Hacker News new | comments | show | ask | jobs | submit login

Python significant white space. Really, that's what preventing me for ever liking Python.



When programmers discuss Python on the web, this comes up again and again. I agree with you - it is a minus for me in terms of using Python. There are "solutions", there are people who say you get used to it, there are people who say it does not matter, yet it comes up again and again. I guess it does matter, to some of us at least.

I can open up a text editor and start writing C, or Java, or Perl, or PHP - all without that much thought on whitespace, indentation etc. I can later run indent -kr, or perltidy, or astyle and so forth and make the code I banged out look nice and tidy. Sometimes I feel like doing the nice indentation myself while coding, sometimes I do not and let the scripts do it. With Python "there is one way to do it" and you have to think about indentation while coding. It is annoying. It is like Java and exceptions, forcing me to put try/catch loops everywhere and messing up my scope - even for a little prototype program which will be run once and will never need to catch an exception. It is treating me like I am an idiot and forcing me to work in a regimented manner.

When I first learned Python, at one point I decided to put a do/while loop in my program. I was not able to discover how this was done in documentation. I go on IRC #python and am told Python has no do-while loops (which I later confirmed is true). I was told, and I am assuming the source knows more than me, that this is due to the indentation system. One screwy thing done as a result of another screwy thing.

There are a lot of nice things about Python. The code is clean and understandable. It sometimes allows you to do in 2-3 lines what might take 10 lines in another language. And so on. But some programmers feel certain aspects of it are drawbacks, and these things are not easily dismissable. If they were, they would not be brought up again and again.


So technically not "do/while", but:

  while(some_condition):
    do_stuff


No, it's actually this:

  do_stuff
  while(some_condition):
    do_stuff
The advantage of do/while in this situation is avoiding the duplicate code.


I prefer:

    while True:
        codeHere()
        if shouldStop():
            break


More like:

  do_stuff
  while(some_condition):
    do_stuff


Ahh yes - point taken.


Opinions on Python aside, I'd never hire somebody saying something like that at an interview, even about a language I don't like. You should really look past petty details when evaluating any tech and come up with actual arguments.


Really, why? You wouldn't hire someone that has some "language aesthetic preferences", although a bit different that the mainstream ones? He said "preventing me for ever liking Python" not "I'll never touch it because of this".


This is really the same as saying, "I won't work for company X because I don't think the brackets in the C++ code should be on the following line[1]!"

[1] E.g.:

  int main() {
  }
vs.

  int main()
  {
  }
This is an apt analogy because most companies will have some sort of code style guide that you have to follow, which doesn't make "coding at company X where bracket placement is enforced" much different from "coding in Python where whitespace is enforced."

I've yet to come across someone that had a severe hatred of Python due to the significant whitespace that was a programmer I respected. The last person that I knew (who had a visceral hatred of Python whitespace) wrote Perl code like this:

  sub func1 { shift->func2(shift)->func3(@_) }
or

  sub func1 { grep { $_[0]->func2($_) } $_[0]->func3() }
Note: These are literally code samples with the function names changed.

I don't know any Perl programmers that would condone this sort of code. It's a maintenance nightmare, especially since these were not one-off instances. Seems to me that the people that have a hatred of Python's handling of whitespace don't like being 'constrained' by their language, and are probably happier with Ruby or Perl.


Actually, saying that you'll never like a given detail about a language is not the same as saying that you'll never work for company X because of that specific detail. It is not a logical conclusion, unless you are implying that for one to work for someone else you really need to like everything about the job at hand. I see no reason why you can't do something for money that you wouldn't do for your personal pleasure.


There are some problems with the whitespace in Python.

I wrote some code with lots of nameless functions as input parameters (think JavaScript style or Ruby blocks), which of course was an ugly pain to port to Python with no multi-line lambdas.

List comprehensions is another example -- a kludge coming from that the lack of multi instruction lambdas stop real map functions. (The fun part is that this necessity is touted as something good, instead of admitting that it is unnecessary complexity to learn.)


I've used both Perl and Python. When using Perl I use anonymous functions all the time. When writing Python, I don't. There are instances where in Perl I would use an anonymous function, but in Python I define a local function an pass the reference. Each language has it's own style.

If you're trying to fight the language's style, then you will obviously run into friction. I'm sure there are plenty of people coming from C/C++/Java that think that dynamically-typed languages are unusable because type can't be enforced at 'compile time.' Defining an anonymous function vs. a one-time use local function is mostly a matter of style/preference[1].

  | which of course was an ugly pain to port to Python with
  | no multi-line lambdas
There are two things you are describing here:

1) It was a pain to name all of these anonymous blocks.

2) The idea of naming the blocks is annoying / aesthetically displeasing to you.

#1 is just part of porting something between different languages. You would equally run into issues porting JS and/or Ruby code to Haskell, even with support to lambdas. #2 is just personal opinion / preference, even if it's one that shared by many.

  | (The fun part is that this necessity is touted as
  | something good, instead of admitting that it is
  | unnecessary complexity to learn.
I could use the same argument to rail against statically-typed languages. "A type system is just unnecessary complexity to learn."

[1] I welcome examples of where anonymous functions are unable to be replaced by local functions passed by reference.


>>Ruby code to Haskell

Uh... different paradigms. The scripting languages, which we are talking about here, are generally very similar.

>> I welcome examples of where anonymous functions are unable to be replaced by local functions passed by reference.

Where did I write "unable" -- did I disprove some Turing machine theorem and forget about it? :-)

In this specific case it was a declarative data structure with some data next to code. It was well suited to have a dozen or two short functions next to the data structures.

You could rethink the whole thing and do something different in Python to keep it readable. But among the normal languages (in the same paradigm), Python is the odd man out, just because of the lack of multi line lambdas.

(I also gave the example of bad influences on map, etc.)


  | >>Ruby code to Haskell
  |
  | Uh... different paradigms. The scripting
  | languages, which we are talking about here,
  | are generally very similar.
Different languages are different, no matter how many similarities that they have. You are going to run into issues porting between languages. Expecting anything else is setting yourself up for disappointment.

C++ and ObjC are both 'C-based', so there should be zero issues porting code between them, right? French and Spanish are both Latin-based, so they are practically the same language, right?

  | >> I welcome examples of where anonymous
  | >> functions are unable to be replaced by
  | >> local functions passed by reference.
  |
  | Where did I write "unable" -- did I disprove
  | some Turing machine theorem and forget about it?
  | :-)
Poorly worded on my part.

  | In this specific case it was a declarative
  | data structure with some data next to code. It
  | was well suited to have a dozen or two short
  | functions next to the data structures.
  |
  | You could rethink the whole thing and do
  | something different in Python to keep it
  | readable. But among the normal languages (in
  | the same paradigm), Python is the odd man out,
  | just because of the lack of multi line lambdas.
1. The inclusion of multi-line lambdas in a language is what makes it 'normal'? This just seems more like a personal opinion than anything objective. I could easily say that Ruby is the odd-man out and Perl/Python are the 'normal' languages, because of this:

  a = 0
  if a
    puts "Evaluates to True"
  end
Or Perl is the odd-man-out and Ruby/Python are the 'normal' languages because they have a GIL.

2. If keeping the data structure and functions together was the best use-case, then maybe Python wasn't the tool for the job. Just like I wouldn't talk about how crappy Ruby is because it doesn't have SciPy/NumPy.


>>The inclusion of multi-line lambdas in a language is what makes it 'normal'?

All the scripting languages are very, very similar (see data structures, etc).

In short, you don't argue against my point.

My point was that Python stops multi-line lambdas for some weird purity argument re white spaces (I haven't found any other motivation).

This hurts quite a few use cases (for a 2nd -- copy paste and automatic indentation of code is at least more complex, and so on) and makes it necessary to learn weird syntax (list comprehensions), which incidentally is touted as the next coming of Christ...

And the inability to admit a wart on Python is quite typical -- didn't Guido tell you guys to stop trolling recently, because you were getting bad reputation..?

>> Or Perl is the odd-man-out and Ruby/Python are the 'normal' languages because they have a GIL.

These three are all similar -- avoid threading. :-(

Not that it is relevant for my original point, either.


> These three are all similar -- avoid threading. :-(

Perl gets a bad rep on threading because it's threads don't map to what many people consider the strong points of threads, which is light weight processes. Perl fails spectacularly in that regard, but if you want multi-process programming with a standard, well understood set of primitives for IPC, it's really quite nice. That is, the 'threads' module isn't all that great by itself, but when paired with 'threads::shared', it's really quite nice (as long as you are pre-spawning workers or don't require quick thread spawning).

For example, I love how locks are scope based, which makes the following possible:

  my $local_var = do { lock($lock_var) }
as well as:

  {
    lock($sync_var);
    $sync_var = "start!";
    cond_broadcast($sync_var);
  }


> There are some problems with the whitespace in Python.

> I wrote some code with lots of nameless functions as input parameters (think JavaScript style or Ruby blocks), which of course was an ugly pain to port to Python with no multi-line lambdas.

As nnq points out elsewhere in this thread, this doesn't have anything to do with significant whitespace. Python only allows you to have a single expression in the body of a lambda, and from what I've read the reason it stays this way is to discourage using a lambda as an argument to map, fold, filter & friends. The usual suggestion from Python programmers is to just use a locally defined, named function, which to be honest is often what I do anyway in languages like ML, because it's clearer.

And anyway, just because it was hard to port some code from language A to language B due to something different about B doesn't really seem like a valid knock against B.


My argument was that a lack of real anonymous functions (> 1 instruction) is a problem in some quite common and useful scenarios.

If you REALLY can give good references that show this is wrong, I'm very interested (like, I would guess, most Ruby users).

(It would also be interesting with a link to a PEP or something with, as you claim, arguments that map etc should be discouraged -- and that is why no multi line lambdas exists. For instance, how would the syntax look like??)


My argument was just that it has nothing to do with significant whitespace, which is what you were complaining about. Python could be changed to have non-significant whitespace but retain its crippled lambda.


>>My argument was just

You made specific claims (regarding map recommendations in Python and that white space has nothing to do with one instruction lambdas). I asked for references. I assume you have none?

Note that with significant white space a usable syntax for multi-instruction lambdas seems... non-trivial. It seems to me as an unusually stupid fan boy position. But a few well argued references, which you lack, would change my mind. I'm no troll.

Edit: I might also note that your reference nng is a low-Karma account that argues that a cometing open source language to his love child is dead -- typical language war troll. Is that your only reference?


I should rest all this rip, but I can't help it: reference to multi-line lambdas in languages with Python-like syntax: http://cobra-language.com/trac/cobra/wiki/AnonMethod, http://boo.codehaus.org/Closures (yep, you're right, having multiple closures on a single line with offsite-rule syntax is "non-trivial" and no one solved this yet, but even one per line would be enough to make map and fold bearable, and ftr, I want to punch Guido or whoever else has a saying in language's design for not "stealing" this and adopting it into Python - I think there was a PEP along the lines that got killed and burried, can't find it now).


>>I want to punch Guido or whoever else has a saying in language's design for not "stealing" this and adopting it into Python

Please relax... no tool is perfect for all situations. It is always a trade off.


For most use cases where you want map/filter/etc, you can chain list comprehensions using generators.


You have no relevant comment on what I wrote either, I take it? Here I asked for references about the GP's claims.

Also, welcome to HN (or are the prepared troll accounts with a few hundreds of Karma running out? :-) ).


Sorry, I wasn't trying to troll you.

You asked about multi-line lambdas in python and I gave you a valid Python syntax for multi-line anonymous functions. Here is PEP 202, which simply says list comprehensions are more concise than for loops and map/filter etc.

http://www.python.org/dev/peps/pep-0202/

RE: accounts - I delete all of my accounts on social websites once a year.


  | (or are the prepared troll accounts with
  | a few hundreds of Karma running out? :-) )
Keepin' it classy.


Considering the way Python people have been trolling most every Perl discussion for the last few years when I've been on HN, I feel like I'm in orbit compared to people in the mud...

I might also note that I got two non sequitur answers in a row, so a joke like that is imho not an extreme reaction. Hardly as sarcastic as you was over a joke, at least...

Edit: I might also note that this article (76 votes) is 5 places below another roughly equally old article (8 hours) with 13 votes. So it is probably flagged to Hell. I do feel classy, compared to some others.


At least in Haskell, I don't really care for the significant white space when it's re-factoring time. I don't see how it wouldn't become more of an issue for a traditional imperative language and their tendency for large blocks.


That doesn't make much sense to me. Let's say if I were interviewed at some place using Java, Python, C and Go, why shouldn't I state my personal language preferences?

For most things as an hiring manager I'm OK with people working with their favourite language on whatever project I assign them to (as long as it isn't too esoteric; OCaml is OK, but Haskell is probably pushing it a bit too far, for instance), except when they have to work on an existing code base, in which case I wouldn't want to hire someone hating it; they'd be miserable and the resulting code would be awful.


Personal language preferences are fine if they are justified by actual reasons like tech details, community, librairies, ...


We'll talk about this Tuesday :) People usually have a very good reason to prefer the language they're more comfortable with, because it makes them more productive. Of course, it could be easy to dispell if some guy insists on writing only assembly, or C++, vs Python. I've known a guy who wrote whole Windows GUIs in assembly, it's pure madness.

There are many valid technical reasons to rather choose C than Python, Java than Perl, or PHP than Go, Erlang rather than C++ ...

But Perl vs Python vs Ruby? Hardly any difference, extremely similar dynamic languages. It's only a matter of personal taste at this point.


Why won't you hire someone who has a unfavorable opinion on a language? Do you feel the same way if he mentioned Java instead of Python?


That's clearly not what he said. He just wants actual, technical arguments against a given language, rather than baseless claims that make no sense.

I think it's quite reasonable to be suspicious of a interviewee who is against Python solely because it enforces sensible code indentation.


>> I think it's quite reasonable to be suspicious of a interviewee who is against Python solely because it enforces sensible code indentation.

Though saying "I don't like python's code indentation" is not the same as "I'm against python". Well, at least I see no relation between the two.


I used to think this was a drawback. Then I started using Python full time four years ago. Significant white space encourages readability, and readability (IMHO) encourages maintainability.

There are a numbers of things to dislike about Python; however, after you used the language for a while, you realize significant white space isn't one of them. It is like saying lisp's parentheses are ugly: sure, you might not like them, but they are a huge reason the language (and, really, the way of thinking about code) is so powerful.


Python significant white space is just like nudism. At first you think something essential is missing (clothes, curly braces), then you notice that it wasn't really needed at all, and everything becomes freer and simpler.


That is not a good argument among non-nudists :).


That's the funny part :-)


I agree, "language aesthetics" matters ...that's why I hate variable prefixing with "$", "%" or "@" :) (and out of all ideas PHP could have gotten out of Perl, it had to be this one, and in a laughable way that makes it useless as arrays also use "$") but I could understand Larry's reasoning that "things that are different should look different" and I wouldn't avoid the language just for this reason.


Get an editor that shows whitespace.

(Non)Problem solved.

Not to mention you should indent your code exactly the same way (in a consistent format, either tabs or spaces) with how you'd indented Python code.


It isn't totally solved. For example, you cannot write a lambda that contains a line-break. Its not a big problem, but it is a trade-off.


Put this in a text editor and run it in python:

  print(reduce((lambda x, y:
                x + y), xrange(10)))
  print(reduce((lambda x, y: x + \
                y), xrange(10)))

HINT: It runs.


significant whitespace is not why you cannot write a lambda that contains a line-break. the problems are not really related. just ditch the difference between statements and expressions and allow multi-line expressions. true, then it will no longer be Python but a whole different language :)


When you're writing code in Perl, Ruby or some other language, do you normally not use any indentation? Do you leave all of your code left-aligned?


I didn't say I hate Python, mind you; but I don't like its significant whitespace, like some don't like Lisp for its parentheses and Perl for its sigils.

I'm using Python from time to time, but I wouldn't spontaneously use it for my personal projects, I prefer Perl for that.

The only languages I really wouldn't want to work with is Java.


Englishalsohassignificantwhitespace.


Not a fair comparison -- English prose

doesn't

care

how

much

whitespace appears between words, only that there's more than none. Python definitely cares how much.

Don't get me wrong -- I like Python and I use it a lot, but its reliance on syntactically significant whitespace isn't in the plus column.


Try submitting either of our comments to a reputable publication. I doubt the editor will preserve our whitespace.

English has well-defined whitespace rules, as does nearly every programming language. Every programming language style guide I've ever read has talked about indentation, so you have to indent according to rules no matter which language you use.

Have you ever thought to yourself, "wow, there's no way I'll be able to finish this task on time because I have to indent according to Python's rules"? I'm guessing no.


> Try submitting either of our comments to a reputable publication. I doubt the editor will preserve our whitespace.

True, but it's because changing the whitespace doesn't break the prose. But in Python it certainly breaks the code.

> so you have to indent according to rules no matter which language you use.

Sort of. I've written any number of programs in languages that use braces as delimiters, but very lazily, then used a beautifier to fix things up at a keystroke.

I agree with your point that the source must eventually follow accepted formatting, but it's not true that I necessarily need to pay attention to that (as I do in Python).

Some of the better beautifiers even enforce optional style choices, like turning this:

    function_name(variables)
    {
Into this:

    function_name(variables) {
Automatically.

The latter style has pretty much taken over since I started programming 40 years ago, to the degree that I've come to like it.

I emphasize I like Python and I find its advantages far outweigh the whitespace issue on my personal radar.

> Have you ever thought to yourself, "wow, there's no way I'll be able to finish this task on time because I have to indent according to Python's rules"? I'm guessing no.

You're right, but only because my editor automatically manages Python indentations in most cases, so I don't think about it any more.

The one caveat are Python source files that, from time to time and for various reasons, get infested with tabs. Then things get complicated as I try to get rid of the tabs without destroying the program's logic. IMHO tabs should be outlawed.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: