patchdoc: emphasize stash "--keep-index" stashes staged content

It's not immediately obvious from the man page that the "--keep-index" option still adds the staged content to the stash, so make that abundantly clear.

Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>

---

"Robert P. J. Day" <rpjday@crashcourse.ca> writes:

I do not think "left intact in the index" is an improvement. The primary reason why people use --keep-index is to get a working tree that has _only_ the changes that are already added to the index and nothing else, so that the state in the index can be built and/or tested. So if you want to add anything to that sentence, it should stress the fact that changes are left intact "in the working tree". Adding "in the index" without doing so makes the result even more confusing, not less, by allowing a mis-read "left intact _only_ in the index? so the working tree files are reverted to the HEAD's version?"

I am still undecided as to the value of saying ", while still being added...".

Hi,

Robert P. J. Day wrote:

Aside from Junio's note about "in the index" vs "in the working tree":

The "Testing partial commits" item in the EXAMPLES section explains what --keep-index is useful for. I wonder if some allusion to that would make the explanation in the OPTIONS section easier to understand.

Something that I end up still curious about when reading this description is what will happen when I "git stash pop". Will it apply only the changes that were stashed away and removed from the working tree, or will it apply the changes that were kept in the index, too? If the latter, why? Is there some way I can turn that behavior off?

E.g. in the "Testing partial commits" example, it seems like the natural behavior for "git stash pop" would be just restore the changes that were removed from the working tree. That would also match an assumption of save/push and pop being symmetrical ('inverse operations').

Is this related to "git stash pop --index"? I notice that the EXAMPLES section doesn't give any examples of that option.

Thanks, Jonathan

On Wed, 11 Oct 2017, Jonathan Nieder wrote:

  yes, that was a good point, i will ponder further.

  at risk of embarrassing myself, it seems that the simplest way to
explain stashing WRT to those --keep-index and --index options is to
first explain that, regardless of --keep-index with push, stash will
*always* stash all of your changes in the working tree, and will
further distinguish between staged and unstaged content. that's based
on the diagram in the man page:

                  .----W
                 /    /
           -----H----I

so the initial explanation should be that the above *always* happens, no matter what.

  the next sentence should then say, "if you add --keep-index, then
staged changes are preserved in the working tree and index", or
something like that. but the use of --keep-index does not (unless i'm
reading this incorrectly) in any way affect what is stashed, correct?

  in that same vein, the explanation should then go on to explain that
popping always restores the *entire* stash to the working tree -- all
of it -- and the use of "--index" simply means that the portion of
the stash representing what had been staged will be restaged.

  not to belabour the point, but i think it's important to emphasize
early that --keep-index and --index in no way affect what is stashed,
and what is popped, which i think is a common misunderstanding.

rday