Pierre Gaston wrote: > I have a couple of suggestions about coprocesses. > If I understood correctly how coproc works, I think that > instead of : > coproc [NAME] command [redirections] > > the documentation would be a little clearer with something like: > > coproc simple-command [redirections] > coproc NAME compound-command [redirections]
I agree. I will make it clearer that NAME cannot be used if the coproc command is a simple command, to avoid confusion with the first word of the command. > The other suggestions is to change the index of the descriptor array, > I find much more natural to think about NAME[0] as the descriptor > number connected to the standard input (0) of the coproc NAME and NAME[1] > as the fd number connceted to the standard output (1) of the coproc NAME > rather than the opposite. > (Maybe you have another point of view on this) It's a shell-centric implementation. You read from the first descriptor and write to the second, just like a pipe, so the shell reads from NAME[0] and writes to NAME[1] to communicate with the coprocess. The coproc doesn't have access to NAME at all. > Also coproc is not documented via help. Good catch, thanks. Chet -- ``The lyf so short, the craft so long to lerne.'' - Chaucer Chet Ramey, ITS, CWRU c...@case.edu http://cnswww.cns.cwru.edu/~chet/
