Hacker News Reader

Comment by desireco42 | Hacker News Reader

This is really great little language. I really enjoy using it and executables I get from it can be very small which brings smile to my face.

View on HN

I like everything Andreas did differently from Python, except the unusual normalization rules for variable names. This is crazy:

    VariableName == variable_name == VaRiAbLe_NaMe

I brought it up recently in the #nim irc channel, trying to get a good explanation. The creator still seemed to think it was a good idea, and I could not convince him otherwise.

Yeah, I feel your pain. This is a deal breaker for me.

My OS of choice for the last 25 years has a case sensitive filesystem, all the programming languages I learned (well, except PHP) are case sensitive, it is too hard to unlearn.

BTW, PHP has weird case sensitivity rules. Case sensitive (both user defined and PHP defined):

  * variables
  * constants
  * array keys
  * class properties
  * class constants

Case insensitive (both user defined and PHP defined)

  * functions
  * class constructors
  * class methods
  * keywords and constructs (if, else, null, foreach, echo etc.)

Why would you need to unlearn it? Simply write code as you would write it in a case sensitive language.

I've been programming Python for a living for the last 8 years or so.

In Python there is a convention for naming classes with `CamelCase` and instances with `snake_case`.

I could do that in a case insensitive language, but I can't do this in Nim because it is not only case insensitive, it violates the "principle of least astonishment" by normalizing variable names in a way that makes `FooBar` equivalent to `foo_bar` or `F_o_O_b_A_r`.

If you can't see the trouble with this, I rest my case.

If I understand you correctly, the convention is basically the following, right?

    class FooBar(): pass
  
    foo_bar = FooBar()

The equivalent in Nim:

    type FooBar = object
    
    var foo_bar = FooBar()

Compiles and runs perfectly fine.

How can this be the case? Well, the identifier equality rules in Nim have changed a while back, to make the first letter of every identifier case sensitive. So, FooBar == Foo_Bar == Foobar, but fooBar != Foo_Bar.

You are right. Nim is not case insensitive anymore, it is case sensitive for the first letter. But in my defense, the last time I checked it used to be! Quoting from the wayback machine:

> Nimrod is a style-insensitive language. This means that it is not case-sensitive and even underscores are ignored: type is a reserved word, and so is TYPE or T_Y_P_E. The idea behind this is that this allows programmers to use their own preferred spelling style and libraries written by different programmers cannot use incompatible conventions.

I can see potential problems, but have you actually tried it and found any problems in practice?

And you can always just learn another naming convention. That's not hard. Almost every language comes with its own way of naming things. For me - I'm using really a lot of languages at the same time - it's actually helpful because it makes easier to tell which language I'm writing right now...

Case insensitive identifiers have no bearing on using case sensitive file names. Nim's system is admittedly baroque but it is an attempt to have (mostly) insensitivity without breaking the most common forms needed for C-interop. Just use it as if it were a fully insensitive language, choosing unique identifiers, and the partial case sensitivity will rarely be a problem.

Is there a way to enforce a particular style across a project? From what I've read the abiliity for projects to be internally consistent, even with third party library usage, seems to be one of the big advantages of this design-wise. So if you had an optional setting somewhere that enforces it, it might help communicate the intention of the feature.

There are actually good reasons to reject similar variable names in general. They are a no-go in scripting languages. Fortunately the Nim compiler catches typos.

http://blog.codinghorror.com/the-case-for-case-insensitivity...

Sounds sensible. Case sensitivity makes life harder for beginners and the advantages are minimal (would you really want to have two different variables called VariableName and variable_name in scope?). There are plenty of potential pitfalls (e.g. code that only compiles in Turkey), but the principle is a great idea.

In Python there is a convention for naming classes with `CamelCase` and instances with `snake_case`. It works very well, you can always tell if you are dealing with a class or an instance.

Personally I believe code is harder to read than to write, so I favor languages that are easier to read.

If such a convention applied consistently that would be great. But my Python experience is that it isn't, and an inconsistent convention is worse than useless. If such a distinction is to be meaningful then IMO the language itself needs to enforce it.

I agree, it should be mandatory - one of the things I like about Nim is static typing.

Oddly enough, the "we are all consenting adults" attitude works in Python. Most patterns and interfaces are by convention instead of being enforced.

People don't abuse Pythons permissive nature very often, and when they do they are frowned upon by the community. I think such deviations are easier to spot in Python compared to other languages, don't really know why.

Exactly why I like Go's conventions for public/private, when I was younger and less experienced I always thought that it was great to have freedom to name things however I wanted but now I can't find a good argument against opinionated and enforced conventions for naming.

I actually would appreciate if someone could provide me one.

Keep in mind that Nim is statically typed and the typing prevents ambiguity: foo.say_hi() and SayHi() are not the same.

> Case sensitivity makes life harder for beginners

I think that I disagree. In general (not all the time, but often), things that look different should be different; allowing them to be treated the same is offering beginners a short-time advantage, while preventing them from making, or even understanding, distinctions that they may want to make in the long term. Imagine, for example, a car where both pedals behave the same, and where the actual function is determined by smart traffic-analysis sensors in the car. It would certainly be easier for beginners, but I imagine that it would produce worse drivers; for example, you couldn't drive anyone else's car.

> would you really want to have two different variables called VariableName and variable_name in scope?

Yes, sometimes! (Not that particular example, but there have definitely been cases when I have wanted variables with the same letters in their names but different capitalisations.)

> In general (not all the time, but often), things that look different should be different

Case-sensitivity is the opposite though: it's making two things that look the same be different. I could maybe support a rule (or even just a linter) that each variable had to be written in a consistent case throughout the project, but two different variables that differ only in case should definitely be disallowed, IMO.

> It would certainly be easier for beginners, but I imagine that it would produce worse drivers; for example, you couldn't drive anyone else's car.

I don't think people buy their cars on the basis of how easy they make it to drive someone else's car, and nor should they.

> Case-sensitivity is the opposite though: it's making two things that look the same be different.

No it isn't. fooBar and foo_bar have the same words in them, but they look very different.

Well, sure, because one is before the "and", and one is after it. Not the difference you meant? There are all kinds of differences, but kids usually learn to mostly ignore casing (in English) for meaning when learning to read, and programming makes them unlearn that rule. There are three exceptions (that I think of offhand): starting a sentence, trademarks and acronyms, and proper nouns. The first two don't usually change the meaning much or at all, and the last is pretty much Python's class/instance rule (except reversed).

The thing is, variable names are more akin to proper nouns. They're not words, they're names.

> Case-sensitivity is the opposite though: it's making two things that look the same be different.

I can see why you would say that, but (as my comment suggests) I'm with wtetzner (https://news.ycombinator.com/item?id=10932471) on this one: I think that `variable` and `VARIABLE` look different. I am willing to believe, though, that this may just be a result of my training (as a mathematician); my students often genuinely see no difference between `h` and `H` (as variables), for example, and will interconvert freely.

> I don't think people buy their cars on the basis of how easy they make it to drive someone else's car, and nor should they.

I agree, but people often learn to drive with someone else's car, and I think that's the relevant analogy here.

I think the point in the "variable"/"VARIABLE" difference is that, although they visually look very different, their meaning does not. In a more practical case, if I read some code which has two variables within the same score with names "last_value" and "LAST_VALUE", I would not be able to tell what should they contain: one has the last value in a sequence (which one?), and the other... what?

I agree that, in that example, it's hard to think why one would want both. I not infrequently find it happening (that I want the same name with different capitalisations to mean different things) in TeX, but that's probably not a good example of programming practice (and I can't think of a recent example anyway). It is probably most prevalent in mathematics, where, for example, one will often use `a` for the length of the side of a triangle opposite the angle labelled `A`. Again, though, it's fair to argue that good mathematical practice in naming variables is not good programming practice!

> I could maybe support a rule (or even just a linter) that each variable had to be written in a consistent case throughout the project, but two different variables that differ only in case should definitely be disallowed, IMO.

By the way, it occurred to me several hours later that this sounds a lot like HFS's "case preserving" approach, where case is ignored when matching (so that you cannot have a directory with separate files `a` and `A`), but preserved when displaying (so that a file created as `a` will never display as `A`, and vice versa). I don't want to speak for everyone, but my understanding is that this behaviour is regarded as a wart.

> By the way, it occurred to me several hours later that this sounds a lot like HFS's "case preserving" approach, where case is ignored when matching (so that you cannot have a directory with separate files `a` and `A`), but preserved when displaying (so that a file created as `a` will never display as `A`, and vice versa). I don't want to speak for everyone, but my understanding is that this behaviour is regarded as a wart.

That behaviour's very surprising because it means you can e.g. open a file by name, then check its name, and get a different name from the one you opened it under.

What I'm suggesting is equivalent to: variable names are case-sensitive, but it's an error to ever have two variables in scope that differ only in case (maybe only applied when you actually access one - depends on your views on wildcard imports).

Actually, VariableName != variable_name

You can write ClassesLikePython and variables_like_python in Nim.

Sometimes people end up having dangerously similar variable names in the same scope, e.g. username and user_name, where an incorrect tab-completion might introduce a bug. Nim would require the developer to choose better names.

Two identifiers are considered equal if the following algorithm returns true:

    proc sameIdentifier(a, b: string): bool =
      a[0] == b[0] and
        a.replace(re"_|–", "").toLower == b.replace(re"_|–", "").toLower

So, right, looks like it is case sensitive for the first character allowing the same convention from Python. Still, I find it dangerous:

    m_insensitive == min_sensitive

This would be caught by the compiler. If you have m_insensitive declared in current scope you won't be able to define min_sensitive in this scope, the compiler would throw an error.

Reply on news.ycombinator.com