git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH v1 01/11] checkout: split part of it to new command 'restore'

From
Elijah Newren <newren@gmail.com>
Date
Mar 8, 2019, 18:01 UTC
Message-ID
<CABPp-BFXZMorrHph3hGFnqfceHs68byWNgffNKGp1ov6X5-o5A@mail.gmail.com>
In-Reply-To
<20190308101655.9767-2-pclouds@gmail.com>

Thanks for working on this; overall looks really good. I've got a few comments here and there on the wording and combinations of options...

On Fri, Mar 8, 2019 at 2:17 AM Nguyễn Thái Ngọc Duy <pclouds@gmail.com> wrote:
> +SYNOPSIS

It might be worth adding some words somewhere to differentiate between `reset` and `restore`. E.g.

`git restore` modifies the working tree (and maybe index) to change file content to match some other (usually older) version, but does not update your branch. `git reset` is about modifying which commit your branch points to, meaning possibly removing and/or adding many commits to your branch.

It may also make sense to add whatever description you use to the reset manpage.
> +--------
> +[verse]
> +'git restore' [<options>] [--source=<revision>] <pathspec>...
> +'git restore' (-p|--patch) [--source=<revision>] [<pathspec>...]

So one cannot specify any special options with -p? Does that mean one cannot use it with --index (i.e. this command cannot replace 'git reset -p')? Or is this an oversight in the synopsis?

Show 7 quoted lines
> +DESCRIPTION
> +-----------
> +Restore paths in the working tree by replacing with the contents in
> +the restore source or remove if the paths do not exist in the restore
> +source. The source is by default the index but could be from a commit.
> +The command can also optionally restore content in the index from a
> +commit.

The first sentence makes it sound like two different operations, when I think it is one. Perhaps:

Restore paths in the working tree by replacing with the contents in the restore source (and if the restore source is missing paths found in the working tree, that means deleting those paths from the working tree).

> +
> +When a `<revision>` is given, the paths that match the `<pathspec>` are
> +updated both in the index and in the working tree.

I thought the default was --worktree. Is this sentence from an older version of your patch series that you forgot to update?

Show 8 quoted lines
> +
> +OPTIONS
> +-------
> +-s<tree>::
> +--source=<tree>::
> +       Restore the working tree files with the content from the given
> +       tree or any revision that leads to a tree (e.g. a commit or a
> +       branch).

I think that's a little hard to parse. We may not be able to avoid "working tree" vs. "tree" confusion, but the spelling of <tree> feels like it should be a separate sentence. Maybe:

Restore the working tree files with the content from the given tree. It is common to specify the source tree by naming a commit, branch, or tag.

?
Show 17 quoted lines
> +
> +-p::
> +--patch::
> +       Interactively select hunks in the difference between the
> +       `<revision>` (or the index, if unspecified) and the working
> +       tree. See the ``Interactive Mode'' section of linkgit:git-add[1]
> +       to learn how to operate the `--patch` mode.
> +
> +-W::
> +--worktree::
> +-I::
> +--index::
> +       Specify the restore location. If neither option is specified,
> +       by default the working tree is restored. If `--index` is
> +       specified without `--worktree` or `--source`, `--source=HEAD`
> +       is implied. These options only make sense to use with
> +       `--source`.

Seems like this contains a minor contradiction. Perhaps start the final sentence with: "Otherwise, ..." ?

Show 10 quoted lines
> +-q::
> +--quiet::
> +       Quiet, suppress feedback messages.
> +
> +--progress::
> +--no-progress::
> +       Progress status is reported on the standard error stream
> +       by default when it is attached to a terminal, unless `--quiet`
> +       is specified. This flag enables progress reporting even if not
> +       attached to a terminal, regardless of `--quiet`.

I'm assuming this means there are feedback messages other than progress feedback?

Show 5 quoted lines
> +-f::
> +--force::
> +       If `--source` is not specified, unmerged entries are left alone
> +       and will not fail the operation. Unmerged entries are always
> +       replaced if `--source` is specified, regardless of `--force`.

This may be slightly confusing, in particular it suggests that --index (or --worktree and --index) are the default. Is --force only useful when --index is specified? If it has utility with --worktree only, what does it do? Also, what happens when there are unmerged entries in the index and someone tries to restore just working tree files -- are the ones corresponding to unmerged entries skipped (if so, silently or with warnings printed for the user?), or does something else happen?

Show 8 quoted lines
> +--ours::
> +--theirs::
> +       Check out stage #2 ('ours') or #3 ('theirs') for unmerged
> +       paths.
> ++
> +Note that during `git rebase` and `git pull --rebase`, 'ours' and
> +'theirs' may appear swapped. See the explanation of the same options
> +in linkgit:git-checkout[1] for details.

So sad, but yes you need to mention it. I'm curious what we say for cherry-pick; it's not clear to me whether people think of the current branch as "ours" or the commit they wrote themselves and are trying to bring to this branch as "ours". There are probably examples of both.

Not that I think you can do anything about that here or need to change your description. I'm just very sad about --ours and --theirs in general.

Show 11 quoted lines
> +
> +-m::
> +--merge::
> +       Recreate the conflicted merge in the specified paths.
> +
> +--conflict=<style>::
> +       The same as `--merge` option above, but changes the way the
> +       conflicting hunks are presented, overriding the merge.conflictStyle
> +       configuration variable.  Possible values are "merge" (default)
> +       and "diff3" (in addition to what is shown by "merge" style,
> +       shows the original contents).

Should you mention that these are incompatible with --source and --index? And perhaps also make sure the code aborts if either of these options are combined with either of those?

Show 5 quoted lines
> +--ignore-skip-worktree-bits::
> +       In sparse checkout mode, by default update only entries
> +       matched by `<pathspec>` and sparse patterns in
> +       $GIT_DIR/info/sparse-checkout. This option ignores the sparse
> +       patterns and unconditionally restores any files in `<pathspec>`.

The first sentence is slightly confusing; it sounds like you are saying what the flag does rather than what the default is without the flag. Perhaps:

"In sparse checkout mode, the default is to only update entries matched by `<pathspec>` and sparse patterns in $GIT_DIR/info/sparse-checkout...."

Show 7 quoted lines
> +
> +--recurse-submodules::
> +--no-recurse-submodules::
> +       Using `--recurse-submodules` will update the content of all initialized
> +       submodules according to the commit recorded in the superproject. If
> +       local modifications in a submodule would be overwritten the checkout
> +       will fail unless `-f` is used.
This suggests that your documentation for -f/--force is incomplete.
> +       If nothing (or `--no-recurse-submodules`)
> +       is used, the work trees of submodules will not be updated.

This seems slightly awkward. Perhaps "If `--no-recurse-submodules` (which is the default) is used, the work trees of submodules will not be updated.

Show 6 quoted lines
> +       Just like linkgit:git-submodule[1], this will detach the
> +       submodules HEAD.
> +
> +--overlay::
> +--no-overlay::
> +       In overlay mode, `git checkout` never removes files from the
Why are you talking about `git checkout` here?  Shouldn't this be `git restore`?
> +       index or the working tree. In no-overlay mode, files that
> +       appear in the index and working tree, but not in `--source` tree
> +       are removed, to make them match `<tree-ish>` exactly. The
> +       default is no-overlay mode.

A minor comment: This seems slightly weird because it feels like you start out with a negative ("never removes") and then describe the opposite option which adds another negation to the negative. But the wording avoids an explicit double negative (which is good), and I don't have a better suggestion for wording here. Just thought I'd flag it in case anyone else reading over this notices it too and has a wording suggestion.

Show 155 quoted lines
> +
> +EXAMPLES
> +--------
> +
> +The following sequence checks out the `master` branch, reverts
> +the `Makefile` to two revisions back, deletes hello.c by
> +mistake, and gets it back from the index.
> +
> +------------
> +$ git switch master
> +$ git restore --source master~2 Makefile  <1>
> +$ rm -f hello.c
> +$ git restore hello.c                   <2>
> +------------
> +
> +<1> take a file out of another commit
> +<2> restore hello.c from the index
> +
> +If you want to check out _all_ C source files out of the index,
> +you can say
> +
> +------------
> +$ git restore '*.c'
> +------------
> +
> +Note the quotes around `*.c`.  The file `hello.c` will also be
> +checked out, even though it is no longer in the working tree,
> +because the file globbing is used to match entries in the index
> +(not in the working tree by the shell).
> +
> +To restore all files in the current directory
> +
> +------------
> +$ git restore .
> +------------
> +
> +or to restore all working tree files with 'top' pathspec magic (see
> +linkgit::gitglossary[7])
> +
> +------------
> +$ git restore :/
> +------------
> +
> +To restore a file in the index only (this is the same as using
> +"git reset")
> +
> +------------
> +$ git restore --index hello.c
> +------------
> +
> +or you can restore both the index and the working tree
> +
> +------------
> +$ git restore --source=HEAD --index --worktree hello.c
> +------------
> +
> +SEE ALSO
> +--------
> +linkgit:git-checkout[1]
> +
> +GIT
> +---
> +Part of the linkgit:git[1] suite
> diff --git a/Makefile b/Makefile
> index 8e91db73ad..ffe7e4f58f 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -799,6 +799,7 @@ BUILT_INS += git-format-patch$X
>  BUILT_INS += git-fsck-objects$X
>  BUILT_INS += git-init$X
>  BUILT_INS += git-merge-subtree$X
> +BUILT_INS += git-restore$X
>  BUILT_INS += git-show$X
>  BUILT_INS += git-stage$X
>  BUILT_INS += git-status$X
> diff --git a/builtin.h b/builtin.h
> index c64e44450e..6830000e14 100644
> --- a/builtin.h
> +++ b/builtin.h
> @@ -214,6 +214,7 @@ extern int cmd_remote_fd(int argc, const char **argv, const char *prefix);
>  extern int cmd_repack(int argc, const char **argv, const char *prefix);
>  extern int cmd_rerere(int argc, const char **argv, const char *prefix);
>  extern int cmd_reset(int argc, const char **argv, const char *prefix);
> +extern int cmd_restore(int argc, const char **argv, const char *prefix);
>  extern int cmd_rev_list(int argc, const char **argv, const char *prefix);
>  extern int cmd_rev_parse(int argc, const char **argv, const char *prefix);
>  extern int cmd_revert(int argc, const char **argv, const char *prefix);
> diff --git a/builtin/checkout.c b/builtin/checkout.c
> index 4903359b49..11dd2ae44c 100644
> --- a/builtin/checkout.c
> +++ b/builtin/checkout.c
> @@ -37,6 +37,11 @@ static const char * const switch_branch_usage[] = {
>         NULL,
>  };
>
> +static const char * const restore_files_usage[] = {
> +       N_("git restore [<options>] [<branch>] -- <file>..."),
> +       NULL,
> +};
> +
>  struct checkout_opts {
>         int patch_mode;
>         int quiet;
> @@ -1528,3 +1533,24 @@ int cmd_switch(int argc, const char **argv, const char *prefix)
>         FREE_AND_NULL(options);
>         return ret;
>  }
> +
> +int cmd_restore(int argc, const char **argv, const char *prefix)
> +{
> +       struct checkout_opts opts;
> +       struct option *options = NULL;
> +       int ret;
> +
> +       memset(&opts, 0, sizeof(opts));
> +       opts.dwim_new_local_branch = 1;
> +       opts.switch_branch_doing_nothing_is_ok = 0;
> +       opts.accept_pathspec = 1;
> +
> +       options = parse_options_dup(options);
> +       options = add_common_options(&opts, options);
> +       options = add_checkout_path_options(&opts, options);
> +
> +       ret = checkout_main(argc, argv, prefix, &opts,
> +                           options, restore_files_usage);
> +       FREE_AND_NULL(options);
> +       return ret;
> +}
> diff --git a/command-list.txt b/command-list.txt
> index 13317f47d4..b9eae1c258 100644
> --- a/command-list.txt
> +++ b/command-list.txt
> @@ -151,6 +151,7 @@ git-replace                             ancillarymanipulators           complete
>  git-request-pull                        foreignscminterface             complete
>  git-rerere                              ancillaryinterrogators
>  git-reset                               mainporcelain           worktree
> +git-restore                             mainporcelain           worktree
>  git-revert                              mainporcelain
>  git-rev-list                            plumbinginterrogators
>  git-rev-parse                           plumbinginterrogators
> diff --git a/git.c b/git.c
> index 39582cf511..6d439e723f 100644
> --- a/git.c
> +++ b/git.c
> @@ -558,6 +558,7 @@ static struct cmd_struct commands[] = {
>         { "replace", cmd_replace, RUN_SETUP },
>         { "rerere", cmd_rerere, RUN_SETUP },
>         { "reset", cmd_reset, RUN_SETUP },
> +       { "restore", cmd_restore, RUN_SETUP | NEED_WORK_TREE },
>         { "rev-list", cmd_rev_list, RUN_SETUP | NO_PARSEOPT },
>         { "rev-parse", cmd_rev_parse, NO_PARSEOPT },
>         { "revert", cmd_revert, RUN_SETUP | NEED_WORK_TREE },
> --
> 2.21.0.rc1.337.gdf7f8d0522
>
Previous: Nguyễn Thái Ngọc DuyNext: Duy Nguyen
Message 3 of 97 in “And new command "restore"”
  1. 00/11 And new command "restore"Nguyễn Thái Ngọc Duy, Mar 8, 2019
  2. 01/11 checkout: split part of it to new command 'restore'Nguyễn Thái Ngọc Duy, Mar 8, 2019
  3. Elijah NewrenMar 8, 2019
  4. Duy NguyenMar 9, 2019
  5. Elijah NewrenMar 9, 2019
  6. Duy NguyenApr 9, 2019
  7. Duy NguyenMar 25, 2019
  8. Elijah NewrenMar 25, 2019
  9. Johannes SchindelinMar 13, 2019
  10. 02/11 restore: take tree-ish from --source option insteadNguyễn Thái Ngọc Duy, Mar 8, 2019
  11. Elijah NewrenMar 9, 2019
  12. Eric SunshineMar 10, 2019
  13. 03/11 restore: make pathspec mandatoryNguyễn Thái Ngọc Duy, Mar 8, 2019
  14. Elijah NewrenMar 9, 2019
  15. 04/11 restore: disable overlay mode by defaultNguyễn Thái Ngọc Duy, Mar 8, 2019
  16. Elijah NewrenMar 9, 2019
  17. 05/11 checkout: factor out worktree checkout codeNguyễn Thái Ngọc Duy, Mar 8, 2019
  18. 06/11 restore: add --worktree and --indexNguyễn Thái Ngọc Duy, Mar 8, 2019
  19. Elijah NewrenMar 9, 2019
  20. Eric SunshineMar 10, 2019
  21. Duy NguyenMar 13, 2019
  22. Junio C HamanoMar 13, 2019
  23. 07/11 restore: default to --source=HEAD when only --index is specifiedNguyễn Thái Ngọc Duy, Mar 8, 2019
  24. Elijah NewrenMar 9, 2019
  25. 08/11 restore: support --patchNguyễn Thái Ngọc Duy, Mar 8, 2019
  26. Elijah NewrenMar 9, 2019
  27. 09/11 t: add tests for restoreNguyễn Thái Ngọc Duy, Mar 8, 2019
  28. Andrei RybakMar 9, 2019
  29. Johannes SchindelinMar 13, 2019
  30. Duy NguyenMar 13, 2019
  31. Johannes SchindelinMar 13, 2019
  32. Duy NguyenMar 14, 2019
  33. Junio C HamanoMar 14, 2019
  34. Junio C HamanoMar 14, 2019
  35. 10/11 completion: support restoreNguyễn Thái Ngọc Duy, Mar 8, 2019
  36. Elijah NewrenMar 9, 2019
  37. Duy NguyenMar 11, 2019
  38. Duy NguyenMar 11, 2019
  39. Elijah NewrenMar 11, 2019
  40. 11/11 doc: promote "git restore"Nguyễn Thái Ngọc Duy, Mar 8, 2019
  41. Elijah NewrenMar 9, 2019
  42. Randall S. BeckerMar 11, 2019
  43. Duy NguyenMar 11, 2019
  44. Junio C HamanoMar 13, 2019
  45. Duy NguyenApr 11, 2019
  46. Duy NguyenMar 10, 2019
  47. Jacob KellerMar 10, 2019
  48. Elijah NewrenMar 11, 2019
  49. Junio C HamanoMar 13, 2019
  50. Duy NguyenMar 14, 2019
  51. 00/16 Add new command 'restore'Nguyễn Thái Ngọc Duy, Apr 11, 2019
  52. 01/16 checkout: split part of it to new command 'restore'Nguyễn Thái Ngọc Duy, Apr 11, 2019
  53. 02/16 restore: take tree-ish from --source option insteadNguyễn Thái Ngọc Duy, Apr 11, 2019
  54. 03/16 restore: make pathspec mandatoryNguyễn Thái Ngọc Duy, Apr 11, 2019
  55. 04/16 restore: disable overlay mode by defaultNguyễn Thái Ngọc Duy, Apr 11, 2019
  56. 05/16 checkout: factor out worktree checkout codeNguyễn Thái Ngọc Duy, Apr 11, 2019
  57. 06/16 restore: add --worktree and --stagedNguyễn Thái Ngọc Duy, Apr 11, 2019
  58. 07/16 restore: reject invalid combinations with --stagedNguyễn Thái Ngọc Duy, Apr 11, 2019
  59. 08/16 restore: default to --source=HEAD when only --staged is specifiedNguyễn Thái Ngọc Duy, Apr 11, 2019
  60. 09/16 restore: replace --force with --ignore-unmergedNguyễn Thái Ngọc Duy, Apr 11, 2019
  61. 10/16 restore: support --patchNguyễn Thái Ngọc Duy, Apr 11, 2019
  62. 11/16 t: add tests for restoreNguyễn Thái Ngọc Duy, Apr 11, 2019
  63. 12/16 completion: support restoreNguyễn Thái Ngọc Duy, Apr 11, 2019
  64. 13/16 user-manual.txt: prefer 'merge --abort' over 'reset --hard'Nguyễn Thái Ngọc Duy, Apr 11, 2019
  65. 14/16 doc: promote "git restore"Nguyễn Thái Ngọc Duy, Apr 11, 2019
  66. 15/16 rm: add --staged as alias for --cachedNguyễn Thái Ngọc Duy, Apr 11, 2019
  67. 16/16 help: move git-diff and git-reset to different groupsNguyễn Thái Ngọc Duy, Apr 11, 2019
  68. Junio C HamanoApr 12, 2019
  69. Duy NguyenApr 13, 2019
  70. Duy NguyenApr 17, 2019
  71. Junio C HamanoApr 18, 2019
  72. Duy NguyenApr 18, 2019
  73. Johannes SchindelinApr 18, 2019
  74. Duy NguyenApr 18, 2019
  75. 00/16 Add new command 'restore'Nguyễn Thái Ngọc Duy, Apr 25, 2019
  76. 01/16 checkout: split part of it to new command 'restore'Nguyễn Thái Ngọc Duy, Apr 25, 2019
  77. 02/16 restore: take tree-ish from --source option insteadNguyễn Thái Ngọc Duy, Apr 25, 2019
  78. 03/16 restore: make pathspec mandatoryNguyễn Thái Ngọc Duy, Apr 25, 2019
  79. 04/16 restore: disable overlay mode by defaultNguyễn Thái Ngọc Duy, Apr 25, 2019
  80. 05/16 checkout: factor out worktree checkout codeNguyễn Thái Ngọc Duy, Apr 25, 2019
  81. 06/16 restore: add --worktree and --stagedNguyễn Thái Ngọc Duy, Apr 25, 2019
  82. 07/16 restore: reject invalid combinations with --stagedNguyễn Thái Ngọc Duy, Apr 25, 2019
  83. 08/16 restore: default to --source=HEAD when only --staged is specifiedNguyễn Thái Ngọc Duy, Apr 25, 2019
  84. 09/16 restore: replace --force with --ignore-unmergedNguyễn Thái Ngọc Duy, Apr 25, 2019
  85. 10/16 restore: support --patchNguyễn Thái Ngọc Duy, Apr 25, 2019
  86. 11/16 t: add tests for restoreNguyễn Thái Ngọc Duy, Apr 25, 2019
  87. SZEDER GáborJun 13, 2019
  88. 12/16 completion: support restoreNguyễn Thái Ngọc Duy, Apr 25, 2019
  89. 13/16 user-manual.txt: prefer 'merge --abort' over 'reset --hard'Nguyễn Thái Ngọc Duy, Apr 25, 2019
  90. 14/16 doc: promote "git restore"Nguyễn Thái Ngọc Duy, Apr 25, 2019
  91. 15/16 help: move git-diff and git-reset to different groupsNguyễn Thái Ngọc Duy, Apr 25, 2019
  92. 16/16 Declare both git-switch and git-restore experimentalNguyễn Thái Ngọc Duy, Apr 25, 2019
  93. Emily ShafferMay 7, 2019
  94. Junio C HamanoMay 7, 2019
  95. Duy NguyenMay 7, 2019
  96. Emily ShafferMay 7, 2019
  97. Duy NguyenMay 8, 2019

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.