Elijah Newren <newren@xxxxxxxxx> 于2022年11月16日周三 13:49写道：
>
> On Mon, Nov 14, 2022 at 8:03 PM ZheNing Hu <adlternative@xxxxxxxxx> wrote:
> > Elijah Newren via GitGitGadget <gitgitgadget@xxxxxxxxx> 于2022年11月6日周日 14:04写道：
> [...]
> > > +sparsity patterns: patterns from $GIT_DIR/info/sparse-checkout used to
> > > + define the set of files of interest. A warning: It is easy to
> > > + over-use this term (or the shortened "patterns" term), for two
> > > + reasons: (1) users in cone mode specify directories rather than
> > > + patterns (their directories are transformed into patterns, but
> > > + users may think you are talking about non-cone mode if you use the
> > > + word "patterns"), and (b) the sparse specification might
> >
> > nit: s/(b)/(2)/g
>
> Thanks.
>
> > > +sparse specification: The set of paths in the user's area of focus. This
> > > + is typically just the tracked files that match the sparsity
> > > + patterns, but the sparse specification can temporarily differ and
> > > + include additional files. (See also "Sparse specification
> > > + vs. sparsity patterns")
> > > +
> > > + * When working with history, the sparse specification is exactly
> > > + the set of files matching the sparsity patterns.
> > > + * When interacting with the working tree, the sparse specification
> > > + is the set of tracked files with a clear SKIP_WORKTREE bit or
> > > + tracked files present in the working copy.
> >
> > I'm guessing what you mean here is:
> > Some files are stored with a flag bit of !SKIP_WORKTREE in its index entry.
> > But files are "vivifying" (restore to worktree) or new files added to
> > index (tracked files),
> > they also belong to the sparse specification.
>
> For this case, when interacting with the working tree, I mean the
> union of the following two sets of files:
> * files with !SKIP_WORKTREE in the index entry
> * files present in the working copy whose index entry has the
> SKIP_WORKTREE bit set.
>
Precise description. This is consistent with the behavior in
clear_skip_worktree_from_present_files().
> The fact that files might be new index entries (i.e. newly tracked
> files) is irrelevant; it only matters whether such files fit into one
> of the two categories above or not.
>
Okay.
> The fact that files have been "vivified" is slightly ambiguous and
> thus a bad way to define this set. When git vivifies files, it'll
> clear the SKIP_WORKTREE bit. If an editor, or script external to git,
> or something else restores the file, it will likely overlook that
> detail. We want vivified files to be part of the sparse specification
> when interacting with the working tree regardless of whether the
> SKIP_WORKTREE bit was correctly updated, so I defined it the way I did
> to remove such ambiguity. (I guess I should note that as per
> af6a51875a ("repo_read_index: clear SKIP_WORKTREE bit from files
> present in worktree", 2022-01-14), git will clear the SKIP_WORKTREE
> bit for files present in the working copy as one of the first things
> it does, but that could leave people wondering whether I meant the
> SKIP_WORKTREE bit was set as of the time of git invocation. So, I
> explicitly call out files present in the working copy for which the
> index entry has the SKIP_WORKTREE bit set, so folks know these files
> are definitely included in the sparse specification.)
>
You are right: the definition of vivifying explicitly clears the
SKIP_WORKTREE bit from the index, So the behavior described
here is not vivifying, but very much like vivifying: clear the
SKIP_WORKTREE bit from index_entry in memory instead of actually
clearing it from the index on disk.
Anyway, for the file on "worktree", we can use ce_skip_worktree(ce)
to check if it belongs to the sparse specification.
> > I think we can add some examples to describe these terms.
> >
> > #!/bin/sh
> >
> > set -x
> >
> > rm -rf mono-repo
> > git init mono-repo -b main
> > (
> > cd mono-repo &&
> > mkdir p1 p2 &&
> > echo a >p1/a &&
> > echo b >p1/b &&
> > echo a >p2/a &&
> > echo b >p2/b &&
> > git add . &&
> > git commit -m ok &&
> > git sparse-checkout set p1 &&
> > git ls-files -t &&
> > echo a >>p1/a &&
> > echo b >>p1/b &&
> > mkdir p2 p3 &&
> > echo next >>p2/a &&
> > echo next >>p3/c &&
> > git add p3/c &&
Here I forget a "--sparse"
> > # p2/a and p3/c vivify
> > git ls-files -t &&
> > # compare wortree/commit
> > git --no-pager diff HEAD --name-only
> > )
>
> You've added a bunch of code with this example, but you have not said
> what the output should be, so how exactly does this help describe the
> terms?
>
We create a repo with two sub projects p1/ p2/, then set
sparsity directory p1.
First git ls-files -t outputs:
H p1/a
H p1/b
S p2/a
S p2/b
It shows that index entries in p2 are skip-worktree.
Then we restore p2/a in the working tree and create a
new file p3/c and add it to the index.
The second git ls-files -t output:
H p1/a
H p1/b
H p2/a
S p2/b
H p3/c
p2/a and p3/c are not in sparse patterns, but they are in
sparse specification. It's like a special "vivifying".
> > > + * When modifying or showing results from the index, the sparse
> > > + specification is the set of files with a clear SKIP_WORKTREE bit
> > > + or that differ in the index from HEAD.
> >
> > #!/bin/sh
> >
> > set -x
> >
> > rm -rf mono-repo
> > git init mono-repo -b main
> > (
> > cd mono-repo &&
> > mkdir p1 p2 &&
> > echo a >p1/a &&
> > echo b >p1/b &&
> > echo a >p2/a &&
> > echo b >p2/b &&
> > git add . &&
> > git commit -m ok &&
> > git sparse-checkout set p1 &&
> > git update-index --chmod=+x p2/a &&
> > # compare commit/index
> > git --no-pager diff --cached --name-only
> > )
>
> Same issue here; you haven't stated the expected output of these
> commands, so I don't see how they help with the description at all.
>
Here only output p2/a:
p2/a is out of sparse patterns, but this index entry mode has been
changed compared to HEAD. So we should consider it as a part of
sparse specification.
> Perhaps it's worth noting why I think the sparse specification should
> be extended when dealing with the index:
>
> * "mergy" commands (merge, rebase, cherry-pick, am, revert) can
> modify the index outside the sparsity patterns, without creating a
> commit.
> * `git commit` (or `rebase --continue`, or whatever) will create a
> commit from whatever staged versions of files there are
> => `git status` should show what is about to be committed
> => `git diff --cached --name-only` ought to be usable to show what
> is to be committed
> => `git grep --cached ...` ought to be usable to search through what
> is about to be committed
>
> See also https://lore.kernel.org/git/CABPp-BESkb=04vVnqTvZyeCa+7cymX7rosUW3rhtA02khMJKHA@xxxxxxxxxxxxxx/
> (starting with the paragraph with "leery" in it), and the thread
> starting there. If the sparse specification is not expanded, users
> will get some nasty surprises, and the only other alternative I can
> think of to avoid such surprises would be making several commands
> always run full tree. Running full-tree with a non-default option to
> run sparse forces behavior A folks into a "pick your poison"
> situation, which is not nice. Extending the sparse specification to
> include files whose index entries do not match HEAD for index-related
> operations provides the nice middle ground that avoids such usability
> problems while also allowing users to avoid operating on a full tree.
>
I can understand the reason why we need to extend sparse specification:
index often needs to handle files that are not in the sparse pattern.
> [...]
> > I think A's users might need a little more refinement.
> >
> > A: Some users are _only_ interested in the sparse portion of the repo,
> > but they don't want to download all the blobs, they can accept to download
> > other data later using partial-clone, which can reduce the local storage size.
> > (Current default behavior)
>
> Behavior A is definitely not the current default behavior. Also,
> behavior A is not tied to partial clones; some users may well want it
> even with a dense clone, so we need to avoid suggesting it is only for
> users with partial clones. (Though, if users are using partial clones
> with behavior A, then I agree with the part you wrote other than your
> parenthetical comment.)
>
Makes sense. This should be considered scalar-style behavior.
> > A** : Some users are _only_ interested in the sparse portion of the repo,
> > but they want to download all the blobs in it to avoid some unnecessary
> > network connections afterwards.
>
> Here you just repeated `A*` but relabelled it as `A**`. Yes, this one
> is explicitly tied to partial clone behavior.
Ah, `A*` part say “so things like `git log -S${SEARCH_TERM} -p`
or `git grep ${SEARCH_TERM} OLDREV ` would need to be prepared to provide
partial results that depend on what happens to have been downloaded."
So I think it's probably a lot like the behavior after a shallow
clone: git log -p or other
git commands returning partial results.
The expectation of A** is to have all blobs under the entire sparse-patterns.
>
> [...]
> > > + The fact that files can move between the 'tracked' and 'untracked'
> > > + categories means some commands will have to treat untracked files
> > > + differently. But if we have to treat untracked files differently,
> > > + then additional commands may also need changes:
> > > +
> > > + * status
> > > + * clean
> > > +
> >
> > I'm a bit worried about git status, because it's used in many shells
> > (e.g. zsh) i
> > in the git prompt function. Its default behavior is restricted, otherwise users
> > may get blocked when they use zsh to cd to that directory. I don't know how
> > to reproduce this problem (since the scenario is built on checkout to a local
> > unborn branch).
>
> Could you elaborate? I'm not sure if you are talking about an
> existing problem that you are worried about being exacerbated, or a
> hypothetical problem that could occur with changes. Further, your
> wording is so vague about the problem, that I have no idea what its
> nature is or whether any changes to status would even possibly have
> any bearing on it. But the suggested changes to git status are
> simply:
>
I just might have caused this in one particular case. So it's not very
important at the moment. But it's worth noting that many shells, IDEs’
git plugins
may also need to understand sparse-checkout properly, otherwise it can
cause some usability problems.
> > > + In particular, `status` may need to report any untracked files outside
> > > + the sparsity specification as an erroneous condition (especially to
> > > + avoid the user trying to `git add` them, forcing `git add` to display
> > > + an error).
>
> [...]
> > > + ls-files may be slightly special in that e.g. `git ls-files -t` is
> > > + often used to see what is sparse and what is not. Perhaps -t should
> > > + always work on the full tree?
> > > +
> >
> > Recently git ls-files added a --format option, perhaps this can be modified to
> > show if a file is SKIP_WORKTREE in the future.
>
> If so, then you've made my question also applicable for `--format`;
> much like -t, should --format always work on the full tree? (Or maybe
> just when the format specifies the skip worktree bit?)
>
My personal opinion is to default to "restrict" , but we can use "git
ls-files -t -scope=all"
to check all index entries
.
> [...]
> > > + * If a config option is added (sparse.scope?) what should the values and
> > > + description be? "sparse" (behavior A), "worktree-sparse-history-dense"
> > > + (behavior B), "dense" (behavior C)? There's a risk of confusion,
> > > + because even for Behaviors A and B we want some commands to be
> > > + full-tree and others to operate sparsely, so the wording may need to be
> > > + more tied to the usecases and somehow explain that. Also, right now,
> > > + the primary difference we are focusing is just the history-querying
> > > + commands (log/diff/grep). Previous config suggestion here: [13]
> > > +
> >
> > Maybe sparse.scope={sparse, all}?
>
> I guess that's people's common first guess. However, when you dig in,
> I think this would be badly broken -- see my response to Stolee I just
> sent out.
Ok...
