We might want to calculate (commit_id, range) pairs of consecutive lines in
order to "absorb" changes, for example.
This should also be cheaper since Vec<u8> doesn't have to be allocated per line.
Perhaps, get_same_line_map() could return an iterator, but implementing an
iterator to be "pull"-ed is much harder than writing a function to "push",
especially when lifetime is involved.
This function was short, and this change makes it clear that !.is_empty() was
redundant. Duplicated doc comment is also removed. I feel the inline comment is
easier to follow here.
It no longer makes sense to initialize Source line_map and build
HashMap<Commit, Source> in one function. Let's extract the line_map
initialization to a function instead.
All intermediate nodes are changed to RevWalk of Result<IndexPosition, _> type
to pass BackendError around from filter predicates. Leaf ancestors/descendants
computation is unchanged, and mapped to Result at revset_engine layer. This is
simpler than converting all RevWalk impls to Result<_, _>.
We'll need to propagate error from predicate function, so .filter() will no
longer be usable. .map() will be used in order to wrap infallible ancestry
lookup with Ok(_).
Some RevsetImpl methods are migrated to .map() as example.
I don't see measurable performance difference, but VecDeque is theoretically
simpler than BTreeSet. The input is sorted, so we never do random insertion.
This also allows some minor optimizations to be performed, such as
avoiding recomputation of the connected target set when
`MoveCommitsTarget::Roots` is used since the connected target set is
identical to the target set (all descendants of the roots).
is_empty() could also return Result<bool, _>, but I think the current definition
is also good. If an error occurred, revset.iter() would return at least one
item, so it's not empty.
Let's say we're updating one parent of a merge:
```
E E'
/|\ /|\
B C D -> B C D'
\|/ \|/
A A
```
When rebasing `E` to create `E'` there, we do that by merging the
changes compared to the auto-merged parents. The auto-merged parents
before is `B+(C-A)+(D-A)`, and after it's `B+(C-A)+(D'-A)`. Then we
rebase the diff, which gives us `E' = B+(C-A)+(D'-A) + (E -
(B+(C-A)+(D-A))) = D' + (E - D')`.
However, we currently don't do quite that simplification because we
first resolve conflicts when possible in the two auto-merged parent
trees (before and after). That rarely makes a difference to the
result, but it's wasteful to do it. It does make a difference in some
cases where our merge algorithm is lossy, which currently is only the
"A+(A-B)=A" case. I added a test case showing where it does make a
difference. It's a non-obvious cases but I think the new behavior is
more correct (the old behavior was a conflict).
This was added at f5f61f6bfe36 "revset: resolve 'HEAD@git' just like other
pseudo @git branches." As I said in this patch, there was no practical use case
of the HEAD@git symbol.
Suppose we implement colocated workspaces/worktrees #4436, there may be multiple
Git HEAD revisions. This means HEAD can no longer be abstracted as a symbol of
the "git" remote.
Custom backends may rely on networking or other unreliable implementations to support revsets, this change allows them to return errors cleanly instead of panicking.
For simplicity, only the public-facing Revset and RevsetGraph types are changed in this commit; the internal revset engine remains mostly unchanged and error-free since it cannot generally produce errors.
Clippy 1.83 (currently in beta) detects more cases of unneeded lifetimes,
namely in trait implementation declarations. Since this lint is warn by
default, we need to fix those instances to get a clean CI.
This is required when performing `rebase -s a -s b` where "b" is a
descendant of "a". Both "a" and "b" should be regarded as the roots of
the target set and be rebased onto the new destination.
This allows for `RebaseOptions` to be respected. This will be used when
migrating `rebase --source`/`rebase --branch` to use `move_commits` to
respect the `--before`/`--after` options.
The `coalesce` function takes a list of revsets and returns the commits in the
first revset in the list which evalutes to a non-empty set of commits.
It can be used to display fallbacks if a certain commit cannot be found,
e.g. `coalesce(present(user_configured_trunk), builtin_trunk)`.
A new module is added to jj_lib which exposes a function
get_annotation_for_file. This annotates the given file line by line with
commit information according to the commit that made the most recent
change to the line.
Similarly, a new command is added to the CLI called `jj file annotate` which
accepts a file path. It then prints out line by line the commit
information for the line and the line itself. Specific commit
information can be configured via the templates.annotate_commit_summary
config variable
This appears to be a bit faster if there are tons of unchanged ranges.
```
group new old
----- --- ---
bench_diff_git_git_read_tree_c 1.00 58.5±0.12µs 1.07 62.7±0.60µs
bench_diff_lines/modified/10k 1.00 34.2±0.72ms 1.08 37.0±1.09ms
bench_diff_lines/modified/1k 1.00 3.1±0.08ms 1.12 3.5±0.01ms
bench_diff_lines/reversed/10k 1.00 28.0±0.15ms 1.01 28.4±0.51ms
bench_diff_lines/reversed/1k 1.00 616.0±16.20µs 1.00 617.0±9.29µs
bench_diff_lines/unchanged/10k 1.00 3.5±0.04ms 1.10 3.9±0.06ms
bench_diff_lines/unchanged/1k 1.00 328.4±4.44µs 1.07 352.0±1.41µs
```
One particular use case for these is escape sequences -- and to that
end, I'm also adding `\e` as a shorthand for `\x1b`.
Change-Id: Id000000040ea6fd8e2d720219931485960c570dd
