The focused main view's click/enter/e/G handlers all need the same thing: given
a rendered diff row, the patch-space line it corresponds to. Until now that came
solely from delta's lazygit-edit:// hyperlinks, which only carry a path and a
single line number — no side. That's lossy: for a deletion the number is the old
line, but the consumers fed it into new-file lookups, and two consecutive
deletions (which share a new-file line number) couldn't be told apart at all.
Replace GetFileAndLineForClickedDiffLine with GetDiffLineInfo, returning the
fuller (file, type, new-line, old-line) record from diff-line-metadata-notes.md.
This is mechanism #1: parse the decolorized view buffer — walk up to the file's
"diff --git" section, reuse patch.Parse on it (splitting multi-file commit diffs
on the "diff --git" boundaries), and read the type and line numbers off the patch
arithmetic. It serves the structure-preserving renderings — no pager, git diff
--color, and delta --color-only without line numbers — with no external
dependency.
To avoid trusting a mis-parse, the parser bails when a hunk's body no longer
matches its header (Patch.IsWellFormed). That's what happens when a pager keeps
the diff/hunk headers but restructures the body: delta's line-number gutters push
the +/- marker off the start of each line, so every body line reads as context.
Such renderings fall through to the next backend rather than yielding a confident
wrong answer. (diff-so-fancy goes further and rewrites the headers too, so it
fails even earlier, on the missing "diff --git".)
GetDiffLineInfo is a seam with swappable backends: the buffer parser first, then
the old hyperlink reader as a fallback for renderings the parser can't handle
(delta's default mode, or delta with line-number gutters). The future #2 OSC
per-cell metadata reader plugs in ahead of both, behind the same record shape.
Wire the consumers to the record per that doc's field mapping:
- dive into staging/patch building lands on the exact patch line, looking a
deletion up by its old-file line number (PatchLineForOldLineNumber) so the
two-deletions case resolves correctly;
- `e` edits at the new-file line;
- `G` anchors the PR link on the left (old) side for a deletion, the right (new)
side otherwise.
The hyperlink fallback can't convey the side, so it reports DiffLineOther, which
the consumers treat as a non-deletion — i.e. exactly today's behavior.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The patch package can already map a patch line index to its new-file line
number and back (LineNumberOfLine / PatchLineForLineNumber). The diff-line
parser being built on top of this needs the old-file equivalents: a deletion
has no distinct new-file position (two consecutive deletions share one), so to
land on the exact deletion when diving into staging we have to look it up by
old-file line number.
Add OldLineNumberOfLine and PatchLineForOldLineNumber as direct mirrors of the
new-file functions, counting DELETION+CONTEXT lines instead of ADDITION+CONTEXT.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was already possible, but only when a file was selected, and it woudln't
always land on the right line when a pager was used. Now it's also possible to
do this for directories, and it jumps to the right line.
At the moment this is a hack that relies on delta's hyperlinks, so it only works
on lines that have hyperlinks (added and context).
The implementation is very hacky for other reasons too (e.g. the addition of the
weirdly named ClickedViewRealLineIdx to OnFocusOpts).
The custom-patch git operations (move/pull/delete patch, and their rebase
continuations) run on worker goroutines and call PatchBuilder.Reset when
they've consumed the patch, clearing To and the fileInfoMap. Meanwhile the
UI thread reads that state every layout — the options bar and the mode
indicator both call Active() — so the reset raced the render.
Add a mutex. The map's entries are only ever touched on the UI thread, so
the lock only has to serialize the To field and the fileInfoMap pointer:
readers snapshot the pointer under the lock and iterate the local, and
getFileInfo drops the lock across its git diff I/O rather than holding it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
When loading the files of a commit we passed --no-renames, so a rename
showed up as a separate delete and add rather than a single R entry.
That made it impossible to work with a rename that also modifies the
file: the modifications were spread across a full deletion and a full
addition instead of appearing as the handful of lines that actually
changed. The staging view already shows renames and lets you stage
their hunks, so there was no good reason for the patch builder to
differ; the flag was only there because the commit-file parser couldn't
cope with the rename record format.
Switch the commit-file loader and the per-file diff to --find-renames,
teach the parser about the rename record (a status followed by two
paths), and carry the previous path through the patch builder so the
diff for a rename is loaded with both paths, which is what makes git
emit the rename in the first place.
A whole-file selection keeps the rename in the header, so the rename
moves or is discarded together with the file's contents. A partial
selection instead strips the rename metadata and points the header at
the new path, so applying the patch only changes the contents and
leaves the rename in place; the blob index line is kept so that a 3-way
apply can still fall back to a blob merge.
Discarding a renamed file from a commit now discards both the new and
the old path, so the new file is removed and the old one is restored.
Changing the rename similarity threshold refreshes the commit files
panel too, not just the files panel, so that a rename can turn into a
delete and add or back. It is disabled while building a patch, however,
because the patch builder caches each file's diff by path and would
desync if a rename changed into a delete and add underneath it.
Finally, copying a file's diff from the commit files panel now passes
both paths for a rename, so the copied diff shows the rename instead of
a new-file add.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This fixes the problem; a consequence of this change is that given the following
scenario:
@@ -1,3 +1,3 @@
1
-2
+2b
3
staging only the line `+2b` will put it *before* the unchanged `2` line, rather
than after it as you might expect (the changed unit tests demonstrate this).
Since this should be a pretty uncommon scenario, I guess it is an ok compromise.
As you can see in the changed tests, while the behavior of what gets staged is
fixed now, it doesn't always correctly select the next line to stage. We'll
address this in the next commit.
When entering staging (or patch building) for an added or deleted file, it
doesn't make sense to use hunk mode, because pressing space would stage/unstage
the entire file, and if the user wanted to do that, they would have pressed
space in the Files panel. So always use line mode for added/deleted files by
default, even if the useHunkModeInStagingView user config is on.
Not because it's terribly important to test here (doesn't hurt though), but
because it will be useful in the next commit for a new test we're adding there.
While it's true that the behavior is a little different from the staging panel,
where the staged lines are actually removed from the view and in many cases the
selection stays more or less in the same place, it is still very useful to move
to the next stageable thing in the custom patch building view too.
This improves the experience when selecting a hunk generously with the mouse, by
dragging over it including some context lines above and below. Previously we
would consider the "moving end" of the selection range for whether things need
to be added or removed, but this doesn't make sense if it's a context line. Now
we consider the first actual change line that is included in the range.
It is confusing to get header lines, hunk headers, or context lines rendered as
being included in a custom patch, when including these makes no difference to
the patch.
This is only a visual change; internally, we still record these non-patch lines
as being included in the patch. It doesn't matter though; you can press space on
a header line and nothing happens.
It would probably be cleaner to only record + and - lines in the includedLines
array, but that would be a bit more work, and doesn't seem worth it.
I took the set of enabled checks from revive's recommended configuration [1],
and removed some that I didn't like. There might be other useful checks in
revive that we might want to enable, but this is a nice improvement already.
The bulk of the changes here are removing unnecessary else statements after
returns, but there are a few others too.
[1] https://github.com/mgechev/revive?tab=readme-ov-file#recommended-configuration
There are two ways to jump to the editor on a specific line: pressing `e` in the
staging or patch building panels, or clicking on a hyperlink in a delta diff. In
both cases, this works perfectly in the unstaged changes view, but in other
views (either staged changes, or an older commit) it can often jump to the wrong
line; this happens when there are further changes to the file being viewed in
later commits or in unstaged changes.
This commit fixes this so that you end up on the right line in these cases.
Several custom patch commands on parts of an added file would fail with the
confusing error message "error: new file XXX depends on old contents". These
were dropping the custom patch from the original commit, moving the patch to a
new commit, moving it to a later commit, or moving it to the index.
We fix this by converting the patch header from an added file to a diff against
an empty file. We do this not just for the purpose of applying the patch, but
also for rendering it and copying it to the clip board. I'm not sure it matters
much in these cases, but it does feel more correct for a filtered patch to be
presented this way.
We're going to add another argument in the next commit, and that's getting a bit
much, especially when most of the arguments are bool and you only see true and
false at the call sites without knowing what they mean.
In go 1.22, loop variables are redeclared with each iteration of the
loop, rather than simple updated on each iteration. This means that we
no longer need to manually redeclare variables when they're closed over
by a function.
We're not fully standardising here: different contexts can store their range state however
they like. What we are standardising on is that now the view is always responsible for
highlighting the selected lines, meaning the context/controller needs to tell the view
where the range start is.
Two convenient benefits from this change:
1) we no longer need bespoke code in integration tests for asserting on selected lines because
we can just ask the view
2) line selection in staging/patch-building/merge-conflicts views now look the same as in
list views i.e. the highlight applies to the whole line (including trailing space)
I also noticed a bug with merge conflicts not rendering the selection on focus though I suspect
it wasn't a bug with any real consequences when the view wasn't displaying the selection.
I'm going to scrap the selectedRangeBgColor config and just let it use the single line
background color. Hopefully nobody cares, but there's really no need for an extra config.
Previously, when pressing right-arrow when the cursor is already in the last
hunk, it would jump back to the beginning of that hunk. This can be confusing if
the hunk is long, maybe the start of the hunk is already scrolled off the top of
the window, and then pressing right-arrow actually scrolls *backwards*, which is
counter-intuitive. It's better to do nothing in this case.
Same for left-arrow when the cursor is already in the first hunk, although here
the problem is not so severe (unless diff context was increased by a huge
amount, and the start of the first hunk is scrolled off the bottom of the
window).
There's a bug in LineNumberOfLine, but the existing test coverage doesn't catch
it, as the only test case for this was one where oldStart and newStart were the
same for all hunks. Add a test case where newStart is different for one of the
hunks; this demonstrates a bug, where all expected results from index 12 on are
off by one.
We've been sometimes using lo and sometimes using my slices package, and we need to pick one
for consistency. Lo is more extensive and better maintained so we're going with that.
My slices package was a superset of go's own slices package so in some places I've just used
the official one (the methods were just wrappers anyway).
I've also moved the remaining methods into the utils package.
This fixes the problem that patching would stop at the first file that has a
conflict. We always want to patch all files.
Also, it's faster for large patches, and the code is a little bit simpler too.
All callers in this file now use reverseOnGenerate=false and
keepOriginalHeader=true, so hard-code that in the call to ModifiedPatchForLines
and get rid of the parameters.
The loop is pointless for two reasons:
- git apply --3way has this fallback built in already. If it can't do a
three-way merge, it will fall back to applying the patch normally.
- However, the only situation where it does this is when it can't do a 3-way
merge at all because it can't find the necessary ancestor blob. This can only
happen if you transfer a patch between different repos that don't have the
same blobs available; we are applying the patch to the same repo that is was
just generated from, so a 3-way merge is always possible. (Now that we fixed
the bug in the previous commit, that is.)
But the retry loop is not only pointless, it was actually harmful, because when
a 3-way patch fails with a conflict, git will put conflict markers in the
patched file and then exit with a non-zero exit status. So the retry loop would
try to patch the already patched file again, and this almost certainly fails,
but with a cryptic error message such as "error: main.go: does not exist in
index".
There's no reason to have two different ways of applying patches for whole-file
patches and partial patches; use --reverse for both. Not only does this simplify
the code a bit, but it fixes an actual problem: when reverseOnGenerate and
keepOriginalHeader are both true, the generated patch header is broken (the two
blobs in the line `index 6d1959b..6dc5f84 100644` are swapped). Git fails to do
a proper three-way merge in that case, as it expects the first of the two blobs
to be the common ancestor.
It would be possible to fix this by extending ModifiedPatchForLines to swap the
two blobs in this case; but this would prevent us from concatenating all patches
and apply them in one go, which we are going to do later in the branch.
We are going to add one more flag in the next commit.
Note that we are not using the struct inside patch_manager.go; we keep passing
the individual flags there. The reason for this will become more obvious later
in this branch.
