Re: [PATCH 5/6] completion: list existing working trees for 'git worktree' subcommands
On Fri, Oct 18, 2019 at 11:00 AM SZEDER Gábor <szeder.dev@gmail.com> wrote:
Show 18 quoted lines
> On Thu, Oct 17, 2019 at 02:08:12PM -0400, Eric Sunshine wrote:
> > It is a long-standing To-Do[1] for "git worktree list [--porcelain]"
> > to indicate whether a worktree is locked, prunable, etc. Looking at
> > the implementation of builtin/worktree.c:show_worktree_porcelain(), it
> > should be easy enough to add.
>
> I didn't look at the implementation, but only at the docs, which says:
>
> --porcelain
> With list, output in an easy-to-parse format for scripts. This
> format will remain stable across Git versions and regardless of
> user configuration. See below for details.
>
> I'm not sure whether introducing a new boolean attribute (i.e. a line
> containing only "locked") would still be considered acceptable, or
> would count as changing the format. I can imagine that a too strict
> parser would barf upon encountering the unrecognized "locked"
> attribute; but yeah, no sensible parser should be that strict IMO.
The stanza-based format of the porcelain output was chosen with the specific intention of being easy to parse _and_ to allow extension such as the introduction of new attributes. It is unfortunate that the documentation you quoted, as well as the description of the porcelain format itself, does a poor job of (or utterly fails at) conveying that intention. (Had I been around to review the later iteration(s) of the series which introduced the porcelain format, I likely would have pointed out these shortcomings, however, Real Life had other ideas, and I didn't manage to review it until weeks after the series had made it into an actual release.)
I'm not convinced, though, that we're locked into exactly the few attributes shown only in an example in the porcelain section of the documentation. That documentation is so woefully underspecified -- indeed, it doesn't even talk about what attributes are available, but only gives a single example showing some of the (possible) attributes -- that it effectively makes no promises about what will or will not appear in a stanza. (The only thing is says strongly is that stanzas will be separated by a blank line -- despite the intention all along being that each new stanza would start with a "worktree" attributes, and that blank lines, if output, were to be ignored.)
So, I am of the opinion that we are not prevented from adding new attributes, such as "locked" and "prunable".
Show 6 quoted lines
> Furthermore, I'm not sure what to do with the reason for locking. In
> general I would think that it makes sense to display the reason in an
> easy-to-parse format as well. However, doing so will inherently make
> the format less easy to parse, because the reason could span multiple
> lines, so without some sort of encoding/escaping it would violate the
> "a line per attribute" format.
Yep, I believe my thinking at the time was that the lock reason and the prunable reason would be escaped if needed. So, for instance:
worktree /blah
branch refs/heads/blah
locked Sneaker-net removable storage\nNot always mounted> I would say that this is beyong the scope of this patch series :)
Oh, I wasn't suggesting that this series do anything of the sort. Instead, I was merely responding to the point in the commit message that the porcelain format was lacking information about the lock, and to say that it could (eventually) be solved by extending the output.
SZEDER Gábor