Again in 2015 I wrote that “Python’s information science coaching wheels more and more result in the R language,” suggesting that the extra critical corporations get about information science, the extra they’ll need the heft of R. Boy, that perspective hasn’t aged nicely.
In actual fact, as a current Terence Shin evaluation of greater than 15,000 information scientist job postings suggests, Python adoption retains rising even because the extra specialist R language is in decline. This isn’t to recommend that information scientists will drop R anytime quickly. Extra probably, we’ll proceed to see each Python and R used for his or her respective strengths.
Even so, if Nick Elprin is appropriate and “2021 is the 12 months through which [data science] will develop into an enterprise-wide functionality that impacts each line of enterprise and purposeful division,” then the language most certainly to dominate is the one that’s most accessible to the broadest inhabitants inside the enterprise.
Recreation. Set. Python.
Fueling the info science increase
The applied sciences and expertise topping the info science charts in 2021 ought to look acquainted:
In any case, they’re fairly just like what we noticed in 2019, as detailed by Jeff Hale:
Yet there are some trends that appear if you squint a bit at the charts. As Shin calls out:
- There is a huge increase in skills related to the cloud.
- Similarly there is also a large increase in skills related to deep learning, like PyTorch and TensorFlow.
- SQL and Python continue to grow in importance, while R remains stagnant.
- Apache products, like Hadoop, Hive, and Spark, continue to decline in importance.
Easy does it
Dig a bit deeper, and the technologies/skills that seem to be growing fastest are those that are easiest to learn. Hence, while TensorFlow and PyTorch both saw growth, PyTorch’s growth significantly outpaced TensorFlow, for reasons I’ve outlined before. PyTorch’s popularity is starting to play out in the projects themselves, too, with cumulative PyTorch contributors set to exceed the number of TensorFlow contributors in the near future (whereas the number of contributors to PyTorch over the last 12 months already surpasses that of TensorFlow).
A few years back Redmonk analyst James Governor decreed that “convenience is the killer app” where developers are concerned. From MongoDB to Fastly to GatsbyJS, our go-to defaults across a wide range of technologies are those that enable developers to become productive faster.
Which brings us back to Python. And R.
R remains highly relevant in data science, something that we shouldn’t expect to change in the near future. Yet we’ve seen far more data scientists switch from R to Python than vice versa (twice as many, in fact). Reasons include better usability, performance, ecosystem, and more for Python, argues Emmett Boudreau. R remains broadly used for statistical computing, but as more and more companies (and their developers and data scientists) embrace data science from a technical, not scientific, standpoint, Python will continue to soar.
Copyright © 2021 IDG Communications, Inc.