ast VS Shell

I suspect the intended point of comparison is the monolithic Unix of old, where the entirety of the code of the system is (hopefully, ostensibly, allegedly) developed with a single vision, except for the applications in the narrowest sense of the word (e.g. the particular piece of numerics you need to design your airplane or whatnot, but not the editor you used to write it, the cluster scheduler you used to run it, the typesetter you used to write down your results, or the 3D viewer you used to present them).

One example that’s just plain interesting to read to see what things people were exploring (because, let’s admit it, historical Unix source gets plain boring after a while) is AT&T Research’s https://github.com/att/ast.

Aha, I shall have a dig through https://github.com/att/ast/tree/master/src/lib/libtksh then.

Thank you kindly.

Concerning the nested typedefs, yes, it's mainly just about nice code organization for me, but there are definitely other usecases where it's pretty essential, I'd say mostly in generic programming. I stumbled hard on this one: when writing a linked list - for each linked list variant, you have a node type, and a handle type that carries reference to the first node and other metadata like list length etc. . But you need to somehow be able to deduce the concrete handle type from the concrete node instance, so that the user can obtain handle instance by calling list_init() macro on the node and do other convenient stuff. Best solution I found so far how to solve this runtime-overhead-free in og GCC C is by defining zero-length array of the handle type inside the node type (like here) and then using __typeof__(node->_handle_typeinfo[0]) or something like that xDD, which is just so insanely ugly.