Which one will be the dominating programming language for next 5 years for analytics , machine learning . R verses python verses SAS. Advantage and disadvantage.

  • 5
    $\begingroup$ This is probably not a great question for StackExchange. It's going to be mostly opinions and speculation. $\endgroup$
    – Sean Owen
    Aug 24, 2014 at 9:24

2 Answers 2


Due to the very Big increase in Big Data (pun intended) and the desire for robust stable scalable applications I actually believe it to be Scala. Spark will inevitably become the main Big Data Machine Learning tool, and it's main API is in Scala. Furthermore you simply cannot build a product with scripting languages like Python and R, one can only experiment with these languages. What Scala brings is a way to BOTH experiment and produce a product. More reasons

  1. Think functionally - write faster code and more readable code
  2. Scala means the end of the two team development cycle. So better product ownership, more agile cross functional teams, and half as many employees required to make a product as we will no longer need both a "research" team and an engineering team, Data Scientists will be able to do both. This is because Scala is;

    • A production quality language - static typing, but with the flexibility of dynamic typing due to implicits
    • Interoperable with rest of Java world (so Apache Commons Math, Databases, Cassandra, HBase, HDFS, Akka, Storm, many many databases, more spark components (e.g. graphx, SparkStreaming)
  3. Step into Spark code easily and understand it, also helps with debugging

  4. Scala is awesome:

    • Amazing IDE support due to static typing
    • Property based tests with ScalaCheck - insane unit testing
    • Very concise language
    • Suits mathematicians perfectly (especially Pure Mathematicians)
  5. A little more efficient as compiled not interpreted

  6. Python Spark API sits on Scala API and therefore will always be behind Scala API

  7. Much easier to do Mathematics in Scala as it's a Scalable Language where one can easily define DSLs and due to being so functional

  8. Akka - another way other than storm to do High Velocity

  9. Pimp my library pattern makes adding methods to Spark RDDs really easy


There is a great survey published by O'Reilly collected at Strata.

You can see that SAS is not widely popular, and there is no reason why that should change at this point. One can rule that out.

R is barely ahead of Python, 43% vs 41%. You can find many blogs expressing the rise of Python in data science. I would go with Python in the near future.

But 5 years is a very long time. I think Golang will steal a lot of developers from Python in general. This might spill over to data science usage as well. Code can be written to execute in parallel very easily, which makes it a perfect vehicle for Big Data processing. Julia's benchmarks for technical computing are even more impressive, and you can have iPython like stuff with iJulia. Hence Python is likely to lose some steam to both. But there are ways to call Julia functions from R and Python, so you can experiment using best sides of each.


Not the answer you're looking for? Browse other questions tagged or ask your own question.