Add :pipe-to typable command that ignores shell output

[alxshine]

I often use an editor in conjunction with a REPL open in another terminal.
To quickly send commands from my editor to the REPL I pipe the editor selection into kitty's remote control feature: kitty @ send-text --match recent:1 --stdin.

In kakoune I could set up a keybinding utilizing the pipe-to command <Alt-|> and pass the shell command directly.
For helix I can only pass the command to typed arguments, but there is currently no typed command to pipe selections to shell commands while ignoring the output.

My suggestion is this:
Add a :pipe-to command which is essentially a copy of the :pipe command, but uses ShellBehavior::Ignore instead of the hard-coded ShellBehavior::Replace.

[QiBaobin]

Is this better to implement in async approach as we don't need the result?

[alxshine]

I can't reply for some reason, so here's my response:

I tried to keep it close to how the shell_pipe_to static command works.
All variants currently internally use the shell function to handle the subshell for the command.
The only outlier is the run_shell_command, which calls shell_impl directly.

If you want the new command to be async, then IMO so should the shell_pipe_to static command.
I don't know if it would be better to add an argument to the shell function, or add a separate shell_async function.
There is an underlying shell_impl_async function, which is currently only used wrapped in a tokio barrier, so a change would not be too complex.

If we wanted we could enable async shell commands, but I'm not sure if thats good default behavior.

[QiBaobin]

I can't reply for some reason, so here's my response:

I tried to keep it close to how the shell_pipe_to static command works. All variants currently internally use the shell function to handle the subshell for the command. The only outlier is the run_shell_command, which calls shell_impl directly.

If you want the new command to be async, then IMO so should the shell_pipe_to static command. I don't know if it would be better to add an argument to the shell function, or add a separate shell_async function. There is an underlying shell_impl_async function, which is currently only used wrapped in a tokio barrier, so a change would not be too complex.

If we wanted we could enable async shell commands, but I'm not sure if thats good default behavior.

Make sense to me, thanks. Although '&' or nohup command & on the command don't work as we might need wait for the shell completes.

[alxshine]

Make sense to me, thanks. Although '&' or nohup command & on the command don't work as we might need wait for the shell completes.

We could create an issue for this, and shift both the named and static pipe-to command to async.
It's just a larger rewrite and not really related to this PR.

That said, I would be happy to work on that once it's clear how such an implementation should be architected in the frame of helix's codebase.