I’m sure a lot of people have seen and tried Google Code Search by now, so I’ve had a look and compared it with my favorite code search at Krugle which has saved me hours of work on many occasions. The interface of Krugle is clearly modelled on Eclipse, but it handles plenty of languages alongside Java and covers a few extremely useful bases which Google doesn’t. The two I find most useful are thye API/TechDocs searches which allow you to find easily how something works, and the function search which allows you to quickly find the definition of a particular function. They are incredibly useful for finding your way round a new project, especially where the documentation is still in flux.
My impression of the Google Code Search is that it works as a standard Google search over the domain of software, so it doesn’t take advantage of the domain knowledge offered within the source code and related documentation. I have a feeling this may be a trend to come. Specialised searches within a domain can exploit knowledge and connections that are hard to generalise to other domains. Krugle is a great example of how this extra knowledge can really make searching easier and more powerful. The generic Google PageRank just can’t cut it, at least not on the basis of what I’ve seen.