Bad Metrics for Trends in Programming Languages
Was reading a posting about Trends in web development according to Google one section got me me thinking. Their trends show end user behavior as it relates to popularity. However, when I think about language popularity I think in terms of programs written or lines of code produced… The Google trend I think is a misplaced metric.
Specifically, when I’m writing various languages I’ve got these behaviors:
- PHP – I’m frequently found typing the following query “php strpos” or “php call_user_func”. Basically using “php” as a google keyword to get me to man page about the specific php function. It’s fast and easy.
- Perl – My man page behavior is a bit different.. it’s “perldoc -f strpos”, since perl has all of the documentation built into the system with simple command line tools to bring it forth.
- Java – Very little of java is focused on language issues, its all about the libraries. You’re going to be doing a query like HashMap without the Java qualification.
- C++ – Very similar to Java – you know the language, it’s queries like “STL list iterator” or other library or feature specific bits. Sure, once in a while I might use “C++ abstract template” but when you’re in C++ land usually the word “template” and %$(#*(#^ — that’s the current g++ error on your screen — is sufficient to get the right help
- JavaScript – Can’t really leave it out, it is a language that’s web search centric and I do qualify my queries with JavaScript to get to the search results, but I’ve found that this isn’t very good (the pagerank for the site that has the good info is low)
That’s it, unless you do an aggregation of all of the sub-search terms I think it’s premature to use Google Trends as a valid qualification of true language trends.