clearly named/prefixed etc. in ways that don't conflict with frequently used words (and we know we can have sigils like $ for namespacing, # for separating function names from variable names etc. etc.
Yet very few can write an enjoyable novel which others can read.

Try reading documentations to see how the 70k+ vocabulary helped the writer and reader.

Sigils like $ and # are even worse. I had to maintain an old perl codebase once (# @_ holds all arguments passed)

Never worked professionally on a Perl codebase but I kind of liked it when I learned a tidbit and feel a bit nostalgic that it went out of fashion :P
Try reading this Date module (which is very cleanly written compared to the codebase which I was supposed to maintain)

https://github.com/mojolicious/mojo/blob/master/lib/Mojo/Dat...

I just find it impossible to glance over it and understand whats happening even though I know what is supposed to be happening there.

Something similar in golang

https://github.com/araddon/dateparse/blob/master/parseany.go

Hmm, that Perl code seems quite irregular / badly structured, I see some regexps are inline while others in named variables, no clear docs of what is the return of functions etc.

Beside that, the extra clarity in Go is from using a static language. Also, Perl is inexcusable for having no readable function signatures, that's really a no-go for a modern language :|

But for me at least, when I see that wall of boilerplatish Go code, my mind instantly ADHDs away, while those dense squiggles of Perl make me WANT to spend time and focus and solve the puzzle of what they do, I have 10x more chances of paying attention to the Perl code while reading it and maybe even spotting a bug in it :P ...

We have an entire profession built around the often-vain attempt to clarify and resolve ambiguities in the use of those 70k+ words: law.
Cleverness is not related complexity in any way. Software is going to be complex. Cleverness is just going to make it harder to read/maintain.

I prefer Go over other languages. I like the part that there is no cleverness around it.

But why are you bothered about what google thinks of its engineers ? You can write everything in Haskell, why stop at Java/Rust ?

1. enabling cleverness enables creating tool to manage complexity, and hopefully, to manage it with fewer people needed

2. Google "sets the intellectual trend/fashion", whether we like it or not, their attitude get copied... there was quite a bit of experimentation in the 1995 - 2015 decades, we saw stuff like real macros coming to mainstream programming languages (Julia, Scala, Rust), we saw advanced type systems become liked and used (see even TypeScript which basically supports "poor man's ADTs" now), but the "new wave" which seems inspired by G's language design trends seems to be all about throwing it all away, we're scared s'less of complexity and unmaintainabilty that using anything more that string substitutions for code generation when we really-really-really have to seems heresy.

The trends seems to have creeped even into languages like Haskell where nobody's talking about template Haskell anymore and everyone or adding dependent types or whatever, it's all about focusing on package managers and limiting new features and building more tools... Python has a growth spur with 3.0 - 3.6 but that's also freezing in a way, no bold new additions coming without endless discussions.

> 1. enabling cleverness enables creating tool to manage complexity, and hopefully, to manage it with fewer people needed

No. Complexity just makes it harder to read. Go try to read your own code from 5 years ago. I just fail to read even my own code flow every where I used reflections in Java to make it clever (small code which can handle multiple scenarios). Yet I can read every golang codebase in github super easily.

> 2. Google "sets the intellectual trend/fashion"

No. People use what they learn quickly and what gets them productivity. I want to write a software to solve business problems, not want to write software because its cool or something.

Go try to read k8s code. And try to image how it would look like in haskell.

It just looks like you are trying to hate google for the wrong reasons.

> It just looks like you are trying to hate google for the wrong reasons.

Ok, maybe I was a bit inflamed, actually I'm in a love-hate relationship Google's languages, both Go and Dart/Flutter, and using them more and more... but it happens again and again and again that I just wish there were some extra features that if we had them, code would be 10x smaller but still readable :)

Yeah, it's "over-reacting Monday" I guess :P

Is your clear faith in industry at large "leveraging powerful programming language features to massively increase developer productivity" substantiated or just taken as a given because it makes intuitive sense?
What makes sense is that the number of bugs tends to end up proportional with the amount of code...

If a powerful language allows you do the same thing with 10x less code, you'll also have up to 10x fewer bugs (in practice maybe only 2x fewer, but that would still be huge).

> the number of bugs tends to end up proportional with the amount of code

This happens when the original code was clever to begin with and someone (else) came try to fix an edge case and lo and behold.

Technically, you only need a very limited feature set to do whatever you want.

For example - https://en.wikipedia.org/wiki/One_instruction_set_computer.

Technically, most people like using things like the internet, a monitor, keyboards, etc, and ideally would like computations to finish in a reasonable amount of time and be presented in an understandable format. Bonus points if this can occur on physically realizable machines.
Programming languages are like culture. If you're embedded in one then moving to another sometimes seems like a maze of funhouse mirrors.

I once had the misfortune of changing from an Algol environment to a Cobol environment. After a year I'd gotten over my revulsion and became fluent and productive in Cobol.

It's like saying English has lots of irrationalities to it, which is true. But if it's your mother tongue and you're fluent that doesn't matter.

Go is not impaired. Go involves a set of sophisticated choices for software engineering, not for computer science research. If you're going to have many millions of lines of code and some shifting around in your staff, you don't want tricky code that takes a long time to figure out, or that obscures functionality, or that obscures bugs.

Because simple is better. K.I.S.S.
Yes. To put it in other words "simplicity scales" in terms of number of people. It helps to commoditise things faster and lower the cost (i.e learning, operating, hiring, tooling cost etc).
I recently tried out Dart on a project to play with Flutter. It felt pretty much the same as TypeScript/React. I prefer using Flutter/Dart to either of UIKit/Swift or AndroidSDK/* so they're doing something right.