[rbranson]

If you want to learn fundamentals, you need to master C. The language itself is very simple and readable, but what it will teach you is how memory works.

Every algorithm & data structure worth implementing has been implemented in C and is probably available in an open source project. You can download Redis, which is fantastic code and covers all the basics. Take one of these projects and start trying to mold it to your will.

Write specialized string parsers in C for doing simple things you'd normally do with a script. If you do it right, it'll be several orders of magnitude faster.

Learn to use the GNU debugger.

Take blocks of algorithmic code you've written, learn pthreads, make them concurrent, and actually able to achieve linear increases in speed as you add cores.

Your greatest ally will be well-written open source code.

[bartonfink]

"Write specialized string parsers in C for doing simple things you'd normally do with a script. If you do it right, it'll be several orders of magnitude faster."

Note that this ONLY makes sense in the context of a learning exercise. If it's simple and you'd normally do it with a script, the performance you'd gain from a C rewrite isn't worthwhile. The knowledge you'll gain from making C do something it doesn't make easy, though, is.

[rbranson]

In most situations, I definitely concur. There are a few cases where I've had to process >10g flat files and I whipped up a specialized tool in C that could run through it nearly as fast as the disk could feed it.

[bartonfink]

Sure, but I doubt you'd seriously try to handle that with pipes and Unix anyway. 10g warrants someone who knows what they're doing and knows, like you, when the overhead of writing in C pays for itself in faster processing.

[ansgri]

Unless you're in scientific community, where it's common to have to convert GBs of data from one weird format to another, or possibly analyzing them, nevertheless knowing only e.g. R and Matlab.