| Message ID | xmqqzh4dk3ey.fsf@gitster.c.googlers.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | apply: clarify description of --index | expand |
Am 23.10.20 um 05:49 schrieb Junio C Hamano: > Instead of explaining the requirement for the paths to be up-to-date, > as if it is an afterthought, state it upfront. > > The updated description matches how the checks actually are > performed. A path that is "dirty" stops the patch application from > being attempted to either working tree files or to the index. > > Hopefully this change would help users to form a better mental > model. > > Signed-off-by: Junio C Hamano <gitster@pobox.com> > --- > > * Just noticed while reviewing how "apply" (and "am") are explained. > > Documentation/git-apply.txt | 12 +++++------- > 1 file changed, 5 insertions(+), 7 deletions(-) > > diff --git a/Documentation/git-apply.txt b/Documentation/git-apply.txt > index 91d9a8601c..1be7751f58 100644 > --- a/Documentation/git-apply.txt > +++ b/Documentation/git-apply.txt > @@ -61,13 +61,11 @@ OPTIONS > file and detects errors. Turns off "apply". > > --index:: > - Apply the patch to both the index and the working tree (or > - merely check that it would apply cleanly to both if `--check` is > - in effect). Note that `--index` expects index entries and > - working tree copies for relevant paths to be identical (their > - contents and metadata such as file mode must match), and will > - raise an error if they are not, even if the patch would apply > - cleanly to both the index and the working tree in isolation. > + After making sure the paths the patch touches in the working > + tree have no modifications relative to their index entries, > + apply the patch both to the index entries and to the working > + tree files or see if it applies cleanly, when `--check` is in > + effect. I don't think that this is an improvement. The purpose of --index *is* to apply the patch to both index and worktree, and that should be mentioned first. The check that both are identical, is a prerequisite and not the primary objective of the option. > > --cached:: > Apply the patch to just the index, without touching the working > -- Hannes
Johannes Sixt <j6t@kdbg.org> writes: >> - Apply the patch to both the index and the working tree (or >> - merely check that it would apply cleanly to both if `--check` is >> - in effect). Note that `--index` expects index entries and >> - working tree copies for relevant paths to be identical (their >> - contents and metadata such as file mode must match), and will >> - raise an error if they are not, even if the patch would apply >> - cleanly to both the index and the working tree in isolation. >> + After making sure the paths the patch touches in the working >> + tree have no modifications relative to their index entries, >> + apply the patch both to the index entries and to the working >> + tree files or see if it applies cleanly, when `--check` is in >> + effect. > > I don't think that this is an improvement. The purpose of --index *is* > to apply the patch to both index and worktree, and that should be > mentioned first. The check that both are identical, is a prerequisite > and not the primary objective of the option. Yeah, but this was an attempt to clarify what that "apply to both", which is the central part of the operation, exactly means. The only mode of operation we offer is that we start from identical index and working tree, and make the same change so that we arrive at the same outcome. It is not like you can have some changes in the working tree file as long as they do not overlap and collide with the incoming patch, make the same change with the patch to arrive at different contents as the outcome. We explicitly forbid that, but "apply to both" does not exactly tell it to the readers. But I am OK to drop this, if you do not think it clarifies the description. Thanks.
Am 23.10.20 um 16:38 schrieb Junio C Hamano: > Johannes Sixt <j6t@kdbg.org> writes: > >>> - Apply the patch to both the index and the working tree (or >>> - merely check that it would apply cleanly to both if `--check` is >>> - in effect). Note that `--index` expects index entries and >>> - working tree copies for relevant paths to be identical (their >>> - contents and metadata such as file mode must match), and will >>> - raise an error if they are not, even if the patch would apply >>> - cleanly to both the index and the working tree in isolation. >>> + After making sure the paths the patch touches in the working >>> + tree have no modifications relative to their index entries, >>> + apply the patch both to the index entries and to the working >>> + tree files or see if it applies cleanly, when `--check` is in >>> + effect. >> >> I don't think that this is an improvement. The purpose of --index *is* >> to apply the patch to both index and worktree, and that should be >> mentioned first. The check that both are identical, is a prerequisite >> and not the primary objective of the option. > > Yeah, but this was an attempt to clarify what that "apply to both", > which is the central part of the operation, exactly means. > > The only mode of operation we offer is that we start from identical > index and working tree, and make the same change so that we arrive > at the same outcome. It is not like you can have some changes in > the working tree file as long as they do not overlap and collide > with the incoming patch, make the same change with the patch to > arrive at different contents as the outcome. We explicitly forbid > that, but "apply to both" does not exactly tell it to the readers. Your have point that the original text muddies the preconditions a bit, but I still think that "what it does" must be the first thing to be mentioned, and the preconditions the second. -- Hannes
Johannes Sixt <j6t@kdbg.org> writes: > Am 23.10.20 um 16:38 schrieb Junio C Hamano: >> Johannes Sixt <j6t@kdbg.org> writes: >> >>>> - Apply the patch to both the index and the working tree (or >>>> - merely check that it would apply cleanly to both if `--check` is >>>> - in effect). Note that `--index` expects index entries and >>>> - working tree copies for relevant paths to be identical (their >>>> - contents and metadata such as file mode must match), and will >>>> - raise an error if they are not, even if the patch would apply >>>> - cleanly to both the index and the working tree in isolation. >>>> + After making sure the paths the patch touches in the working >>>> + tree have no modifications relative to their index entries, >>>> + apply the patch both to the index entries and to the working >>>> + tree files or see if it applies cleanly, when `--check` is in >>>> + effect. >>> >>> I don't think that this is an improvement. The purpose of --index *is* >>> to apply the patch to both index and worktree, and that should be >>> mentioned first. The check that both are identical, is a prerequisite >>> and not the primary objective of the option. >> >> Yeah, but this was an attempt to clarify what that "apply to both", >> which is the central part of the operation, exactly means. >> >> The only mode of operation we offer is that we start from identical >> index and working tree, and make the same change so that we arrive >> at the same outcome. It is not like you can have some changes in >> the working tree file as long as they do not overlap and collide >> with the incoming patch, make the same change with the patch to >> arrive at different contents as the outcome. We explicitly forbid >> that, but "apply to both" does not exactly tell it to the readers. > > Your have point that the original text muddies the preconditions a bit, > but I still think that "what it does" must be the first thing to be > mentioned, and the preconditions the second. Yeah. I offhand do not think of a better phrasing within the constraint that "apply to only identical state" must be said after saying "to both the index and the working tree", though, so I'll leave it up to the list's wisdom ;-) Thanks.
diff --git a/Documentation/git-apply.txt b/Documentation/git-apply.txt index 91d9a8601c..1be7751f58 100644 --- a/Documentation/git-apply.txt +++ b/Documentation/git-apply.txt @@ -61,13 +61,11 @@ OPTIONS file and detects errors. Turns off "apply". --index:: - Apply the patch to both the index and the working tree (or - merely check that it would apply cleanly to both if `--check` is - in effect). Note that `--index` expects index entries and - working tree copies for relevant paths to be identical (their - contents and metadata such as file mode must match), and will - raise an error if they are not, even if the patch would apply - cleanly to both the index and the working tree in isolation. + After making sure the paths the patch touches in the working + tree have no modifications relative to their index entries, + apply the patch both to the index entries and to the working + tree files or see if it applies cleanly, when `--check` is in + effect. --cached:: Apply the patch to just the index, without touching the working
Instead of explaining the requirement for the paths to be up-to-date, as if it is an afterthought, state it upfront. The updated description matches how the checks actually are performed. A path that is "dirty" stops the patch application from being attempted to either working tree files or to the index. Hopefully this change would help users to form a better mental model. Signed-off-by: Junio C Hamano <gitster@pobox.com> --- * Just noticed while reviewing how "apply" (and "am") are explained. Documentation/git-apply.txt | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)