Hacker News Reader

Comment by exabrial | Hacker News Reader

> It is so intuitive and beautifully designed

No it's not. Significant whitespace, underscores that mean things, lack of type system (but people treat variables as typed anyway), et all, make it quite a difficult language to understand. And we won't get into performance problems that it's cult following turns a blind eye to.

I will say Python has inspired a generation of languages that got away from the C-inspired syntax. We're moved to critically thinking about _why_ we write code the way we do, rather than simply accepting things because "that's the way it's done" and for that we owe Python a great debt.

View on HN

For a beginner, compared to many languages out there, it is very intuitive. I taught many 1st/2nd year college students Python as their first foray into programming, and it's flexibility and readability was a great way to get people in the door.

I won't defend beauty, as that's extremely subjective. I happen to find the whitespace significantly more intelligible than nested braces and parens, but that's me. I know plenty of people who strongly prefer to have braces and parens, so no argument from me.

Everyone loves to bring up performance as a huge drawback to Python, but it never holds up as a fundamental problem. Yes, that means for specific problems, it's not suitable. But the same can be said about many languages. Fortran is a horrible language to try and build a web framework on top of. Lisp is a bad language to write an OS in. Programming languages are tools, choose the right one for the job.

If you need blazing fast processing of numerical data, choose the proper tool. If you want fast iterative development with a broad set of packages and an active community, Python is a suitable tool.

I do agree that the growth in new languages is amazing, and greatly welcome them to the the toolbox that is programing and software development.

You don't need significant whitespace to get rid of nested braces - see every language with a pascal-like syntax.

In Pascal, the braces are simply larger: BEGIN END;

Yes, same in Python:

  '    '

> Lisp is a bad language to write an OS in

Beg to differ. Lisp machines were huge, back in the day, and there are still hobbyist projects like https://github.com/froggey/Mezzano

> and there are still hobbyist projects

Given that it's been decades, I think it may be safe to say "If Lisp were a good language to write an OS in, we'd see more than hobbyist projects and the Lisp machine companies would still be offering them."

(The need for garbage collection caused a bit of a stumble for the Lisp machines on the market, if I recall correctly).

"It was done at least once" != "it was a good choice", though.

There's a reason why most OSes are not written in Lisp, and it's not because the OS writers aren't smart enough to learn Lisp. It's not because they don't know it exists. It's because there are better languages to write an OS in.

I think it's pretty obviously wrong to say that a language is unpopular (for a task) because it isn't good (at that task). Otherwise no one would be using java, javascript, or c++; and lisp and haskell would be a lot better used.

OSes aren't written in lisp because lisp isn't terribly popular, and subset of lisp programmers who are interested in OSes is pretty small--this is a much more interesting factor to look at than subset of OS programmers who are interested in lisp.

> I think it's pretty obviously wrong to say that a language is unpopular (for a task) because it isn't good (at that task). Otherwise no one would be using java, javascript, or c++; and lisp and haskell would be a lot better used.

I think you are confusing "languages that are good" with "languages that you think should be good". I will admit that there are a fair number of programmers that are fashion-following zombies, but there are a large number who are not easily led by the crowd. If Lisp and Haskell were better suited for the bulk of real-world programming, they would be better used. (The alternative - that you're one of the enlightened few, and almost everyone else is a mindless sheep - may feel satisfying, but the very degree to which it is satisfying should make you more wary of accepting it as truth.)

> OSes aren't written in lisp because lisp isn't terribly popular, and subset of lisp programmers who are interested in OSes is pretty small--this is a much more interesting factor to look at than subset of OS programmers who are interested in lisp.

OS programmers in particular are not among those easily led by fashion. If Lisp gave them a real advantage, some of them would use it. ("Some" meaning "more than just Lisp Machines way back in the day".)

> I will admit that there are a fair number of programmers that are fashion-following zombies, but there are a large number who are not easily led by the crowd

Perhaps not, but they are led by money. No one is hiring lisp programmers.

> The alternative - that you're one of the enlightened few, and almost everyone else is a mindless sheep - may feel satisfying, but the very degree to which it is satisfying should make you more wary of accepting it as truth.

I should hope that's not how I come off. My meaning was more that which languages are popular is more a result of inertia than anything. My impression from reading internet forums is that most people who use java and javascript are unhappy with those languages.

> OS programmers in particular are not among those easily led by fashion. If Lisp gave them a real advantage, some of them would use it.

And they do. OS programming is notoriously hard to get into, but even more so if you leave the beaten path. It's inertia again: everyone else is using c, and there's a lot of existing literature about how to write OSes in c, so that's what people do. Rust--which has recently been popular for hobbyist OSes like redox--and c++ were explicitly designed to support the same kinds of programming paradigms as c, so OSes written in them aren't indicative of a deviation from c. But those people who have enough expertise to move past toy OSes and stolen architecture can and do write OSes in other languages, like the lisp os I linked above.

> I should hope that's not how I come off.

Well... you didn't really. You seemed to be somewhat in the neighborhood, which is why I mentioned it, but not really there, so I tried to be gentle. (I hope I came across that way...)

> My meaning was more that which languages are popular is more a result of inertia than anything.

To perhaps strain a metaphor, I think it might be more like a gas cloud gravitationally collapsing. You get denser regions, which have more gravity, so they are able to attract more gas and get even denser. In the same way, a language ecosystem that has more... something... attracts more developers, which makes it a better market for third-party libraries and how-to books and consultants and tools, all of which gives it even more "something". ("Mass" is the word in the analogy; I'm not sure what the word is for language ecosystems.)

But for a project I'm going to start right now, it's perfectly rational for me to take that "mass" (or whatever) into account when choosing a language, because it means I can find programmers, and tools, and how-to guides, even if the language as a language is no better than the alternatives.

I guess that means that I'm agreeing with you about "inertia". I don't think it's just mindshare, though. It's the ability of the ecosystem to help you get your project done.

And the languages that attain that kind of "mass" are languages that were able to attract it when they didn't already have it (when they were just starting out), which is either some indication of merit, or an indication of organizational support. For organizational support, see Java (Sun publicized Java heavily, but also wrote massive libraries). For merit, consider C. (When it started out, it didn't have tons of publicity or great libraries. But it was good enough, and it was free. That was enough to start gaining a following.)

Writing an OS in Lisp basically means adapting or developing, and then growing, a Lisp implementation on the bare hardware. Someone would only do this if they are simultaneously interested in OS research R&D and Lisp implementation R&D.

This is different from someone working on an OS in, say, C. They just pick some C compiler and use it, without becoming a C researcher and implementor.

(They might bemoan some things happening in the development of the compiler they are using that affect their work, but not actually take that on.)

I'm not sure I agree with this. How much assembly would you actually have to write to bootstrap a Lisp? If I understand correctly, not much. And you could cross-compile from some other platform for as long as needed as you use the Lisp to write the OS.

Cross-compiling from another platform (say, SBCL) only brings to your target whatever you have written yourself, not any attributes of that platform (say, SBCL).

You don't magically get SBCL's reader, garbage collector, hash table implementation, object system ... or anything else you might want in the OS.

(You could fork that system though to borrow those things instead of doing it from scratch.)

Why not, if I can get the basics of Lisp running, and those other things are written in Lisp? Isn't the hash table implementation written in Lisp?

Or are you saying that I wouldn't automatically have the source for those things? If so, then... yeah, I was kind of assuming that I would have it. If the licenses don't work that way, or if the source simply wasn't available, then my devious plan would not work.

It's a blog post about the creator of python leaving Dropbox a company who uses a TON of Python. Did you expect them to complain about Python?

Of course there is issues, as there is with ANY language.

Stop these silly language wars...

When I first learned and took up Python in 1996 I didn't adopt it because it was elegantly designed, but because it was pragmatic and the only reasonable (to me) tool for a job that most people used Perl or Tcl or shell scripts for, which I didn't like.

Since then I've watched people use it as an application or systems programming language and I don't really get it.

To me it's a great Swiss army knife scripting language.

But yes, I think you're right. There are far more "elegant" programming languages. I never got the sense that that was Python's niche, though.

A comment like this being upvoted and responsible for a large portion of the "discussion" on the language creator's goddamn retirement announcement makes me embarrassed for this community. Keep the pointless bickering over programming language superiority somewhere else.

We all have our preferences. I find Python to be intuitive. I find Ruby to not be. I am sure someone would spend half a day trying to convince me otherwise, but, it's not worth arguing over in the end. There are more languages than most would care to count. Devs should pick a few that speak to them and move on.

But intuition is not something fixed like an extra toe or even ingrained through a lifetime like a sexual fetish. The core of what humans do is acquire new abstractions in such a way that what was previously convoluted is now intuitive.

"Intuitive" is a word for "stuff I already know". Symplectic geometry is unintuitive until you learn it. Learn stuff.

Yeah. Jef Raskin pointed out that we use "intuitive" to mean "familiar" (in IT.) The computer mouse seems so "intuitive" today, but when he first handed a mouse to someone else, an architect IIRC, to use she turned it upside down and tried to use it like a little trackball. ("The Humane Interface")

I think you mean lack of static type checking. Python is more strongly typed than e.g. Perl and JavaScript.

> lack of type system (but people treat variables as typed anyway)

Python has a type system. The types are on the objects, not the variablenames. If people threat variables as types then it's their misunderstanding.

Correct. It's not a type system that lends itself to simple static analysis, but it's there.

That being said, simple static analysis definitely has its advantages, but it's not so useful that it needs to eat the world.

Python is well-designed for what it is, and it certainly seems intuitive compared to bash or perl or whatever people were using before, but it is opinionated in a way that makes it hard to write anything but procedural-style code. That quality draws some people to it, but it's not a language that can be everything to everyone.

> it's not a language that can be everything to everyone.

Unlike say, for example...?

I'll probably get shit on for saying this, but JS comes closer. It's multi-paradigm, performant, and TS provides static type checking for those that want it. Of course, JS is also a language that gets excoriated for being poorly designed, so maybe intuitive design and adaptability are fundamentally at odds with each other.

Nothing can be "everything to everyone". People are different. No point in trying to conquer the world.

My point exactly. It's unfair to criticise a language on the grounds that it doesn't do what no other language does.

"Lack of type system". That is an absurd claim. Python is dynamic but strongly typed.

Python also can now be optionally typed, which the article discusses. More: http://mypy-lang.org/

I'm sure you know this, but just to explain "optionally typed" a bit more, Python objects are always strongly typed (can't add a string and an int together like in JS, etc.), by default Python variables are not typed - a variable can refer to objects of different types at different times.

Python 3 has introduced syntax to optionally declare types in the source (type annotations). This does not affect the runtime. Mypy is a program that does static analysis to tell you where your code is inconsistent with your type annotations, so:

    a: int = "foo"

Will compile to bytecode and run fine, but if you run mypy it will warn you that you're not supposed to assign a string to a.

Since you brought it up I would like to pile on. There are scads of Academic applications that will remain in use for decades and every single one of them has a half baked Windows port that is used 10000% more often in research institutions where security concerns mandate managed OS not fashion OS or hobby OS.

I don't hate python just lazy windows ports.

Reply on news.ycombinator.com