Re: [PATCH 1/2] [RFC] wt-status: show sparse checkout status as well

From: Derrick Stolee <stolee@gmail.com>
Date: Jun 17, 2020, 14:50 UTC
Message-ID: <e8e59662-7e8b-6aae-da28-c4b5ee225ac8@gmail.com>
In-Reply-To: <462cee857ef30c413eda0abe655b123f32b3877e.1592350413.git.gitgitgadget@gmail.com>

On 6/16/2020 7:33 PM, Elijah Newren via GitGitGadget wrote:

Show 9 quoted lines

> From: Elijah Newren <newren@gmail.com>
> 
> Some of the early feedback of folks trying out sparse-checkouts at
> $dayjob is that sparse checkouts can sometimes be disorienting; users
> can forget that they had a sparse-checkout and then wonder where files
> went.  Add some output to 'git status' in the form of a simple line that
> states:
> 
>     You are in a sparse checkout with 35% of files present.

I like this idea.

Aside: I think there was some ideas around a "git sparse-checkout stats"
subcommand that would help users identify how large their sparse-checkout
definition is compared to the full index. That would also want to drill
down into directories, so it would need to be more advanced than what you
have here.

There is some prior work in the "gvfs health" commmand in VFS for Git. We use this to help users self-diagnose when they have "over hydrated" their working directory. Here is a sample use:

  /c/_git/t/ForTests/src (master)
  $ gvfs health
  
  Gathering repository data...
  
  Health of directory:
  Total files in HEAD commit:           548 | 100%
  Files managed by VFS for Git (fast):   30 |   5%
  Files managed by Git:                   1 |   0%
  
  Total hydration percentage:                   6%
  
  Most hydrated top level directories:
   19 / 524 | GVFS
    0 / 8   | GitHooksLoader
    0 / 3   | Scripts
  
  Repository status: OK
  
  /c/_git/t/ForTests/src (master)
  $ cd GVFS
  
  /c/_git/t/ForTests/src/GVFS (master)
  $ gvfs health
  
  Gathering repository data...
  
  Health of directory: GVFS/
  Total files in HEAD commit:           524 | 100%
  Files managed by VFS for Git (fast):   19 |   4%
  Files managed by Git:                   0 |   0%
  
  Total hydration percentage:                   4%
  
  Most hydrated top level directories:
   0 / 33 | FastFetch
   0 / 92 | GVFS.Common
   0 / 97 | GVFS.FunctionalTests
   0 / 22 | GVFS.GVFlt
   0 / 7  | GVFS.Hooks
  
  Repository status: OK

This is just a reference for something that has been helpful for some users. In VFS for Git, the hydration is not explicit but instead implicit by filesystem use. There could still be some value in the sparse-checkout case.

Show 24 quoted lines

> where, obviously, the exact figure changes depending on what percentage
> of files from the index do not have the SKIP_WORKTREE bit set.
> 
> Signed-off-by: Elijah Newren <newren@gmail.com>
> ---
>  wt-status.c | 35 +++++++++++++++++++++++++++++++++++
>  wt-status.h |  1 +
>  2 files changed, 36 insertions(+)
> 
> diff --git a/wt-status.c b/wt-status.c
> index 98dfa6f73f9..687d2ab1ba1 100644
> --- a/wt-status.c
> +++ b/wt-status.c
> @@ -1484,6 +1484,16 @@ static void show_bisect_in_progress(struct wt_status *s,
>  	wt_longstatus_print_trailer(s);
>  }
>  
> +static void show_sparse_checkout_in_use(struct wt_status *s,
> +					const char *color)
> +{
> +	if (s->state.sparse_checkout_percentage != -1)
> +		status_printf_ln(s, color,
> +				 _("You are in a sparse checkout with %d%% of tracked files present."),
> +				 s->state.sparse_checkout_percentage);

There is some concern over breaking third-party tools that parse "git status". However, we have had the machine-readable formats out for a long time.

I briefly thought about putting this message to stderr to be safe, but some tools scan stderr (incorrectly) to say "any output to stderr must mean an error in the process."

Show 17 quoted lines

> +	wt_longstatus_print_trailer(s);
> +}
> +
>  /*
>   * Extract branch information from rebase/bisect
>   */
> @@ -1623,6 +1633,27 @@ int wt_status_check_bisect(const struct worktree *wt,
>  	return 0;
>  }
>  
> +static void wt_status_check_sparse_checkout(struct repository *r,
> +					    struct wt_status_state *state)
> +{
> +	int skip_worktree = 0;
> +	int i;
> +
> +	if (!core_apply_sparse_checkout) {

Since this method is static, we can assume that it is only called after the index was loaded, which requires reading the config and initializing this value. Good.

Show 44 quoted lines

> +		state->sparse_checkout_percentage = -1;
> +		return;
> +	}
> +
> +	for (i = 0; i < r->index->cache_nr; i++) {
> +		struct cache_entry *ce = r->index->cache[i];
> +		if (ce_skip_worktree(ce))
> +			skip_worktree++;
> +	}
> +
> +	state->sparse_checkout_percentage =
> +		100 - (100 * skip_worktree)/r->index->cache_nr;
> +}
> +
>  void wt_status_get_state(struct repository *r,
>  			 struct wt_status_state *state,
>  			 int get_detached_from)
> @@ -1658,6 +1689,7 @@ void wt_status_get_state(struct repository *r,
>  	}
>  	if (get_detached_from)
>  		wt_status_get_detached_from(r, state);
> +	wt_status_check_sparse_checkout(r, state);
>  }
>  
>  static void wt_longstatus_print_state(struct wt_status *s)
> @@ -1681,6 +1713,9 @@ static void wt_longstatus_print_state(struct wt_status *s)
>  		show_revert_in_progress(s, state_color);
>  	if (state->bisect_in_progress)
>  		show_bisect_in_progress(s, state_color);
> +
> +	if (state->sparse_checkout_percentage != -1)
> +		show_sparse_checkout_in_use(s, state_color);
>  }
>  
>  static void wt_longstatus_print(struct wt_status *s)
> diff --git a/wt-status.h b/wt-status.h
> index 73ab5d4da1c..4550004003a 100644
> --- a/wt-status.h
> +++ b/wt-status.h
> @@ -90,6 +90,7 @@ struct wt_status_state {
>  	int bisect_in_progress;
>  	int revert_in_progress;
>  	int detached_at;
> +	int sparse_checkout_percentage; /* -1 == not in sparse checkout */

My only complaint is that maybe this "-1" should be a #define macro with a name such as SPARSE_CHECKOUT_DISABLED.

Thanks, -Stolee