Slashdot Mirror
The Python Paradox, by Paul Graham
GnuVince writes "Paul Graham has posted a new article to his website that he called "The Python Paradox" which refines the statements he made in "Great Hackers" about Python programmers being better hackers than Java programmers. He basically says that since Python is not the kind of language that lands you a job like Java, those who learn it seek more than simply financial benefits, they seek better tools. Very interesting read."
...a civil discussion will ensue.
Ironically, the word ironically is often used incorrectly.
Right on. I wish employers/customers would take a look at the verity of your programming skills and not the number of years programming in one language. So what if you have 10 years of programming VB. Even if you have to program a VB app if you see a person with 3 years experience in VB and 7 Years in a buch of other languages it shows that this guy know how to program and is flexible to work around problems. Unlike possibly the 10 year VB guy who knows all the prebuilt widgets but something outside those widgets becomes impossible for him. I can't even count the number of times I helped people program on languages that I never used before (and they were soposed to be the experts) (I even helped out the Microsoft Guy in the 2003 Linux world expo in .Net) It is because I know a lot of languages and I can use concepts from the different ones and relate it to different languages. I also hate it when a customer tells you that they need an application written in this language to do this. My view is use the correct language for the job and I hate being forced to use a language that is not well optimized for the job. It is like someone telling someone when they build a house that they have to use this screwdriver and only this screwdriver to build a house. Someone who is truly a professional knows the different tools available and will use them when needed.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
I like Python because I can figure out what's going on. If you know C, then it is not hard to trace what happens at the lowest levels since the C-Python is written in a remarkably clean C and very well documented. So it's a simple, powerful, easy to read language where you have some assurance of being able to track down most problems. My experience with Java was quite the opposite - every Java book I read always had mysterious claims about threads, JVM, synchronization, garbage collection that seemed like some sort of "insider knowledge" and I was expected to just believe it. So I think it's not just about the language itself as much as it is about the implementation; for me at least.
We used to script most of our processes (digital chip design) with Perl. In reality, only a few people really bothered automating boring tasks. At one point I started writing everything in Python because of all the good reasons (readability, easy to learn and MUCH easier to maintain later on) and gradually spread the gospel. As very nice side effect is that my collegues are much more likely to get their hands dirty themselves and write scripts with it's useful. Anyway, as for the article: I think a language shouldn't only be beautiful in the way it allows one to express intent, but also or even much more so in the way it looks esthetically. This is one of the biggest problems I have with Lisp (after reading Paul Graham's other articles, I bought this ANSI Common Lisp Book and printed out out 'On Lisp'). It's a fascinating language, but it looks to incredibly dense. Sigh.
I''ll be willing to buy his theory that python hackers are better than java hackers...
I think the main reason I write so much stuff in Python and pretty much nothing in Java is simple: Open source and comprehensive standard library.
When I did have to write something in Java - well, better hope you have the right API. Oh, and what's the difference between the 3587324 different XML parsing packages? Oh and download Java Super Micro PDA Library! Only available for Solaris and Windows!
The promise of "write-once-run-anywhere" was pretty much dead. Not that I was even going for portability - I just wanted it to run on Linux. And I wanted to do it without having to download Beans, Java DynamicManagement, Java Metadata, ad nauseum.
Python, on the other hand, simply works. Sure, it doesn't have a standard GUI toolkit (although wxPython is pretty much the de facto standard now...), but it does everything I need it to do. With a clean, nice syntax, no less.
Oh, and I still don't get how Java doesn't let you write a freakin routine to get the permissions on a file without resorting to writing a C function. Sure, it's not "portable", but just do what Perl and Python have done - on platforms with no permissions, return a sane default value. Don't just NOT include the function at all...
Anyway, forgive my rant. Python - good standard library. Java - tons of confusing frameworks and platforms, etc, etc.
...but it's being eaten...by some...Linux or something...
The Army reading list
The immense number of system and third party libraries is what keeps most Java developers coding Java. The close second is Java developer's tools. Until there is an IDE for Python that's as extensive, capable and sophisticated as Eclipse I'm going to stay with Java. Most Java development is in the enterprise server side space and the sheer amount of tools that Java offers is just mind boggling.
Also us, enterprise developers tend to work in environments that are much more conservative from the technology standpoint (banking, insurance, brokerage). As far as I'm concerned it's a big win for us that we got Cobol elbowed out with Java. Trying to push the language of the month at those executives will cause us more harm than good.
I know that Python has some very nice features (I read the tutorial) but it's hardly the sort of paradigm shift that merits ditching Java and rewriting everything because of some neat syntax flavoring. Besides Java is hardly a frozen language and we have some exciting stuff coming down the pipe here. JDK 1.5 will introduce shared VM model which may make java compeling on the desktop, more elaborate iterators, annotations (my favourie in 1.5), and generics (although without primitive types support they are kinda lame).
Java is now frequently used in CS research as well. It looks almost certain that the next milestone in CS evolution will come in the form of Aspect Oriented Programming and AspectJ has been the leading implementation.
Sometimes it feels that Paul G. just has an ax to grind into the collective Java community but I wonder how closely did he actually look at Java before dismissing it?
Your pizza just the way you ought to have it.
n 1960, a researcher interviewed 1500 business-school students and
classified them in two categories: those who were in it for the
money - 1245 of them - and those who were going to use the degree to do
something they cared deeply about - the other 255 people. Twenty years
later, the researcher checked on the graduates and found that 101 of
them were millionaires?and all but one of those millionaires came from
the 255 people who had pursued what they loved to do!
Research on more than 400,000 Americans over the past 40 years
indicates that pursuing your passions - even in small doses, here and
there each day - helps you make the most of your current capabilities
and encourages you to develop new ones.
Dyslexics have more fnu.
I use Perl, Python, Ruby, Lisp, Scheme...
Out of those, Python is probably the least Lisp-like, and the worst designed (well, okay, yes, Perl's design is *much* worst, but it can be massaged into doing lots of cool stuff).
Yet PG brings it up all the time when he talks about Lisp. To me Lisp and Python are like night and day. How do you create anonymous functions and pass them as variables in Python? You can't, only "lambda *expressions*" which is a strange and arbitrary distinction. In Lisp, creating functions on the fly is the norm. Python doesn't have macros, or even blocks like smalltalk or Ruby, which again is one of the best things about Lisp, allowing you abstract and refactor the flow of code itself. In Python, you have arbitrary but fixed structures bolted on, like comprehensions, or tuples. In Lisp, you can create new constructs on the fly, using the same syntax as everything else.
In my opinion Python is like the Java of the open source world: tons of people use it, they think it's great because it's a lot better than whatever they used last year (C++, Perl, etc), and they don't realize (or care) there are better languages that can help them work faster.
Maybe Paul should adjust his spiel to simply say: the more obscure a language a person has mastered, the more likely he is a smart self-motivated programmer.
But even that isn't true all the time...
I'm not a programmer, I'm just an engineer who can do impressions when he has to. This language is relatively simple and organized and, with the growing number of math and scientific libraries available, it's becoming more and more a regular tool in my arsenal.
I guess the point I'm making is that this language has a strong appeal to people like myself who are just looking to use it to solve problems and/or make problem solving tools. I don't know anything about Java, but maybe the reason that the author feels that Python has the better hackers is because Python, by virtue of it's simplicity and no-nonsense syntax, tends to attract your more "problem solver" type person.
Just a thought.
A goal is a dream with a deadline
Pardon me while I go build a better mouse trap, pontificate on how much better it is and what a great mouse catcher I am, and then put it in my hamster's cage to prove it.
Hmm...let's see.
Both languages have a virtual machine.
Both languages can use JIT compilers for improved efficiency.
Both languages support OO development.
Both languages have large standard libraries.
Both are turing complete.
Even their performance is similar.
Please explain how this is an inappropriate comparison.
*sigh* back to work...
Anyway, something like that. Here is the closest he comes to a paradox:
First, employers are always looking for people who go above and beyond the bare-minimum, including people who like what they're doing enough to do it even when they aren't paid. That doesn't constitute a paradox. Second, as soon as people widely believe that it is a good standard for hiring programmers (meaning it really is the language to learn to get a job), people will start learning it merely to get a job, so I'm not sure his statement really even makes sense.
So, while I'm not saying anything about his statement that python programmers are better (since I'm not a very good programmer in any sense, and wouldn't know to argue), describing it as a "paradox" seems like pseudo-intellectual camouflage for a "Python RULES!" article.
Not trolling here, but this opinion piece is stupid. Hell, just look at that quote above. "He couldn't be a first rate hacker since he obviously chose NT voluntarily." According to the author, there's no way to succeed if you choose to build on NT.
Knowing Python doesn't make you a "first rate hacker". Any decent programmer can pick up a language like python in a day or two. A good hacker (i.e. a programmer that a company would want to hire) is someone who can take their previous experience and apply that to the problem at hand, using the tools available. Saying "...but I know Python" is the same as saying "...but I know Assembly" when you have a bunch of C++ code to write.
"The market alone cannot provide sufficient constraints on corporation's penchant to cause harm." -- Joel Bakan
Don't really see how you can compare a scripting language with an OO development language.
Python clearly is an object-oriented development language; it even has multiple inheritance. Python is pretty close in its semantics to Smalltalk, and there are several native compilers and environments for Python. So, Python really is much more than a "scripting language".
It's not clear that Java should even be called "object oriented". Alan Kay said "I invented the term Object-Oriented, and I can tell you I did not have C++ in mind." Well, Java's object system is even more restrictive than C++'s.
So, yes, it does make sense to talk about Java and Python and compare them.
I think more people who are java programmers should be trying Python for smaller programs, cause I was able to pump small programs out much faster than with Java.
I think this point is interesting. Paul Graham makes a big deal about why he doesn't think much of Java, and it all seems to go back to the big/small issue. For example, in his essay on how he put together the Yahoo stores using Lisp, he mentioned that interpreted languages (or at least languages that can be both interpreted and compiled like Lisp) are better because they let you do things incrementally, as opposed to languages such as C/Java/etc, which force you to do the whole project all at once(I'm generalizing his statements here, but that seemed to be the gist of it).
My point is this: Graham seems to really like the "small is better" approach, for which interpreted languages really shine. For larger projects, Java/C++/etc. would (might?) be more appropriate, but for Graham's projects and ideas, not really relevant.
Every once in a while, Graham comes out with something that seems to get Java users up in arms, but it's entirely possible he's operating from point-of-view that involves projects that Java wasn't really designed for to begin with.
Just a curiousity I thought I'd note.
--Erik
It's a really nice idea, with great direct effects and horrible side-effects.
If at first you don't succeed, skydiving is not for you
Google turned this up, which cites a source: http://www.mindview.net/WebLog/log-0037
From http://www.paulgraham.com/javacover.html :
...
I've never written a Java program, never more than glanced over reference books about it
Do I need to add more?
First off, let me be clear: I like Python. I think the throw-back to the yesteryear of line-oriented programming with enforced indentation style is quaint, bordering on painful and that the ultra-dynamic typing without dynamic conversion is kind of a strange choice, but I like Python, and I intend to use it quite a bit.
However, back in the "old days" (late 80s, early 90s), Perl too was a bastion of those who wanted to throw off the shackles of scripting (I hate when that word is mis-used to refer to interpreted languages). Perl was the way to start giving some real structure to all of those tasks like report printing and systems toolsmithing that had traditionally involved totally unmaintainable "scripts" which eventually had to be re-written in a low-level language.
So what happened? Nothing really. Perl developed some nice features, but ultimately LOTS of bad programmers learned it and in a language that makes it easy for people to write programs, you quickly develop a robust collection of REALLY bad code (along with the really good). Just look at C for confirmation of that.
Python is where Perl was in the early 90s now. Lots of folks who know bad code from good are using it, and it looks like the next great island to swim to. It's easy to look back at Perl and say "it was the dollar-signs that FORCED people to be bad coders," or to look at Java and say, "the low-level types are what SEDUCED people into writing crappy code."
In reality it was the popularity and subsequent influx of bad programmers. Python is becoming popular and I guarantee that in about 5 years Python programmers will be listening to, "[Span, Ruby, something else] is so much better than Python... just look at how much cleaner the code is."
I wonder if we'll ever figure out that joe blow who barely understands what programming is will always produce unmaintainable shlock, no matter what language he writes in and no matter what book on abstract modeling he's just read.