Closes#47064Closes#20806
The previous implementation used Zed's custom fsevent crate, which
spawned a dedicated thread for each watched path. On large projects,
this could result in 100+ threads just for filesystem watching.
This PR removes the `fsevent` crate and switches to using `notify's`
FsEventWatcher directly. The notify implementation maintains a single
FSEvents stream that watches all paths on one thread. When paths are
added or removed, the stream is stopped, the path list is updated, and
the stream is restarted.
Ref: https://github.com/notify-rs/notify/blob/main/notify/src/fsevent.rs
As a result, Zed now uses one thread for filesystem watching regardless
of how many paths are watched.
Release Notes:
- On macOS, Zed now uses significantly fewer resources when watching
filesystem changes
---------
Signed-off-by: Marco Mihai Condrache <52580954+marcocondrache@users.noreply.github.com>
Co-authored-by: Anthony Eid <anthony@zed.dev>
This PR swaps the @ icon button in the message editor for a + one, which
opens a dropdown that displays context options you can add to the agent.
Aside from removing one step if you're wanting to add context first with
the mouse (in comparison to just inserting @ in the message editor),
this menu will also house skills you've created, whenever we get to
support to that. It also works to surface images and selections in a bit
more visible way as context options. So, effectivelly, this is a bit of
foundation work for further features to come. Here's what it looks like:
<img width="500" height="586" alt="Screenshot 2026-01-27 at 11 38@2x"
src="https://github.com/user-attachments/assets/551686ba-4629-4317-9177-1e942512a23c"
/>
Note that all the options you see in the menu should also be available
through simply typing @ in the message editor.
Release Notes:
- Agent: Added a menu for inserting context more easily with the mouse
in the agent panel.
The same way you can use `size` when width and height have the same
value, it should also be possible to write `min_size` and `max_size`.
This PR adds these two methods, which are also supported in Tailwind
(but not documented in v4, surprisingly!).
Release Notes:
- N/A
This PR adds some UI improvements to the ACP registry UI, already
including the URL for the soon-to-be-released blog post. I'm also
renaming the action to open the page from `zed: agent registry` to `zed:
acp registry`.
Release Notes:
- N/A
Closes#46104
Release Notes:
- Fixed language injections sticking after language comment removed
---
This is working well for the Python SQL comments
```python
from pyspark.sql import SparkSession
spark = SparkSession.builder.getOrCreate()
# sql
cmd = "SELECT col1, col2 FROM tbl"
df = spark.sql( # sql
"""
WITH cte AS (
SELECT col1, col2, COUNT(*) AS n
FROM tbl
GROUP BY ALL
)
SELECT * FROM cte
"""
)
```
And go comments
```go
package main
func test() {
var _ = /* sql */ `SELECT id, name FROM products`
var _ = /* sql */ "SELECT id, name FROM products"
var _ = /* sql */ `SELECT id, name FROM products`
var _ = /* sql*/ "SELECT id, name FROM products"
}
```
Note: `f04b252dd9` was a simpler implementation that worked for
non-inline comments in Python, but produced the following Go behavior
---------
Co-authored-by: Joseph T. Lyons <JosephTLyons@gmail.com>
This PR reworks the `Plan` type, now that we don't need to be concerned
about the legacy plan versions.
We've also made the deserialization layer more robust, which should make
it easier to add new plan variants in the future without needing to go
through this same song and dance.
Release Notes:
- N/A
Add support for Vim's `gdefault` option which makes the `:substitute`
command replace all matches in a line by default, instead of just the
first match. When enabled, the `/g` flag inverts this behavior.
- Add `vim.gdefault` setting
- Add `:set gdefault`, `:set nogdefault` (and short forms `:set gd`, `:set nogd`)
- Fix handling of multiple `/g` flags so that each one inverts the one before
Closes#36209
Release Notes:
- vim: Add `vim.gdefault` setting to make `/g` (replace all matches in a line) the default for substitutions, along with `:set gdefault` and `:set nogdefault` commands (short forms: `gd`, `nogd`)
---------
Co-authored-by: dino <dinojoaocosta@gmail.com>
This is in preparation of merging
https://github.com/zed-industries/zed/pull/47322, because our MacWatcher
implementation was able to watch root directories, and I wanted to
maintain that behavior to decrease the chance of any bugs occurring from
a merge
Note: Release candidate 9 for Notify was created two days ago which
means that a new version is coming out soon. We should update to v9 once
it's released to include bug fixes from upstream.
Release Notes:
- N/A
Closes#47656
Release Notes:
- Fixed extension uninstall button not working when an upgrade is
available.
---------
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
The `}` and `{` paragraph motions now correctly treat only truly empty
lines (zero characters) as paragraph boundaries, matching vim's
documented behavior. Whitespace-only lines are no longer treated as
boundaries.
Changed `start_of_paragraph()` and `end_of_paragraph()` in
`editor/src/movement.rs` to check `line_len() == 0` instead of
`is_line_blank()`.
Note: This change does NOT affect the `ap`/`ip` text objects. Per vim's
`:help ap`, those DO treat whitespace-only lines as boundaries, which is
the existing (correct) behavior in `vim/src/object.rs`.
Closes#36171
Release Notes:
- Fixed vim mode paragraph motions (`}` and `{`) to correctly ignore
whitespace-only lines
---------
Co-authored-by: dino <dinojoaocosta@gmail.com>
This PR extends support to run Go table-test subtests (#35657), handling
tests without explicitly declaring variables for test scenarios.
<img width="611" height="318" alt="go-table-tests"
src="https://github.com/user-attachments/assets/e001b661-b512-4183-b6df-6c25f3af4f27"
/>
Release Notes:
- Improved support to run Go table-test subtests, handling tests without
explicitly declaring variables for test scenarios.
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Implements relative line jumping, a common feature i missed from
Vim/Neovim where you can jump to lines relative to your current cursor
position.
- New `{"relative": true}` flag for `go_to_line::Toggle` action
- Supports both `-5` (5 lines up) and `b5` (backward 5 lines) syntax
- Full lines only (no column support in relative mode, wasnt sure if
that would be necessary)
- Unbound by default - users can add e.g. `"ctrl-j":
["go_to_line::Toggle", {"relative": true}]` to their keymap
Example usage:
- `5` → jump 5 lines down
- `-3` or `b3` → jump 3 lines up
- `0` → stay on current line
[Screencast_20260115_191312.webm](https://github.com/user-attachments/assets/395c0e7b-8ac1-48c8-a39e-5ade4f9206ec)
Release Notes:
- Added relative line jump support to go-to-line action via
`+/-/f/b/F/B`
# Add "Reopen with Encoding" feature (Local/Single user)
## Summary
This PR adds a "Reopen with Encoding" feature to allow users to manually
specify an encoding and reload the active buffer.
This feature allows users to explicitly specify the encoding and reload
the file to resolve garbled text caused by incorrect detection.
## Changes
1. Added encoding picker logic to `encoding_selector`
- Implemented a modal UI accessible via the command palette, shortcuts,
or by clicking the encoding status in the status bar.
- Allows users to select from a list of supported encodings (Shift JIS,
EUC-JP, UTF-16LE, etc.).
2. Updated Buffer logic (crates/language)
- Added a `force_encoding_on_next_reload` flag to the Buffer struct.
- Updated the `reload` method to check this flag and apply the following
logic:
- **Non-Unicode (e.g., Shift JIS):** Bypasses heuristics (like BOM
checks) to force the specified encoding.
- **Unicode (e.g., UTF-8):** Performs standard BOM detection. This
ensures that the BOM is correctly handled/consumed when switching back
to UTF-8.
3. UI / Keymap
- Made the encoding status in the status bar (ActiveBufferEncoding)
clickable.
- Added default keybindings:
- macOS: cmd-k n
- Linux/Windows: ctrl-k n
- Windows: ctrl-k n
## Limitations & Scope
To ensure stability and keep the PR focused, the following scenarios are
intentionally out of scope:
1. **Collaboration and Remote Connections**
- Encoding changes are disabled when collaboration (is_shared) or SSH
remote connections (is_via_remote_server) are active.
- **Reason:** Synchronizing encoding state changes between host/guest or
handling remote reloads involves complex synchronization logic. This PR
focuses on local files only.
`Remote Connection (SSH/WSL)`
|Via status bar|Via shortcut/command|
|:---:|:---:|
|<img width="767" height="136" alt="remote_tooltip"
src="https://github.com/user-attachments/assets/6c7cb293-2486-4f6d-a3ff-2086d939398e"
width="400" />|<img width="742" height="219" alt="remote_shortcut"
src="https://github.com/user-attachments/assets/5448f199-2066-4baf-b349-a983ab2fa77a"
width="400" />|
`Collaboration Session `
|Via status bar|Via shortcut/command|
|:---:|:---:|
|<img width="734" height="86" alt="collab_tooltip"
src="https://github.com/user-attachments/assets/37de99a9-dd33-4c78-98bf-20654d41fdd0"
/>|<img width="720" height="182" alt="collab_pop"
src="https://github.com/user-attachments/assets/91d03ea7-f029-442a-8236-55234576f7ed"
/>|
2. Dirty State
- The feature is disabled if the buffer has unsaved changes to prevent
data loss during reload.
|Via status bar|Via shortcut/command|
|:---:|:---:|
|<img width="545" height="103" alt="local_dirty_tooltip"
src="https://github.com/user-attachments/assets/d9ae658e-52b3-4ecd-9873-d0ec8bd51b5d"
/>|<img width="707" height="178" alt="local_dirty_pop"
src="https://github.com/user-attachments/assets/d170ea1e-9fcb-42e7-aa3e-0555b4a19d86"
/>|
3. Files detected as Binary
Files that worktree detects as "binary" (e.g., UTF-16 files without BOM
containing non-ASCII characters) are not opened in the editor, so this
feature cannot be triggered.
**Future Work**: Fixing this would require modifying crates/worktree
heuristics or exposing a "Force Open as Text" action for InvalidItemView
to trigger. Given the scope and impact, this is deferred to a future PR.
## Test Plan
I verified the feature and BOM handling using the following scenarios:
### Preparation
Used the following test files:
-
[**test_utf8.txt**](https://github.com/user-attachments/files/24548803/test_utf8.txt):
English-only text file. No BOM.
-
[**test_utf8_bom.txt**](https://github.com/user-attachments/files/24548822/test_utf8_bom.txt):
English-only text file. With BOM.
-
[**test_utf8_jp_bom.txt**](https://github.com/user-attachments/files/24548825/test_utf8_jp_bom.txt):
UTF-8 with BOM file containing Japanese characters.
-
[**test_shiftjis_jp.txt**](https://github.com/user-attachments/files/24548827/test_shiftjis_jp.txt):
Shift-JIS file containing Japanese characters (content designed to
trigger misdetection, e.g., using only half-width katakana).
Used an external editor (VS Code or Notepad) for verification.
### Case 1: English-only file behavior
1. Open an English-only UTF-8 file (test_utf8.txt).
2. Reopen as Shift JIS.
3. **Result:**
- Text appearance remains unchanged (since ASCII is compatible).
- Status bar updates to "Shift JIS".
### Case 2: Fixing Mojibake
1. Open a Shift-JIS file (test_shiftjis_jp.txt) that causes detection
failure.
※Confirm it opens with mojibake
2. Select Shift JIS from the status bar selector.
3. **Result:**
- Mojibake is resolved, and Japanese text is displayed correctly.
- Status bar updates to "Shift JIS".
### Case 3: Unicode file with BOM behavior
1. Open an English-only UTF-8 with BOM file (test_utf8_bom.txt).
2. Reopen as `Shift JIS`.
3. **Result:**
- The BOM bytes are displayed as mojibake at the beginning of the file.
- The rest of the English text is displayed normally (ASCII
compatibility).
- Status bar updates to "Shift JIS".
### Case 4: Non-Unicode file with BOM behavior
1. Open a UTF-8 with BOM file containing Japanese
(test_utf8_jp_bom.txt).
2. Reopen as Shift JIS.
3. **Result:**
- The BOM bytes at the start are displayed as mojibake.
- The Japanese text body is displayed as mojibake (UTF-8 bytes
interpreted as Shift JIS).
- Status bar updates to "Shift JIS" (no BOM indicator).
### Case 5: Revert to Unicode
1. From the state in Case 4 (Shift JIS with mojibake), reopen as UTF-8.
2. **Result:**
- The BOM mojibake at the start disappears (consumed).
- The text returns to normal.
- Status bar updates to "UTF-8 (BOM)".
### Case 6: External BOM removal (State sync)
1. Open a UTF-8 with BOM file in Zed (test_utf8_bom.txt).
2. Open the same file in an external editor and save it as UTF-8 (No
BOM).
3. Refocus Zed.
4. **Result:**
- Text appearance remains unchanged.
- The (BOM) indicator disappears from the status bar.
- Saving in Zed and checking externally confirms the BOM is gone.
### Case 7: External BOM addition
1. From the state in Case 6 (UTF-8 No BOM), save as UTF-8 with BOM in
the external editor.
2. Refocus Zed.
3. **Result:**
- The (BOM) indicator appears in the status bar.
- Saving in Zed and checking externally confirms the BOM is present.
### Case 8: External Encoding Change (Auto-detect sync)
1. Open an English-only UTF-8 file in Zed (`test_utf8.txt`).
* *Status bar shows: "UTF-8".*
2. Open the same file in an external editor and save it as **UTF-16LE
with BOM**.
3. Refocus Zed.
4. **Result:**
* The text remains readable (no mojibake).
* **Status bar automatically updates to "UTF-16LE (BOM)".** (Verifies
that `buffer.encoding` is correctly updated during reload).
Release Notes:
- Added "Reopen with Encoding" feature (currently supported for local
files).
---------
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
This PR makes it so that `open_diff_since`, as used by the branch diff,
is able to return a stable diff entity instead of creating a new one
every time this is called. In particular, the base text buffer for this
diff is now stable across branch diff refreshes, making it usable with
the side-by-side view. The strategy for keeping the diff entities alive
is the same one that we use for the uncommitted and unstaged diffs--the
`GitStore` only holds a weak pointer to each "since" diff, so when these
diffs are no longer in use by the branch diff multibuffer they can be
cleaned up.
Release Notes:
- N/A
---------
Co-authored-by: Zed Zippy <234243425+zed-zippy[bot]@users.noreply.github.com>
Closes#46509
Release Notes:
- Fixed: `workspace::SendKeystrokes` would not allow remapping keys in
different keyboard layouts
---------
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Follow up to #47714 because it is still broken for the extension
workflows post #47699 cache introduction. Keeping the cache around
though in the hopes that it will now finally work.
Release Notes:
- N/A
Following the changes to the namespace caching, CI in the extensions
broke because cache conflicts arose there seemingly. Trying to set the
target here more explicitly in an effort to fix this.
Release Notes:
- N/A
When there are no search results, it's clear that a search is still
ongoing because the landing page says "Searching...". However, once
there's at least one result found, it becomes completely unclear when
search is actually finished. This adds a spinner immediately to the
right of the result counter that stops once the search is finished.
https://github.com/user-attachments/assets/a0ca4e2a-c506-42a4-bc4b-c1eb32d69e79
Release Notes:
- Added spinner to project search, indicating that a search is still
ongoing.
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
Fixes#41776
## Problem
Old AI Text Thread sessions fail to open from History because the
deserializer fails on unknown icon variants. When icons are removed or
renamed in refactors, old saved threads become unloadable with errors
like:
```
unknown variant `AtSign`, expected one of `Ai`, `AiAnthropic`...
```
## Solution
Added a lenient deserializer for the `icon` field in
`SlashCommandOutputSection` that falls back to `IconName::Code` for
unknown variants.
This ensures old saved threads remain loadable even as icons are
added/removed from the codebase.
## Test Plan
- Added unit test for valid icon deserialization
- Added unit test for unknown icon fallback to `Code`
- Added unit test for various unknown icon variants
- Added unit test confirming serialization unchanged
- All tests pass: `cargo test -p assistant_slash_command`
Release Notes:
- Fixed old AI text thread sessions failing to open from History when
they contain icons that were removed in previous updates.
🤖 Generated with [Claude Code](https://claude.ai/code)
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: MrSubidubi <finn@zed.dev>
Closes #ISSUE
Measure the model reversing user edits programatically, for a more
robust and consistent measurement than LLM-as-judge
Release Notes:
- N/A *or* Added/Fixed/Improved ...
