How to properly call compinit (and bashcompinit) in zsh

user_2194119 0
Viewed 4131

The whole issue is discussed here and here. Since no one involved in those discussions was 100% sure about the issue, I'm asking for help here. (For the sake of completeness, I'll start from the beginning.)

Let's say we have two scripts (sourced in ~/.zshrc) that set up some completion logic for ZSH. Now based on what I learned, at some point in the script you need to call compinit and bashcompinit like this (copied from NVM completion script)

if [[ -n ${ZSH_VERSION-} ]]; then
  autoload -U +X compinit && if [[ ${ZSH_DISABLE_COMPFIX-} = true ]]; then
    compinit -u
  else
    compinit
  fi
  autoload -U +X bashcompinit && bashcompinit
fi

Apparently, according to ZSH manual, bashcompinit must be called after compinit, (not sure if it's relevant). Now the problem is, the moment the second script calls compinit, the logic coming from the first script is gone (i.e. no completion from the first script is available). A simple snippet to reproduce this is (copied from here):

complete -W "hello world" one
one <tab>   # to see autocomplete working
compinit 
one <tab>   # to see autocomplete NOT working

Someone proposed (here) something like below to solve the issue (by checking if compinit is already called before calling it):

if [[ -n ${ZSH_VERSION-} ]]; then
  if ! command -v compinit > /dev/null; then
    autoload -U +X compinit && if [[ ${ZSH_DISABLE_COMPFIX-} = true ]]; then
      compinit -u
    else
      compinit
    fi
  fi
  autoload -U +X bashcompinit && bashcompinit
fi

Another idea could be to call compinit and bashcompinit not in the custom completion script, but in ~/.zshrc (which hurts the automated installation process for tools like NVM).

I'd like to know what is the correct way to set up completion in general (or specifically with respect to calling compinit).

Thanks.

bash-completion tab-completion zshrc zsh bash
1 Answers

Installing native Zsh completions

Let's say we have two scripts (sourced in ~/.zshrc) that set up some completion logic for ZSH. Now based on what I learned, at some point in the script you need to call compinit and bashcompinit

Nope, that's not what your script should be doing. Not your script, but the user should call compinit (in their .zshrc file) to enable Zsh's completion system. Additionally, it should be called only once for each shell instance.

This is the proper way to add native completion functions to Zsh:

  1. In /usr/local/share/zsh/site-functions, which is in every Zsh user's $fpath by default, create a symlink to each completion function.
  2. Tell the user to make sure following get called from their Zsh dotfiles, in the order given: 
    autoload -Uz compinit
compinit

compinit will then automatically pick up your completion functions from the user’s $fpath.

Installing Bash completions in Zsh

Apparently, according to ZSH manual, bashcompinit must be called after compinit, (not sure if it's relevant).

Yes, it's relevant, but no, not in the way you think it is. Among other things, bashcompinit defines the function complete, which emulates Bash’s complete builtin and can be used to add Bash completions to Zsh. Like compinit, bashcompinit is meant to be called only once per shell.

If your package supplies Bash completions only, then you can do the following to install them in Zsh:

  1. Tell the user to make sure that the following get called from their Zsh dotfiles, in the order given: 
    autoload -Uz compinit bashcompinit
compinit
bashcompinit
  2. Tell them to then follow the same instructions as for installing your completions in Bash.
user_138327690
Related