On Feb 8, 2004, at 9:31 PM, [EMAIL PROTECTED] wrote:
Java requires even more verbosity.
This is my general impression of Java. Is the verbosity a good thing or not? It seems verbose to the point of redundancy. Is this helpful, or does it just get in the way?
The answer to that is that it's a matter of perspective, but I don't think that Java would be as successful as it is if a majority of people found it -too- redundant. There's no doubt about it -- the verbosity of Java is overkill for simpler tasks like what you've accomplished with your Python script, especially when there are languages like Python, Perl, Ruby, and bash which can make this kind of thing a lot easier. And, if you know you'll only ever invoke the program on a Unix machine, you can do what another poster suggested and simply glue a bunch of existing Unix tools together like awk, sed, grep, find, et cetera, reducing the burden of actually programming the task to a level of merely asking other tools to perform certain actions on the text. But when things get a bit more complicated, this verbosity can be helpful. (see below)
It took longer to write, even though I had already
prototyped the design in Python (the two designs are nearly identical),
Was it just the verbosity of Java which made it take so long?
Yes and no. No if you mean did the verbosity take longer to type. Yes, because Java requires a great deal more syntax to say the same thing that can be said with less syntax in Python, and subtle issues surfaced when trying to compile the code and then run it. For instance, there were numerous times I was attempting to use a class without having imported it first. Another thing is that the verbosity of Java means that there is more text to be conscious of using correctly, so that right there leads to greater potential to make a mistake.
I wrote that Java implementation in a text editor, which is a useable but relatively primitive tool for a language that can be as verbose as Java. It would have been faster to use an IDE. At work we use WSAD, which speeds development by automatically importing any classes I attempt to use, automatic method completion, popup API documentation (sparing me a trip to the docs), realtime compiler error-checking, and other luxuries. So, using a text editor, I'd make a change and then jump back to the shell and type the command to compile the file. This doesn't take long, because I use keyboard shortcuts and the command history, but then I have to examine the compiler output when there's a problem and jump back to the text editor and fix the mistake -- an IDE will highlight the erroneous code, making it much faster to figure out what you've done wrong.
But Python offers the interactive interpreter, which is a godsend when trying to debug a problem or even just sketching out the script. If you write a python script and invoke it with the -i option (python -i scriptname), then after the script completes you are automatically dumped into the interactive interpreter and any variables in your script are now local to your python session so you can invoke functions using arbitrary arguments, evaluate the values of variables, and other conveniences.
(Yeah, I know that Real Programmers (tm) use vi/emacs/ed, but Real Programmers also don't consider Java a Real Programming Language. ;)
and IMHO would also be more work to modify/extend. That said, if handed a several million-line application written by some other development team, I would rather the application be written in Java than Python.
Why? Performance, cleaner code, more robust language? What makes Java better than Python for some things? What types of things is Java best at?
In this particular case, the reason I say that is because Java is a statically typed language, and Python is dynamically typed. There are a lot of arguments about which is better, and I won't say one is better than the other for all occasions -- but I happen to find a statically typed language like Java to be easier to read once the application exceeds a certain level of complexity. I anticipate some dissent on this topic, mind you. But when I'm trying to navigate my way through a twisted legacy framework of poorly-written source code to find a bug, it's nice to see the type declarations reminding me that foo is a FileManager and bar is a BufferedXmlMessage.
Of course, static type declarations are a pain in the ass in smaller programs, or programs that I'll be writing entirely myself, since I can't enjoy such flexibilities as this:
for item in iterableSequence:
item.doSomething()Where iterableSequence could be a reference to a file object (so "item" would be a line), or a database resultset object (so "item" would be a row), or an object representing a list of users (so "item" would be a user), etc. Unlike Java, Python doesn't require that your objects actually declare that they implement a specific interface, so you can substitute any object that will accept the message you are sending to it ("doSomething") -- the compiler never checks to make sure that "item" is the right type before attempting to send it the "doSomething" message. Of course, if "item" doesn't support that message, you'll get an exception. Which, you can trap if you want, or you can allow to propagate upward.
As for performance, cleaner code, robustness of the language: I do not have any evidence on hand, though no doubt there's plenty to be found on the web. But I would bet that in many situations, Python is comparable to Java in performance, in the sense that the difference between Python and Java is nowhere near as great as the difference between Python and C or Java and C. Clean code can be written in Python or Java, though I would say that Python's easier to read in the short run, but I stand by my assertion that for truly complex applications, Java is easier to navigate. And about robustness, I think they're probably pretty comparable.
The other really important thing about Java to keep in mind is that it's designed to be binary-compatible across platforms. I realize that in reality this doesn't always work quite as cleanly as Sun would like to have you believe, but the very code I compiled on my MacOSX machine when I finally tested the working version of EtcGroupToLdif should work just fine on your Linux box or even a Windows box. Should. There's plenty of stories of incompatibilities, and supposedly it's best to compile your code on the architecture in which it's intended to be deployed. But overall, Java code really can be deployed on just about every major operating system. Now, it's true that most Perl, Python, Ruby, and PHP scripts can too -- and if you write them without making use of OS-specific features, they should work even more seamlessly than Java. But Sun does its best to provide a Java-specific way to do everything. The earlier suggestion about glueing a bunch of command line tools together to do something would be great for a shell script, since that's what it's designed for. It would work fine for a Python script, though Python often provides its own libraries for doing some of those things, and you would really limit yourself to a single OS by using them (Unix). But in Java, it's actually a pain in the ass to step outside of the Java mindset and work in collaboration with OS-specific services and tools. In this sense, Java doesn't really fit in well with the Unix philosophy -- when writing Java code, it's more accurate to think of Java as its -own- platform or operating system that happens to run on top of other platforms. It even provides its own multi-threaded programming model, which AFAIU doesn't directly correspond to the threading model of the OS that the program happens to be running on.
By contrast, if you tell a Python programmer that "Python's too slow for my project", then s/he will probably say "write it in Python, then find the bottleneck and implement that part in C". This makes perfect sense, assuming you're not trying to distribute the software as a wholly platform-neutral shrinkwrapped product. I would argue that most Python programs bigger than a homebrewed script/project are deployed by at least somewhat knowledgeable Python users themselves, who can at least try to diagnose a problem, or who will know which forum/mailing list to turn to for help if there's a problem. I bet they are less likely to require a support contract from an ISV before trying an application. And I bet that in a majority of situations, the software is being custom-developed or customized for a set of end users, so having a truly shrinkwrappable product is not an objective of the development effort anyway.
import java.io.BufferedReader; import java.io.File; import java.io.FileReader; import java.io.IOException;
import java.util.Arrays; import java.util.Iterator; import java.util.List; import java.util.SortedSet; import java.util.TreeSet;
This almost seems rediculous :) 9 library imports, it seems that some should be so commonly used that they'd just be built-in or at least combined into something like stdio.
Each one of these entities is a class, and the classes actually came from only two different packages -- java.io and java.util. In all fairness, I should say that I could have simply written:
import java.io.*; import java.util.*;
It's just that using wildcard imports is a pet peeve of mine, because it means that when someone's reading the code and wants to read the documentation on each class being used, they aren't sure whether TreeSet is from the java.io package or the java.util package. (In this case, it's pretty obvious that a Set comes from java.util, but in giganto classes with dozens of imported classes, it's not always so easy.) Also, if you're working on a substantial Java application then you're probably using an IDE, which does all the importing for you (and even tells you when you've imported a class that's not really being used). IDEs also make it much easier to figure out whether a class came from one package or the other even when wildcard imports -are- being used.
In fact, I have to say this about Java -- it's a great language for very large applications composed of many hundreds or thousands of classes IMHO, but it's also a language that pretty much demands an IDE. I learned Java using a text editor and the command line, but those two tools don't scale up well to real application development, and really don't make the most of your time when there are some great free open source IDEs out there that can basically take a lot of the "work" out of using Java. WSAD is based on the open source Eclipse project, and Eclipse is simply an amazing tool if you're a Java developer.
But I have to say this about Python -- if your project is one that is manageable using Python, then I personally feel it's the better language for the job. It's a lot simpler than Java (or other languages IMHO), and I really believe in keep it simple. But it's got a lot of the things I like about Java, such as a clean OO paradigm and strong typing, and a bunch of things that Java just doesn't offer -- such as the ability to use functions, dynamic typing/method dispatch, and just general ease of use.
Although I know a couple of other languages, and make my living using Java, I use Python for nearly *everything* I do in my own time, and even to write quick tools that I use at work to make my life easier.
Thanks a bunch. Part of trouble with learning languages is the lack of "real world" applications to try them out with. One reason I know perl so well is because the language is designed to do exactly what I do all the time; text munging. Something C and Java aren't especially efficient at it seems :)
Perl is the ideal language for text munging IMHO. It has regexes built right into the language's syntax, a lot of conveniences for reading text and files and reformatting text too. I'd say that between an equally skilled Perl programmer and Python programmer, the Perl programmer will get a text munging task done quicker, simply for the very reason that you mention, which is that the language was designed with this task in mind. But I'm glad you're giving it a try in Python. I personally do all my text munging in Python (and am even starting to forget some of the Perlisms that used to be second nature, such as whether the regex match operator is =~ or ~= , or what the various quoting operators are [qq, qr, qx, etc]) and have no problems, nor do I feel less productive than I was when I used Perl for that purpose. Some things are even easier [for me] in Python, like using an XML parser.
I guess every language has its real strengths and weaknesses. It seems to me that C is great when you need really fast code, but I find it harder to work with than Java in terms of maintaining and debugging problems. Perl is great for text munging, but I hate reading other people's Perl code (except that of Randal Schwartz, who writes really clean and understandable Perl) and I really dislike Perl's OO model. Python has a clean syntax and OO model, but isn't a particularly fast language if you need speed, and I don't think it's quite as maintainable as Java for very very large programs. Java is overall a pretty well-designed language, but not designed for scripting, nor as fast as straight C for most tasks. Some people say that Java's biggest advantage is when an HR dept is looking at your resume.
;)
Erik
_______________________________________________ gnhlug-discuss mailing list [EMAIL PROTECTED] http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss
