[stevenwiles]

> The reason is that I think it can still be useful as training data for, say, designing new programming languages/tools around coding patterns seen in the wild.

Do you happen to have any evidence of that ever being done even once in the history of mankind?

It's a nice idea, but you should know by now the world isn't a slave to your desires.

[mcbits]

Here is some evidence of mining source code for common structure and idioms. I didn't even have to dust off my bull whip: http://homepages.inf.ed.ac.uk/csutton/publications/idioms.pd...

I don't have evidence of this approach actually being used to inform language or tool design, but it's not like it's an outrageous concept that nobody's thought of before. Your hostility is perplexing.