Zsh Mailing List Archive
Messages sorted by:
Reverse Date,
Date,
Thread,
Author
Re: Ansible tab completion: re-read cached hosts & groups on directory change
- X-seq: zsh-workers 50033
- From: Daniel Shahaf <d.s@xxxxxxxxxxxxxxxxxx>
- To: Lawrence Velázquez <larryv@xxxxxxx>
- Cc: Moritz Bunkus <mo@bunkus.online>, zsh-workers@xxxxxxx
- Subject: Re: Ansible tab completion: re-read cached hosts & groups on directory change
- Date: Sun, 10 Apr 2022 21:45:07 +0000
- Archived-at: <https://zsh.org/workers/50033>
- In-reply-to: <cd169bfb-62ff-4a17-837a-28194677fea8@www.fastmail.com>
- List-id: <zsh-workers.zsh.org>
- References: <877dbbopzo.fsf@bunkus.online> <cd169bfb-62ff-4a17-837a-28194677fea8@www.fastmail.com>
Lawrence Velázquez wrote on Thu, Mar 31, 2022 at 01:55:49 -0400:
> Hi Moritz,
>
> On Fri, Jan 7, 2022, at 4:21 AM, Moritz Bunkus wrote:
> > we have several Ansible repositories in different directories. I often have
> > to switch between them.
> >
> > Unfortunately the tab completion for all the Ansible commands caches known
> > Hosts & Groups in the global variables _ansible_hosts &
> > _ansible_groups. For me the effect is that the wrong set of hosts & groups
> > is offered when I change to a different Ansible repo.
> >
> > How to reproduce:
> >
> > 1. Have two or more Ansible repositories with different inventories
> > somewhere
> > 2. cd into /path/to/first_ansible_tree
> > 3. Type `anssible <TAB>`, let tab completion cache hosts & groups
> > 4. Hit <TAB> again to verify that the offered completions match expected
> > ones
> > 5. cd into /path/to/other_ansible_tree
> > 6. Type `ansible <TAB><TAB>` and observe that the offered completions still
> > come from the first_ansible_tree, not from other_ansible_tree
> >
> > As a quick workaround I've modified the completion to also cache the
> > working directory where _ansible_hosts & _ansible_groups were created. If
> > the current working directory doesn't match the cached working directory,
> > both variables are unset & their values cached again. This seems to work.
> >
> > The attached patch implements this. Feel free to use it however you want,
> > or to apply a totally different solution.
>
> Thanks for the contribution, and apologies for the delay.
>
> workers: Are we interested in integrating this?
Yes. Tested, and it's correct, but we're in release freeze now ☹
An alternative would have been _store_cache with $PWD as part of the key
(and possibly the hostname too, in case $HOME is shared across multiple
hosts).
Cheers,
Daniel
Messages sorted by:
Reverse Date,
Date,
Thread,
Author