radare2-shell-parser: command substitution

[ret2libc]

A feature that still makes me think a lot is how we want to handle command substitution. I would like to have some feedback and thoughts about it, so that I can implement it in radare2-shell-parser, as the way we implement it will change a lot how the parser works.

1. Commands substitutions can appear anywhere, they are just string-replaced.
   Example: p$(?e "d 10~xor") becomes => pd 10~xor
2. Commands substitutions can only be in arguments and they can generate new arguments
   Example: o $(?e "myfile 0x4000 rwx") is interpreted as o myfile 0x4000 rwx.
3. Commands substitutions can only be in arguments and they can't generate new arguments
   Example: o $(?e "myfile 0x4000 rwx") is interpreted as o "myfile 0x4000 rwx" (which would not work because o expects 3 arguments.
4. Commands substitutions can only be in arguments and they can generate new tokens or new commands as well.
   Example: f $(?e "ciao @ 0xdeadbeef; f-") ciao is interpreted as f ciao @ 0xdeadbeef; f- ciao

Other ideas? Suggestions? What do you think is the clearest/best thing? Bash seems to use a sort of 1/2.

[ret2libc]

I'm going to try 2.

The reason I don't like 3 is because I think it makes sense for a cmd_substitution to generate more than one argument token.
The reason I don't like 4 is because if you allow the cmd_substitution to generate any kind of new token, you can hardly parse the input because until you execute each substitution you don't know the structure of the final input.

[XVilka]

I think restricting them to just arguments->arguments substitution is good enough. I don't remember any case when I needed something more advanced (and if you need something more advanced - just use macroses or r2pipe).