Google programming Frankenstein is a Go

Google I/O Google's back-end infrastructure is now running applications built with its new-age Go programming language.

"We're already using Go internally at Google for some production stuff," Robert Pike, one of Go's developers and a former member of the Unix team at Bell Labs, said today during a question and answer session at the company's annual developer conference in San Francisco. "The language is pretty stable."

Speaking with The Reg, Pike declined to describe these systems. "But I'm not lying," he told us. "This is real stuff, and we're using it."

Google open sourced Go in November, describing it as an effort to crossbreed a dynamic web-centric language like Python with a compiled language like C++. "There is a growing rebellion against cumbersome type systems like those of Java and C++, pushing people towards dynamically typed languages such as Python and JavaScript, [but] some fundamental concepts such as garbage collection and parallel computation are not well supported by popular systems languages," the company said in a blog post announcing Go.

"We believe it's worth trying again with a new language, a concurrent, garbage-collected language with fast compilation."

The language was originally designed for server-side applications, but Pike said it could also be put to use on the desktop and even on mobiles. At this point, its desktop options are limited because the graphic libraries "aren't quite there yet," and mobile hasn't really happened yet either. But Pike hinted that Go will eventually be used to build apps for Android. "ARM, after all, is one of the platforms supported," he said.

The idea, Pike told us, was to build a programming language from the ground up that does all the things the project's developers wished other languages did. It borrows a bit from C, Oberon, and Newsqueak, a language Pike developed back in the 80s, he said, but for the most part, it's a new animal.

The development team also includes Unix founding father Ken Thompson.

Go is specifically designed to run software on multicore machines. Pike describes it as concurrent as opposed to a parallel language. The language is built so that an application can restructure itself as it expands across a parallel environment. The idea is to make it easy to build applications that can scale across a distributed infrastructure. In other words, it's very Google.

Go is object-oriented, Go developer Russ Cox explained today, and not type-oriented. The type system has no hierarchy, and there's no inheritance. Unlike with a Java program, where you have to specify a hierarchy, the Go compiler can determine the relationship between types.

Cox said that Go is so different from something like Java or Python that you can't really translate programs written in these languages into native Go. Google also boasts that the language is fast — at compile-time and at run-time. In a video announcing the new language in November, Cox showed off the Go math package, packed with about 1,000 lines of code across about 20 files. It built in about 20 milliseconds, and it could be tested in under a second.

Today, Cox and Pike also said that they have completed a Go compiler for Native Client, the Google plug-in that lets you run native code inside Google's Chrome browser. But it would seem that very little application work has been done here. Native Client is now bundled with Chrome, and the company reiterated yesterday that it would be an "important" part of its upcoming Chrome OS, the netbook operating system based on the browser. ®

Update: This story has been updated to show that Robert Pike told us that Go borrows some ideas from Oberon as well as C and Newsqueak.