Deciding on a language: Python or Java

Question

I am part of a small (read: two programmer) startup, and we are about to start developing a large (we estimate it will be about a year before customers will see anything) web application… And it's come time to pick a language.

Most of the experience I have is with Python, most of my co-worker's experience is in Java and C++, but they would like to learn Python.

But, if that was all there was to it, I wouldn't be posting here :)

My boss is worried that, if we commit to developing in Python, we will be increasing the risk that it will be hard to find new programmers or replace us if we get hit by a bus.

I have opinions on this matter, but I don't have any experience to back them up... So I'd like to know:

• Are his concerns valid?
• Do you have evidence that it's hard/easy to bring a new programmer on board if they don't have experience with the language/framework you're using?
• Have you been in a similar situation, and have some wisdom you could share?

Thanks!

Edits

• Thanks to dmckee for pointing out a similar question, http://stackoverflow.com/questions/954164/choosing-between-java-and-python
• It would be easier for my current co-worker to learn Python than my replacement, because my current co-worker will have me sitting next to her (not to make myself out to be anything particularly special – but it's easier to learn something if you've got a teacher with experience)
• Some have suggested that my boss is asking a silly question… But the reason I'm asking here is that I've got a lot of respect for him and his business sense (something I, as a programmer, have less of), so I want to take his concerns seriously and give him a good (read: "reliable and devoid of superstition") answer.
• I don't want to suggest that we'd want to hire the next idiot Java developer… But I believe that there are more "good" (read: "maybe not superstar, but solid and competent") Java programmers in the world than there are Python programmers.
• A few people have suggested that statically typed languages are easier to maintain. I'm willing to believe that, but only if someone can point me to evidence which supports it (I have maintained a large (25k LOC) Python application, and I don't think my task would have been significantly simpler had it been statically typed)

Accepted

Thanks, everyone, for your responses – they are much appreciated.

After reading through all the answers and linked articles, I have come to these conclusions:

• When dealing with an established language, the most important factor should be experience (as learning a new toolkit will cost a lot of time, regardless of the language).
• Without exception, everyone has said that any competent programmer can pick up Python (or any other sane language) in relatively short order, so we need not worry too much about being hit by a bus.
• Assuming equal experience in Java and Python, Python seems to be the favorite choice for all the reasons you would expect: time to market, average programmer quality and syntactic clarity/expressiveness.
• Some good arguments have also been made for Java, in that it's got more "stuff", especially enterprise "stuff", available (Python people: if you can't admit this, you're in denial), there are more experienced (and competent) Java programmers and Java can be faster.
• Static typing, however, is not a good reason to consider either language, as for every person who says they think it's helpful there is another (or even two others) saying you're fine without it.
• Python (or any language which can run on the JVM) is a good choice, because it leaves the possibility open of interacting with Java code down the road.

Again, thanks for the responses – they are very much appreciated.

Answer 1

Obligatory TIOBE index reference. Python is definitely a safe bet.

Answer 2

Python, of course. Get your boss to read those two great Paul Graham essays Beating the Averages and The Python Paradox.

Answer 3

I would go with Java. It's probably worth the initial inconveniences in the long run. With a broader set of tools and available labor, it's probably the safer bet here.

Answer 4

I've been there before. I had a simple web app to write once upon a time. The choice was between PHP and Python. I had started writing it using the turbogears framework in python. It was fun, learning python was a lot of fun, its a neat language.

Unfortunately, my boss thought the same thing as yours. What happens if I get lost in the amazon and no one is here to maintain it? Where do we get a python programmer? My response would be to let anyone tackle it, they'll probably have as much fun to learn the language as I had. But that did not satisfy him much...so it ended up getting redone in PHP instead. We have a whole team of PHP developers already on staff, so finding someone to pick up where we left off wouldn't be that big of a deal. Of course later on, we end up getting someone to pick my work up that didn't know python or php, so in hindsight it didn't really matter what language i had picked.

As far as the choice between java and python...i'd pick python.

Answer 5

My advice: Don't do a large application in a dynamic language. Maintaining and refactoring when the line count goes to 10k and above is not fun. The compiler is your friend here.

Answer 6

1. Yes, his concerns are valid, but he has no cause for concern as there are plenty of Python developers around

2. If you don't have much experience of cooking, is it going to hard for you to learn a new recipe? Yes. Would you give up and not bother at all? No. Same goes for coding, it'll be a learning curve for them, but you'll always find somebody to do the work.

3. Your boss pays your wages. He sells the software you code. If it takes you longer to learn and code in Python, which is inevitably will, given the fact that there is a learning curve, then your boss sells less, he makes less money, your wages don't increase as much as they could do. Would you rather get paid less and learn a new language, or get paid more and learn a new language in your own time?

Answer 7

I think his concerns are valid considering that c++ and java developers can be found anywhere. Take a look at google they were forced to hire the creator of python. If I were your boss I would do the same thing. But, on the second note python is becoming more popular everyday and more and more websites out there are being developed in python.

Answer 8

In my opinion, a "good" developer will be able to pick up any language... part of being a good developer is to know how to learn a new language.

Choosing between web languages depends entirely on the problem at hand. It's important to understand what supporting systems you application will run on. If you are going to have a MSFT Sql 2008 database on Windows server 2008, then you may want to consider asp.net and c#. If you are going to be running on linux, MYSQL and apache, another language may be a better fit. You'll find very large websites using any number of web languages:

MySpace: asp.net
FaceBook: php
drop.io: ruby

With python, you'll be working with a dynamic language which makes it absolutely imperative that you unit test. Java is statically typed and having the complier is a nice safety net to have in place. It's important to NOT pick Java just because the majority of your developers are comfortable developing in it. Pick the language that is best for your problem set :-).

My boss is worried that, if we commit to developing in Python, we will be increasing the risk that it will be hard to find new programmers or replace us if we get hit by a bus.

A good litmus test would be to go to monster.com and find all resumes that have [enter language here] on their resume. It may give you a better handle on the risk. But again. A good developer will be able to pick up a new language fairly easily.

Answer 9

Fear and Loathing and Language Politics

If your boss wants to hire programmers that can't learn a new language then that's the problem.

But getting a smarter boss is probably not on the table, so how about this:

1. Agitate for Scala and Lift or Ruby on Rails.
2. The boss complains.
3. Agree with the boss that it's a bad idea. Try to keep a straight face.
4. Compromise on Python.

:-)

There is an argument for Java, btw, but it's not: "I want to use the most popular language because that's what the grads know." Yes, a less popular language has fewer job seekers, but also fewer openings, so it all approximately balances out.

But back to the real argument for java...many other answers reference the conventional wisdom that statically typed languages are better for a large codebase. It's probably quite true, but startups are risky. I think becoming successful would trump any hypothetical what do we do down the road when we are the-next-netscape worry. Gee, in that case you have already won, so go find some great recruit, not just the next warm body that "knows" java.

And I don't think Twitter is going to have the slightest problem finding a Scala developer. In fact, I don't think anyone is having trouble finding developers.

Answer 10

Before this starts the whole "my language is better than your language" debate, let me state up front - I currently write in Java. I have also written in C, C++, Perl, Pascal and a little in PhP (and am currently learning Objective C). If it was my decision, I would choose Java. The reasons are:

1) The best IDEs for Java are free (Netbeans and Eclipse) 2) Being typesafe on large projects is always a plus and a benefit - let the compiler do as much of the work for you as it can 3) In my experience, doing large projects in type-agnostic or soft-typed languages can be much harder than not, maintenance included (and according to Martin Fowler, mainenance is about 90% of the cost of software). 4) The libraries for Java are very rich, and the domain space is very well understood.

Now, do some of these apply to Python? Most likely. I cannot say personally. However, I do like Java as a language, and well written Java server code will perform as well or better than well written server code in another language - I have proven it several times at my current employment.

If the developers believe that the best technology is Python, then you should use Python. However, that is a technology decision that should be made based on available resources, what technology is going to be used to do what, etc.

And of course, the best advice I can give to anyone is to write good code no matter what language you end up using. Some languages are more forgiving of bad code than others, however if you and your team are disciplined about best practices and the like then the language will matter less than the eventual code base that you end up with.

Answer 11

Why not use take advantage of both platforms by using Jython ?

Write your lower-level entities and core objects in Java, and tie together in an agile, flexible fashion using Jython. That way you can easily leverage off the benefits of Java (the language) - static typing, associated unit test frameworks etc.

But using Jython to tie everything together means you can very quickly reassemble your project using the core objects but change the dynamics of it.

Answer 12

IMHO, anyone suggesting you don't need unit tests in a static typed language (i.e. Java) is highly suspect.

Answer 13

Why not use both? This maybe silly but alot of companies already did this.

1. Develop the backend (db related and service related) with Java
2. Developer the front-end (customer facing) with Python

EDIT

The rationale behind this is:

You want to pick up the statically type of Java. You want to make sure what goes in to the backend. You (usually) don't want a dynamically typed language for your service/back-end (especially if it is critical) because you don't want something creeps in your back-end without knowing what type it is. You can ensure what goes in to your backend/web service method beforehand with your unit testing. Another reason is normally backend does not change alot. You want the backend to be stable, robust and stay for a long period of time. Companies (usually) invest alot on the backend, because that is where the business logic lies, that's the part that produce cha-ching.

On the other hand you want something that enables you to change quickly on the front end. Because front-end is the layer that has human interaction. From my experience, front-end technology tends to change really quickly, human interaction requirements changed very fast and using Java for the front-end would be a complete waste (waste of time and money). Compare how productive it is to build a front end with any Java framework and dynamically typed language framework (e.g: rails, django, etc). And also another reason is that most of the time the testing for front-end is functionality testing because you want human interaction and feedback for the front-end. That's why I reckon dynamic language is a perfect fit for this.

Twitter is one example that took this approach. They develop the backend using Scala while the front-end is still using Ruby on Rails. Facebook is also another example that uses PHP on the front end and their data storage (Cassandra) is built using Java.

So I still reckon you are picking up the best of two worlds with this approach.

Answer 14

Time to market.

One of the main reasons why I propose python as a solution is the short time to market needed to produce an application. Java is so painful to develop with. Yes, you end up having a lot of ready to use jars, but in the end, the code you write is just gluing between these libraries, gluing that is painful with its static nature. You end up using all the Design pattern book not to write logic code, but just to fight against the static nature of the language.

On the other hand, python is faster. You can develop a working prototype in no time, and add on that prototype again in no time. The code you write is logic, not scaffold, and you don't need a World War II german tank parked on your RAM to ease all the useless typing you would have to do to do the most trivial things. Yes I am talking about Eclipse.

Java, on the other hand, probably scales better on a large number of developers, but my experience in a company with 2 teams of 6 developers each and around 100k LOC was very positive with python.

Answer 15

...the risk that it will be hard to find new programmers or replace us...

Hard to find programmers? In this market?

I think you have your answer right there.

Answer 16

I doubt that it will be much harder to find qualified Python developers. As mentioned good programmers can pick up other languages quickly, so I don't think replacements will be hard to find if you go with Python. That said, if I were you, I would show your boss why you think Python is the better choice over Java. Give them something that outweighs their smaller job pool argument.

Answer 17

Old (2006), but good, from Joel:

"...the bottom line is that there are three and a half platforms (C#, Java, PHP, and a half Python) that are all equally likely to make you successful, an infinity of platforms where you're pretty much guaranteed to fail spectacularly when it's too late to change anything (Lisp, ISAPI DLLs written in C, Perl), and a handful of platforms where The Jury Is Not In, So Why Take The Risk When Your Job Is On The Line? (Ruby on Rails)."

and

"So for now, my advice is this: don't start a new project without at least one architect with several years of solid experience in the language, classes, APIs, and platforms you're building on. If you have a choice of platforms, use the one your team has the most skills with, even if it's not the trendiest or nominally the most productive."

Anyone who can program can get up to speed on a well-written code-base in any sane language. It's designing and creating that code-base that is the trick.

Just make sure you have an accurate self-assessment of your skills if you think you are up to the job of doing that for a large system.

Answer 18

I think everyone here has made some good points about why either choice could be a good idea.

At some point you and your boss are going to have to choose.

I suspect your boss is challenging you with the assertion of increased risk in order to have you present a compelling case for the choice.

A good way to respond to this challenge is to present an objective summary of the pro's and con's of either choice. Relate the pro's and con's to your business. Remember, compelling points that just don't relate to what you're doing are useless!

It's worth being honest about the pitfalls of either choice.

It's important to understand and appreciate your bosses concerns and openly address them, even should they not support your preferred option.

A little information on the job market trends for language skills could be worth including to offset the hit by a bus concern. In fact, demonstrating a commitment to relevant documentation is a great way of addressing that too, regardless of language choice.

I'd even include a few close but no cigar options that you've dismissed, with a brief explanation as to why, before doing the more in-depth analysis on these final two contenders.

Good luck, I hope we hear more in a year's time!

Answer 19

It's not learning the language per se that will hurt you. Anyone can learn Java or Python quickly. It's all the little bitty decisions about architecture. If you use Java, will you use Spring, Struts, Seam, or something else? Why? Will you use an ORM? Why, and how will it hook in to the layers above and below it. With Python, will you use Django, TurboGears, Zope?

In your first version, you'll probably settle on what you can get working in a reasonable amount of time. Odds are that there could be a far better way to do something but you just don't know it because you're not experienced enough in it yet. After a year you'll probably look back and say, "Holy smokes! Why did I do it that way?"

You should consider it a head start to use the language you're more familiar with, because you've already made all the silly little mistakes. If both of your levels of experience are the same, then you're probably fine either way.

Finding new hires either way should be no problem. At my current place, only 10% of our new hires have had any experience in our language, but the architecture was designed by a very experinced person. The new hires have been able to pick up things quickly.

As long as you, the more experienced, set the architecture you will be fine. Even within a couple of months of you beginning to code, a new hire should be able to pick up what you've laid down and run with it.

Answer 20

I would go for Java -- better performance, better stability.

I know the whole Spring/Hibernate/J2EE is painful when you want to do something simple, but, as you satisfy more complex requirements you begin to realise that they thought of everything except ease of development.

My signal personal experiece might be what you could call the J2EE paradox. I had a pretty complex application written mostly in php with some vanilla Java backend. We needed to ship this out to other parts of the company but we didnt want them messing with the source code, or, get into a fight with thier architecture people about implementing a php/apache system in a universe of J2EE/Websphere so the decision was mad eto refactor the php front end to J2EE. This took just over a month (there is after all no better spec than a working piece of code!) the real surprise was the Java code base (by line count) was smaller than the php code base (including spring/hibernate config files as code). This did not really make sense until we looked more closly at the code. Java exception handling accounted for about 20% of the Java code. But most of the mvc and sql handling just disappeared or was done by a single line in the .xml files.

Even the exception handling bloat turned out to be a blessing. We got proper exception messages which identified the line of code and the problem. With php we either got "variable xxx not set" or nothing on the page. So debugging in the Java code took much less time.

J2EE is just to big a beast for small projects, but its just the job for industrial strength large projects.

((I suspect .NET is actually better but have limited experience of it.))

Answer 21

I had the same decision.

I chose Python.

Why? The flexibility. Our requirements were -- at best -- murky. Realistically, they were incomplete. If you want to take an extreme position, they were little more than lies told by a vendor to an executive (who'd believe anything.)

The rest of language choice is almost a tie. Specifically, finding skilled people is a wash -- all skills are equally hard to staff. We're a staffing company and it's as hard to find Java people as it is to find Python or Informatica.

Bottom Line: Good people are hard to find, irrespective of skills.

Answer 22

Are his concerns valid?

They might be in many languages (I'd be especially worried about functional programming languages: intrinsically wonderful but VERY hard to get people with real-life programming experience in, AND a huge hurdle to learn for most experienced programmers). They're not for Python -- any superb programmer can pick it up with incredibly small effort if they want.

Do you have evidence that it's hard/easy to bring a new programmer on board if they don't have experience with the language/framework you're using?

I could quote (but you could not verify;-) several situations of superb programmers being hired, early in Google history, mainly on the basis of their strength in Java and/or C++ -- and who easily picked up Python (Google's key general purpose languages are C++, Java and Python) and in some cases coming to see it as their primary language (though most probably don't go further than my own mantra: Python whenever I can, C++ if I really must -- maybe s/C++/Java/ in some cases;-).

We're not talking about run-of-the-mill programmers, the kind who are terrified of having to learn a second programming language (horrors!-): if that's what your management is daydreaming about -- cookie-cutter mediocre programmers that they can hire by the gross and lay off any time the phase of the moon is wrong -- Python's probably not their best choice (if that's the kind of people they're gonna be hiring, they're not going to make it big with ANY language, but they're likelier to get by with extremely popular languages).

As Paul Graham (whose ideal language is his own dialect of Lisp, NOT Python) points out in an essay that's already been quoted, you're going to get better programmers if you demand Python expertise (or, I would add, Objective C, Ruby, Erlang, Haskell, F#, Scala, O'Caml, Eiffel, APL-2, SML/NJ, Mozart, Boo, Scheme, Alice, LUA, Prolog, ...!-) than if you demand expertise in a "mainstream" language such as Java -- that's simply because the people who don't program because of a raging inner fire, but to put bread on the table, are NOT going to go for "niche" languages, they're gonna stick with the "tried and true" that's going to get them a temp job at burger-flipper hourly rates (which may be more than they deserve given the effectiveness of such programmers -- productivity ratios best:worst of 30:1 or more have been reported in many empirical studies!-).

Any SUPERB programmer is not going to find any difficulty learning Python (though they may not want to work in it if they can make huge incomes working in their chosen favorite language[s], of course;-). MEDIOCRE programmers wouldn't either, alas!-), were they to devote a modicum of effort to it, but apparently (let's hope it stays that way) most haven't, yet, "thanks" to the Python's community lack of interest and/or ability in marketing and evangelism (for a language to become a serious buzzword it needs either a huge marketing spend, like Java got back when Sun Microsystems still had money for that, or a community of enthusiast practitioners some of whom accidentally happens to have real marketing flair, like Ruby got in its time -- and other languages before then [Pascal, say, and C++]).

Have you been in a similar situation, and have some wisdom you could share?

I've often found myself "selling" Python to upper management A LOT back when I was freelancing (mostly based on it, though I could do many technologies, C++ first and foremost, and side issues such as technical management, development methodologies, &c, if the need arose). At that time there was no YouTube to point to, and Google's vast use of Python was hardly known at all (I learned about it only because my publisher, O'Reilly, let me know that the largest private purchaser of my Python books was this "Google" privately held firm.... thanks Tim!-). Nowadays, what with YouTube's reliance of Python being widely known, reddit.com's use of Python to port their originally-LISP code, Google App Engine, etc etc, you've got a ZILLION more useable "reference customers" than I did;-).

I didn't always succeed, I even sold and executed some consultancy contracts based on C++, MFC, win32, etc etc (same [hefty;-)] hourly rate as I charged for Python, but so MANY more billable hours per delivered function point, that I just couldn't resist the monetary allure of such contracts...!-), but, mostly, I did fine. Having experience and credibility in management roles as well as developer ones no doubt helped -- I could and did spin the same kinds of points I'm making here into profit-and-loss and return-on-investment terms.

Wisdom? I guess I should, given my advanced age, but I'm not gonna claim any -- just arrow scars in my back (you can always tell the pioneers by the arrow scars on their back!-) and the smile on my face (hinting at the success that Python brought into my life and the technical vicissitudes of every single company I was ever able to introduce it in;-).

Answer 23

In my experience, Java is good for building cathedrals over the spans of centuries. If you want to be more agile than that consider Python (JPython to get the JVM).

See http://www.paulgraham.com/avg.html describing what made their start-up a success - using Lisp :)

Answer 24

I agree with all the points above. They're all valid.\

The real answer is it depends. I firmly disagree with the argument 'I won't be able to hire.....'. This simply does not pan out to be the case.

A year sound like an awfully long time - particularly if you're not getting any customer feedback in that time - this seems to be a much higher risk than which platform you choose.

In terms of productivity however, I find that a resonable programmer can be productive in Python in about a day. The expression of business scenarios os clearere also, as Python is a much more transparent language than Java - the question becomes 'how can I best express this in Python?', as opposed to 'how can I translate this to Java?'

In any event, the only way you're going to find out is to give it a go. I think you'll be stunned.

Good luck.

Ben

Answer 25

There are plenty of Python programmers out there. Your boss needn't worry about being able to replace you. Furthermore, Python is a very easy language to learn, so even bringing on people with no Python experience shouldn't be a problem; I've taught Python classes and had complete newbies (with very little programming experience, even) writing non-trivial programs after just a day.

I develop in both Java and Python on a daily basis, and I can say without hesitation that I would choose Python for a web application without hesitation.

If your boss is worried, show him some examples of successful Python usage. See this for more examples.

Answer 26

One thing. If you take group of Java programmers and Python programmers - percentable, with no offence - the second ones will be more skillfull and smart. But It would be easier to find typical programmers (not genius one) with lower salary for Java, not for python.

Answer 27

Java.

If I were the customer, I will pick the Java-based product instead of Python-based product. Simply because Java owns a very positive image of being high quality product.

Thanks to Sun's effective marketing strategy :)