I have a question about the difference(s) between Python and Ruby. I have been working with both languages for about one year each now, and I keep finding differences that turn out to be just syntactical. I will elaborate, but this is the gist of my question:
What can Python do that Ruby can't? What can Ruby do that Python can't? What does Ruby gain from its syntactic quirks, other than easier DSL implementation?
I recognize that this is potentially a very "religious" question, so I'd like to specifically note that I'm not talking about whether one language is "better" or faster or easier to use. I'm just trying to grok the real differences between Ruby and Python, and the reasons behind the design decisions that created them as they are today.
For my comments below, please refer to the attached rar.
One example: A simple Python closure technique can be seen in clos.py. This leverages Python's simple first-class function support. However, with very little effort, we can replicate the technique in clos.rb with a lambda and the call() method.
Conversely, we can look at a good usage of Ruby's block/yield functionality in iter.rb. Again, with only a little bit of head-scratching, we can translate this rather directly to Python using UserList and, again, Python's first-class function capabilities.
Even in terms of real closures, like the example in clos2.rb and clos2.py, which I shamelessly stole from Google, Ruby and Python fail to really outpace each other.
So, how do blocks in Ruby REALLY differ from lambdas in Python? Is there anything you can do in one language that some fancy footwork can't achieve in the other, or is that the point of their differences? I know all of the knee-jerk reactions: Multiple inheritance, syntactical indentation, keyword arguments, regex literals... those are all either matters of taste or easily simulated. What are the "real" differences, if any? Was Matz's intention in creating Ruby to approach existing functionality with a new syntax, or are there things I'm not seeing?
Thanks so much in advance for all of your thoughts and comments. I'm looking forward to seeing what everyone says, and solving this mystery!
On Wed, May 7, 2008 at 5:23 PM, Max Cantor <maxcan...@gmail.com> wrote: > I have a question about the difference(s) between Python and Ruby. I > have been working with both languages for about one year each now, and I > keep finding differences that turn out to be just syntactical. I will > elaborate, but this is the gist of my question:
> What can Python do that Ruby can't? What can Ruby do that Python can't? > What does Ruby gain from its syntactic quirks, other than easier DSL > implementation?
One often-overlooked point is the design of the standard library. For example, one of ruby's major design decisions is the syntactically convenient and addition of one free closure to every method, with a lightweight, performant way to invoke it. That led to things like a collection interface built around the concept of internal iterators, the pervasive use of the open resource/ yield to caller/ close resource pattern, and an avoidance of higher-order-functions and patterns that would have required *two* lambdas passed in. Python both makes inline anonymous closures highly inconvenient, and provides cheap generators (something hard to do in ruby without a large performance hit), leading to the more pervasive use of external iterators. Also, rubyists like method chaining, and pythonists don't, which is why several methods in the python stdlib return None if they are being called for their side effects, where a ruby method would tend to return self.
On Wed, May 7, 2008 at 8:23 PM, Max Cantor <maxcan...@gmail.com> wrote: > What can Python do that Ruby can't? What can Ruby do that Python can't? > What does Ruby gain from its syntactic quirks, other than easier DSL > implementation?
> So, how do blocks in Ruby REALLY differ from lambdas in Python? Is > there anything you can do in one language that some fancy footwork can't > achieve in the other, or is that the point of their differences? I know > all of the knee-jerk reactions: Multiple inheritance, syntactical > indentation, keyword arguments, regex literals... those are all either > matters of taste or easily simulated. What are the "real" differences, > if any? Was Matz's intention in creating Ruby to approach existing > functionality with a new syntax, or are there things I'm not seeing?
First of all, you seem to be coming at it from the assumption that Ruby is a response to Python, or vice-versa. Both Ruby and Python are far more influenced by their shared ancestors - Lisp, Smalltalk, Perl, Modula, C, to name just a few - than they are by each other. BOTH languages are "nothing new" in the sense that Lisp, Smalltalk, Perl et al already had full fledged closure semantics when Ruby and Python arrived on the scene. So yes, in that sense it's just a matter of syntax.
If you are interested in a closure-related feature that DOES differentiate the two languages, look at continuations (Kernel#callcc). Python, with the possible exception of the Guido-disowned Stackless variant, does not support continuations. Of course, with JRuby not supporting continuations and MRI 1.9 apparently moving to Fibers, this may cease to be a point of differentiation in the future. A prospect that I personally find disappointing.
Max Cantor wrote: > What can Python do that Ruby can't? What can Ruby do that Python can't? > What does Ruby gain from its syntactic quirks, other than easier DSL > implementation?
Python's DSL support is pitiful. What you call a "quirk" is instead the ability to make coding _all_about_ those silly fringe quirky DSLs.
On 8 Maj, 04:09, Phlip <phlip2...@gmail.com> wrote:
> Max Cantor wrote: > > What can Python do that Ruby can't? What can Ruby do that Python can't? > > What does Ruby gain from its syntactic quirks, other than easier DSL > > implementation?
> Python's DSL support is pitiful. What you call a "quirk" is instead > the ability to make coding _all_about_ those silly fringe quirky DSLs.
|> Python's DSL support is pitiful. What you call a "quirk" is instead |> the ability to make coding _all_about_ those silly fringe quirky DSLs. |> |> -- |> self.Phlip | | dsl as in?
Domain Specific Language. Rake, Rails, RSpec, Adhearsion..
~ - You know you've been hacking too long when... ..your friends who aren't hackers wonder what happened to you. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
I was about to comment on something but when i read this:
> What does Ruby gain from its syntactic quirks
I could not really continue. It reminds me of the terminology "monkeypatching". I dont really want to comment towards derogative terminology. (To me, a quirk implies something derogative. Same as monkey-patching implies to me as a "not good patch" that was applied under speed pressure, very quickly and without much thought etc but using a derogative term for it .).
> I recognize that this is potentially a very "religious" question
I dont really think it is per se. But one maybe has to be careful which words he uses. Unless that guy in question does not care anyway maybe (mister "rails is a ghetto") :)
However, the python vs ruby comparison is, IMHO, really not such a huge difference, because both share many goals. It is much more a beneficial competition between the two. With perl biting the dust (I could not resist ... ;> )
One example - is that in speed comparisons, python outperforms ruby, but I never understood why anyone who uses such "slow" languages would care about speed anyway.
DSL is also a "difficult" term, because I have seen about 6 different definitions of a DSL. It would be much better if people could all agree about a DSL. And most importantly, if a DSL is a real language *WITHIN* that language, or whether it is not. -- Posted via http://www.ruby-forum.com/.
> However, the python vs ruby comparison is, IMHO, really not such a > huge difference, because both share many goals. It is much more a > beneficial competition between the two.
Really? Python just stands all the time in my way, ok it is the Ruby way I am used to I have to admit. But I know enough Python to know which one is easier on my fingers and my eyes. Do not forget that Ruby's design gool was to please Matz, we should check with Guido if this was his design goal too ;).
> With perl biting the dust (I could not resist ... ;> )
Those been told dead live longest, and this might as well not be the exception confirming the rule. <snip> Robert
On Sat, 10 May 2008, Marc Heiler wrote: > I was about to comment on something but when i read this: >> What does Ruby gain from its syntactic quirks > I could not really continue. > It reminds me of the terminology "monkeypatching". > I dont really want to comment towards derogative terminology. (To me, a > quirk implies something derogative. Same as monkey-patching implies to > me as a "not good patch" that was applied under speed pressure, very > quickly and without much thought etc but using a derogative term for it > ..).
Also, a quirk can only be measured in relation to something non-quirky -- in this case, I guess, the syntax of other languages. (I don't mean to lay this at the feet of the OP of this thread; it's a pretty constant theme.) The sense that Ruby is uniquely accountable for the ways in which it differs from other languages is very persistent. I'm not sure why, especially now that it's got such a large usership. Or maybe that has something to do with it; I don't know. It's all very meta-meta :-)
>> I recognize that this is potentially a very "religious" question
> I dont really think it is per se. But one maybe has to be careful > which words he uses. Unless that guy in question does not care > anyway maybe (mister "rails is a ghetto") :)
> However, the python vs ruby comparison is, IMHO, really not such a > huge difference, because both share many goals. It is much more a > beneficial competition between the two.
I remember hearing an intro presentation about Python where you could have crossed out Python and put in Ruby and it would have been *exactly* what people say all the time about Ruby: concise, elegant, OO, doesn't get in your way, etc. etc. The two languages definite share many goals, and appear to have succeeded at them. (I'm not interested in Python myself, but I do take Python programmers at their word about what they love about it.)
> With perl biting the dust (I could not resist ... ;> )
> One example - is that in speed comparisons, python outperforms ruby, > but I never understood why anyone who uses such "slow" languages > would care about speed anyway.
> DSL is also a "difficult" term, because I have seen about 6 different > definitions of a DSL. It would be much better if people could all > agree about a DSL. And most importantly, if a DSL is a real language > *WITHIN* that language, or whether it is not.
-- Rails training from David A. Black and Ruby Power and Light: INTRO TO RAILS June 9-12 Berlin ADVANCING WITH RAILS June 16-19 Berlin INTRO TO RAILS June 24-27 London (Skills Matter) See http://www.rubypal.com for details and updates!
On Fri, May 9, 2008 at 12:23 PM, Marc Heiler <sheve...@linuxmail.org> wrote: > I was about to comment on something but when i read this: >> What does Ruby gain from its syntactic quirks > I could not really continue. > It reminds me of the terminology "monkeypatching". > I dont really want to comment towards derogative terminology. (To me, a > quirk implies something derogative...
My usage of the term "quirk" was never intended to be derogatory. To me, a "quirk" is simply an idiosyncrasy: something unique to a particular group or idiomatic context. For example: No matter which way you slice it, "arr.each { |item| item.foo }" and its underlying implementation is unique to Ruby, and I'm really interested to see why, specifically, Matz decided to do it that way, for that and other decisions. Maybe a different spin on my question could be, "What problems did Ruby solve, as a new language?"
On Fri, May 9, 2008 at 1:13 PM, Max Cantor <maxcan...@gmail.com> wrote: > For example: No matter which > way you slice it, "arr.each { |item| item.foo }" and its underlying > implementation is unique to Ruby
On Fri, May 9, 2008 at 1:29 PM, Avdi Grimm <a...@avdi.org> wrote: > On Fri, May 9, 2008 at 1:13 PM, Max Cantor <maxcan...@gmail.com> wrote: >> For example: No matter which >> way you slice it, "arr.each { |item| item.foo }" and its underlying >> implementation is unique to Ruby
> It is?
Is there another language out there that lets you pass a "literal" block of code as a virtual argument so the called method can push arguments into the code block specified by | |s?
On Fri, May 9, 2008 at 8:01 PM, Max Cantor <maxcan...@gmail.com> wrote: > On Fri, May 9, 2008 at 1:29 PM, Avdi Grimm <a...@avdi.org> wrote: >> On Fri, May 9, 2008 at 1:13 PM, Max Cantor <maxcan...@gmail.com> wrote: >>> For example: No matter which >>> way you slice it, "arr.each { |item| item.foo }" and its underlying >>> implementation is unique to Ruby
>> It is?
> Is there another language out there that lets you pass a "literal" > block of code as a virtual argument so the called method can push > arguments into the code block specified by | |s?
Of course, with minor syntactic change. It's called Smalltalk IIRC ;)
On 5/9/08, Max Cantor <maxcan...@gmail.com> wrote:
> On Fri, May 9, 2008 at 1:29 PM, Avdi Grimm <a...@avdi.org> wrote: > > On Fri, May 9, 2008 at 1:13 PM, Max Cantor <maxcan...@gmail.com> wrote: > >> For example: No matter which > >> way you slice it, "arr.each { |item| item.foo }" and its underlying > >> implementation is unique to Ruby
> > It is?
> Is there another language out there that lets you pass a "literal" > block of code as a virtual argument so the called method can push > arguments into the code block specified by | |s?
well except for the ||, what about c? Assuming an 'Object' struct has been defined:
Object set[N]; /*<fill set>*/ int idSort(void* l, void* r) { /* Here's my literal block of code */ return ((Object*)l)->id - ((Object*)r)->id;
}
qsort(set, N, sizeof(Object), &idSort); /* Which I pass as argument */ /* so the called method can push items into it */
What's unique to Ruby (at least in your example) is not that I *can* do it, just that I can do it in roughly a quarter of the equivalent c code: set = [] #fill set set.sort_by{|o|o.id}
Michael T. Richter wrote: > I'm not sure that this is a meaningful question. What problems did > *any* language past patch cabling circuit boards solve? If you set the > bar low enough (or high enough) all current computer languages are > imperfect reflections of a Turing machine anyway. (Yes, even the > functional ones based on Church instead of Turing. They're just REALLY > obfuscated.)
Actually, I think it's Turing and Von Neumann that were obfuscated -- Church and McCarthy got it right. ;)
But seriously, I don't recall having read any of Turing's practical work on design of real computers, but I *have* read Von Neumann's, and there is a certain elegance in Von Neumann's design that I don't think has ever been surpassed, and it's certainly much more elegant than a Turing machine.
On 10 May 2008, at 07:31, Michael T. Richter wrote:
> Church's model of calculation is far more appealing to me and the > languages based on it -- Lisp (arguably: there's some evidence that > McCarthy stumbled over this rather than deliberately trying to model > Church), Haskell, etc. -- are increasingly the way I like to work. > But it's all smoke and mirrors. Underneath it all is a von Neumann > machine masquerading as a Church lambda expression engine.
Which of course drives home the point that all languages are about useful abstraction, allowing those who use them to discuss problem spaces without drowning in detail. In my experience Ruby is very good at this even though it lacks some of the abstraction mechanisms I like in other languages (such as Icon's goal direction), but another developer's view may differ completely.
Why? Because language is not just a means of communicating between two or more abstract and indistinguishable entities but a means of transferring information between separate knowledge spaces in a manner that will allow them to move towards a common representation of a given problem domain. Its efficacy is therefore dependent on both the structure of the language itself and the peculiar topography of all the knowledge spaces.
In more concrete terms, there are programmers for whom Python is a much closer fit to their stored knowledge base than Ruby, and vice versa. It is therefore easier for them to investigate a given problem domain using the one language than the other except in cases where the problem domain itself is significantly more compatible with the other language.
The same logic explains why some people love Lisp above all other languages, and others would happily consign it to the bin of 'academic' languages ;)
Michael T. Richter wrote: > Mathematically I agree with you, but in terms of hardware underlying all > this stuff it's basically a real-world Turing machine. (Which is what > the von Neumann architecture is: Turing's machine turned into something > that could actually be implemented. Things like "infinite tapes" and > "infinite decision tables" turned out, surprisingly, to be implausible > at point of implementation. :D)
Again, I have to plead ignorance on Turing's contributions to practical computing. But the Turing machine was introduced in the context of logic and foundations of mathematical logic, *not* as a conceptual foundation for *computing*, either scientific or commercial. The same is true of Church's Lambda Calculus and Schonfinkel's Combinatory Logic, as so eloquently documented by (Haskell!) Curry and Fitch and numerous others.
The Von Neumann machine, on the other hand, was designed from the ground up for the practical solution of equations -- linear, nonlinear, algebraic, differential, difference and integral. It was patterned on successes of the past in this domain using electromechanical (relay) or even mechanical technologies. Indeed, the original Burks/Von Neumann design looks a lot like a programmable Marchant or Friden desk calculator, except that the mechanical calculators were always decimal and the Von Neumann machine was binary.
> Church's model of calculation is far more appealing to me and the > languages based on it -- Lisp (arguably: there's some evidence that > McCarthy stumbled over this rather than deliberately trying to model > Church), Haskell, etc. -- are increasingly the way I like to work. But > it's all smoke and mirrors. Underneath it all is a von Neumann machine > masquerading as a Church lambda expression engine.
McCarthy and others have written extensively on the origins of Lisp, and I think most of the design decisions are well documented. But the key concept McCarthy and the others introduced with Lisp is analogous to Von Neumann's. Von Neumann introduced the practical equivalence of programs and data by storing them in the same address space. McCarthy introduced the practical equivalence of programs and data by expressing programs as S expressions just like the data. I don't think either Turing's or Church's "logics" do this.
Lisp 1 and 1.5 were indeed Von Neumann machines interpreting Church lambda expressions. But later generations of Lisp machines arose that carried out the interpretation at a lower level of hardware abstraction.
> /There are two ways of constructing a software design. One way is to > make it so simple that there are obviously no deficiencies. And the > other way is to make it so complicated that there are no obvious > deficiencies. (Charles Hoare)/
Charles Anthony Richard Hoare, most often referred to as C. A. R. Hoare, although I believe his friends call him "Tony". :)
I am also beginning to really enjoy Erlang's approach to parallelism and wishing I could do cool stuff like that in Ruby as easily
Me, too. That was the impetus for dramatis (http://dramatis.mischance.net; actor concurrency for ruby and python; early but always looking for feedback.)
Apropos of the overall thread, I’m completing the port of dramatis to python and the metaprogramming is pretty different. Python uses metaclasses and doesn’t have the callbacks that ruby does. There’s overlap in the feature set, but they don’t provide exactly the same features.
More importantly, object dereference, e.g., “a.b”, is very different. It’s difficult to do proxy objects in python, for example, rpc stubs, because it requires accessing the object to know whether a.b is calling a function (if b is a property) or simply binding a method pointer (if b is a normal instance method).
|
