This PR adds two new tools to Assistant 2:
- `list-worktrees` - Lists the worktrees in a project
- `read-file` - Reads a file at the given path in the project
I don't see `list-worktrees` sticking around long-term, as when we have
tools for listing files those will include the worktree IDs along with
the path, but making this tool available allows the model to utilize
`read-file` when it otherwise wouldn't be able to.
Release Notes:
- N/A
This PR changes the default value when no input is provided with a tool
use from `null` to `{}`.
This fixes an issue I was seeing where tools that didn't accept input
were not being called correctly.
Release Notes:
- N/A
This PR fixes the language model selector.
I tried to piece together the state prior to #25697 (the state it was in
at 11838cf89e) while retaining unrelated
changes that happened since then.
Release Notes:
- Fixed an issue where language models would not be authenticated until
after the model selector was opened (Preview only).
This fixes a performance problem introduced in #25906 and caused by
calling `BufferDiff::snapshot` too frequently.
Release Notes:
- Fixed a performance regression related to buffer diffs
Co-authored-by: Conrad <conrad@zed.dev>
We would previously return an error if there was at least one non-UTF8
file. Now we just ignore them and only add text files. If no text files
are found we show an error.
Release Notes:
- N/A
https://github.com/zed-industries/zed/pull/25663#issuecomment-2686095807
Renamed the `vim::Backspace` and `vim::Space` actions to
`vim::WrappingLeft` and `vim::WrappingRight` respectively. The old names
are still available, but they are marked as deprecated and users are
advised to use the new names.
Also added a paragraph to the docs describing how to enable wrapping
cursor navigation.
These changes tackle two issues with running terminal commands via vim
mode:
- When using `!!` the tab's title was set to `!!` instead of the
previous command that was run and these changes fix that in order to
always display the previous command in the tab's title when re-running
the command with `!!`
- For a terminal command, pressing the rerun button would actually bring
up the task palette, so this has been updated in order to disable the
rerun button when the terminal tab was spawned via a vim command
Closes#25800
Release Notes:
- Fixed the terminal tab title when using `!!` to rerun the last command
- Improved the terminal tab for when command is run via vim mode, in
order to disable the rerun button, seeing as Zed does not support it
Closes#25647
This PR fixes two issues related to cmd + click on URL:
1. Normally cmd + click on URL, it opens browser. Now, alt + tab back to
Zed. If you cmd + click on link again it won't work, until you normal
click some where else in buffer. It won't even show underline.
2. Again, cmd + click on URL, it opens browser. Now, alt + tab back to
Zed. If you cmd + click, some where else in buffer like just normal
text, and now try to hover on URL it won't show up underline and cmd +
click on it won't work. Unless again, if you plain click somewhere else.
Problem:
Issue is when clicking we set pending anchor (for selection), and when
we mouse up we clear those. This works for normal case without pressing
any modifier.
But, in case of cmd modifier, we set pending anchor (set when
`SelectPhase::Begin`), but we don't clear it once we use that data.
Fix:
Once we end up using selection, anchor, etc data to figure out where to
navigate either URL/defination etc, we clear selection just like how we
do it in normal click. This doesn't require to happen after navigate
task, so we do it right after our usage of it.
Before:
https://github.com/user-attachments/assets/b33d93fc-f490-4fa4-ae22-1da1fd6b77a9
After:
https://github.com/user-attachments/assets/028f039a-cd13-4651-b461-3ba52f2526de
Release Notes:
- Fixed an issue where cmd + click on a URL was not working sometimes.
Closes#21974
`resolve_path_in_worktrees` function looks for provided path in each
worktree until valid file is found.
In this PR we priortize current buffer worktree before other worktrees,
because of edge case where, file with same name might exists in other
worktrees.
Updated tests to handle this case.
Release Notes:
- Fixed an issue where the wrong file from a different worktree would
open when using `Cmd + Click` on a file import.
Closes#18157
Release Notes:
- vim: Added `:reg[isters]` to show the current values of registers
---------
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Closes#22536
Changes:
- Visual and visual block: Cursor at start of selection.
- Visual line: Cursor at start on line.
- Uses different handling since the selection does not actually change
in vline.
Release Notes:
- vim: Fixed insert before (`shift-i`) in visual modes.
Also simplify it to avoid doing a bunch of unnecessary work.
Co-Authored-By: Cole <cole@zed.dev>
Closes #ISSUE
Release Notes:
- git: Fix jumping to the previous diff hunk
---------
Co-authored-by: Cole <cole@zed.dev>
Closes#10400
Closes https://github.com/zed-industries/zed/issues/17947
Changes:
- Let vim::increment find boolean values in the line and toggle them.
Release Notes:
- vim: Added support for toggling boolean values with `ctrl-a`/`ctrl-x`
Fixes two issues with the Vim exchange implementation:
1. The clear exchange implementation **didn't** clear the exchange. This
was due to us asking the editor to clear normal highlights instead of
background highlights.
2. Calling clear exchange also wouldn't cause the operator to be
cleared, so you would be left in operator = "cx".
I've added tests for both of these cases.
Partially closes#25750. It doesn't address the problem with dot repeat
not working for my custom bindings, but I don't know what would cause
that. I'd love to hear some thoughts on why that is. That might be a
problem on my part or it might be something with the code. Input would
be appreciated.
Release Notes:
- Fixed: Vim exchange's "clear exchange" function didn't clear the
exchange and kept you in operator pending mode.
This solves a couple of issues with Vim search by making the search
buffer and `SearchableItem` aware of the direction of the search. If
`SearchOptions::BACKWARDS` is set, all operations will be reversed. By
making `SearchableItem` aware of the direction, the correct active match
can be selected when searching backward.
Fixes#22506. This PR does not fix the last problem in that issue, but
that one is also tracked in #8049.
Release Notes:
- Fixes incorrect behavior of backward search in Vim mode
This fixes a major performance issue in the current git beta.
This PR also removes the PopoverButton component, which was easy to
misuse.
Release Notes:
- Git Beta: Fix frame drops caused by opening the git panel
---------
Co-authored-by: Mikayla Maki <mikayla.c.maki@gmail.com>
We were iterating over the row range of a hunk, and inserting into a
hash map for every row.
Release Notes:
- Fixed a performance problem when a large diff hunk was displayed in an
editor.
Fine-tuning the visuals (namely, reducing font and keybinding size) and
passing `on_click` handlers to the Cancel & Regenerate actions.
Release Notes:
- N/A
Just a basic implementation so we can start trying it out.
Release Notes:
- N/A
---------
Co-authored-by: Marshall <marshall@zed.dev>
Co-authored-by: Michael <michael@zed.dev>
This PR makes it so the copy button on Markdown code blocks stays
absolutely positioned even when scrolled:
<img width="1297" alt="Screenshot 2025-03-04 at 5 28 48 PM"
src="https://github.com/user-attachments/assets/b0d0fae9-ccd6-43c1-bef3-44d8d3c3e669"
/>
We achieve this by inserting a new parent element around both the copy
button and the code block itself so we can position the copy button
absolutely within that element.
Release Notes:
- N/A
Release Notes:
- Git Beta: Synchronized selections between the modal editor and the
panel editor
- Git Beta: Allow opening the commit modal even if we're unable to
commit.
Proper version of https://github.com/zed-industries/zed/pull/25425
When https://github.com/zed-industries/zed/pull/24857 returned font
updates on settings changes, settings values, not in-memory ones should
be compared.
This PR returns back the logic finally, and changes it to explicitly
track the settings values, not the in-memory ones.
Also adds the same tracking for UI font changes, which had never been
tracked before.
Release Notes:
- Fixed font sizes not reacting on settings change
- Add global handlers so these actions can be invoked from the command
palette, etc.
- Tweak spinner to not show itself until a remote has been selected
Release Notes:
- N/A
Check that there are no `visible_worktrees` rather than checking
`worktrees` when deciding whether to display the "Open a file or project
to get started" text
Closes#25395
Release Notes:
- Fixed the "Open a file or project to get started" message not always
showing after all buffers have been closed
This PR adds the `git.hunk_style` setting, allowing setting an alternate
style for hunks – specifically the rendering of unstaged hunks.
It has 2 options:
- `transparent` (unstaged hunks are more transparent/less opaque than
staged hunks)
- `pattern (unstaged hunks are indicated by a visual pattern)
We'll possibly explore a VSCode-style "don't show staged hunks", but the
complexity it adds is a bit out of scope for now.
Transparent:
Pattern:
Release Notes:
- Git Beta: Added `git.hunk_style` setting to allow toggling between git
hunk visual styles.
Closes#22999
# Problem
Currently, the default soft wrap mode of an editor is determined by
reading the language-specific settings of the language _at offset zero_
in the editor's (multi)buffer. While this provides a way to pick a
single soft wrap mode for a multi-language multibuffer, it's a bad
choice for a single-buffer multibuffer that begins with a different
embedded language. For example, Markdown with frontmatter:
```markdown
---
my_front_matter
---
# Hello World
```
Setting this in config:
```json
"languages": {
"Markdown": { "soft_wrap": "bounded" }
},
```
Will not soft wrap the Markdown file as the language at offset zero is
YAML.
# Solution
Instead of using the language at offset zero, use the language of the
first buffer in the multibuffer (the buffer at offset zero). This gives
better behavior for single-buffer editors, and a similar default for
multi-language multibuffers as before.
# Testing
All existing `editor` crate tests pass, but I would appreciate any
guidance for where best to add additional testing.
Release Notes:
- Fixed soft_wrap setting not applying to buffers starting with a
different language
---------
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
If the user already binds `tab`/`alt-tab`/`alt-l` to a different action
in a conflicting context and hasn't assigned a different keybinding for
`editor::AcceptEditPrediction`, we would show broken popovers with no
bindings:
Instead, they will now see an error-variant of every popover which
includes a tooltip with a short description and buttons to open the
keymap, and open a new docs section explaining the issue in detail and
how to fix it.
Note: I included the docs change in this PR because it's ok to deploy
before the release, as it also applies to existing versions.
Release Notes:
- edit predictions: Improve UX when there's no keybinding for accepting
predictions
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
Co-authored-by: Danilo <danilo@zed.dev>
Also
- Recomputes `suggested_commit_message` and no longer stores it, to
ensure things are always up to date
- Reduces indentation in `render_footer`
Release Notes:
- N/A
We modified few actions in
https://github.com/zed-industries/zed/pull/25846, which are:
`"editor::GoToHunk" -> ["editor::GoToHunk", { "center_cursor": true }]`
`"editor::GoToPrevHunk" -> ["editor::GoToPrevHunk", { "center_cursor":
true }]`
Also, recently we changed and added migration for:
`["editor::GoToPrevHunk", { "center_cursor": true }] ->
["editor::GoToPreviousHunk", { "center_cursor": true }] `
This means:
1. User that might still have `editor::GoToHunk` won't be automatically
migrated to `["editor::GoToHunk", { "center_cursor": true }]`. Note
value of `center_cursor` is false, in first case (default), and true in
second case.
2. User that might still have `editor::GoToPrevHunk` won't be
automatically migrated to `["editor::GoToPreviousHunk", {
"center_cursor": true }]`. Note, `editor::GoToPrevHunk` is renamed
since, it is now invalid action.
This PR adds those migrations.
cc: @marcospb19
Release Notes:
- N/A
Closes#22447
When navigate forward/back, the focus moves from the ProjectSearchView's
result editor to the Pane, and then move to the ProjectSearchView, but
the event `on_focus_in` not triggered for ProjectSearchView, causing the
result editor to lose focus eventually.
f6dabadaf7/crates/workspace/src/workspace.rs (L1372)f6dabadaf7/crates/workspace/src/workspace.rs (L1385)
Considering that the navigation might be triggered again in the next
frame, so use `on_next_frame` in `on_focus` event to move focus to
result editor.
Next frame:
- the blur event triggered for result editor.
- focus move from ProjectSearchView to result editor in `on_focus` event
for ProjectSearchView
- navigate again, focus moves from result editor to Pane then move back
to ProjectSearchView
- the focus not change during this frame, so no focus event happened for
ProjectSearchView.
Release Notes:
- Fix lost focus when navigate back in project search result
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
Fixes#16057
In newer versions of clangd, the switch labelDetailsSupport in the json
passed to the language server modifies the format of the returned json.
Zed handles well the old format, but misses the function parameters in
the new one. For example:
The old format looks like this:
```json
...
"label": " Window(int width, int height, const char *name, bool vsync, bool resizable)",
...
```
and with labelDetailsSupport = true:
```json
...
"label": " Window",
"labelDetails": {
"detail": "(int width, int height, const char *name, bool vsync, bool resizable)"
},
...
```
A simple solution is to just to not tell the language server that label
details are supported and force it to use the old format. This is a
dirty fix, but makes the completions behave like in the old versions of
clangd.
I do not know if this will break another language server. From what I've
found out most lsp-s do not depend on that setting and provide all
completion data either way. If not, this switch will need to be exposed
in a config or be at least lsp-dependant.
Lastly, I do not know Rust, maybe will need help to make a better fix
for the issue.
Release Notes:
- Fixed broken C++ completion suggestions
Closes#25885
This PR improves the matching for file icons to tabs.
Previously, the tab icon would be resolved based upon the relative path
in the current project. However, this caused the default file icon being
assigned to all files outside of the project, as the relative path for
these files would be empty.
Instead, `path_for_buffer` is now used which always returns a proper
file name even for paths outside the current project (as also stated [in
this
comment](fee9c67707/crates/editor/src/items.rs (L1689))).
As the file name is sufficient for matching icons to files, this fixes
the linked issue whilst not changing anything for previously properly
matched icons.
| `main` | This PR |
| --- | --- |
| <img width="296" alt="main"
src="https://github.com/user-attachments/assets/e72b8b5d-aa1c-4a8e-903f-14239f5b8764"
/> | <img width="296" alt="PR"
src="https://github.com/user-attachments/assets/a736974a-ce41-4861-be3f-95448cc7ffd0"
/> |
Release Notes:
- Fixed wrong file icons being shown for files outside of the current
project.
This does not fix the bug where, when the commit editor modal is open,
changing the staged file does not update the suggested message in the
commit editor. Conrad mentioned he thought we shouldn't be allowed to
change those when the modal is open, so I'm not attempting to fix that.
Release Notes:
- Made suggested commits placeholders and allow them to be committed.
