I spent a few hours in heap profilers. What I found was:
* Inefficient use of heap when building up runs of
`Action::Print(char)`.
-> Solve by adding `Action::PrintString(String)`
and accumulating utf8 bytes rather than u32 codepoints.
* Inefficient use of heap when building Quad buffers: the default
exponential growth of `Vec` tended to waste 40%-75% of the allocated
capacity, and since we could keep ~1024 of these in cache, there's
a lot of potential for waste.
-> Solve by bounding the growth to 64 at a time. This has similar
characteristics to exponential growth at the default 80x24 terminal
size. May need to add a config option for this step size for users
with very large terminals.
* Lazy eviction from the LFU caches. The underlying cache advisor is
somewhat probabilistic and has a minimum cache size of 256, making
it difficult to maintain low heap utilization.
-> Solve by replacing it with a very simple LFU algorithm. It doesn't
seem to hurt much at the default terminal size with the default
cache sizes. If we make the cache sizes smaller, its overhead is
reduced.
Some further experimentation is needed to adjust defaults, but this
should help reduce heap usage.
refs: https://github.com/wez/wezterm/issues/2626
I've expanded the number of bits from 16->32 without impacting
the overall struct sizes and reserved 2 bits for super/subscript.
I refer to these as vertical alignment properties for conceptual
consistency with css.
SGR 73, 74, 75 are used to set super, sub and normal vertical alignment.
These are compatible with mintty.
However, mintty just added support for setting both attributes to render in
small caps in 06ac446049
(https://github.com/mintty/mintty/issues/1171)
According to its benchmarks, it's almost 2x faster than
unicode_segmentation. It doesn't appear to make a visible
difference to `time cat bigfile`, but I'll take anything
that gives more headroom for such little effort of switching.
Action is used to encode parsed terminal output and shuttle it between
the thread that does the parsing and the main gui thread that applies
it to the terminal model.
Take it down from 184 bytes to 40 bytes (on 64-bit systems). This seems
to boost `time cat bigfile` by reducing the runtime to ~40% of its prior
duration: down from 8s -> 4.5s on an M1 macbook air.
Size reductions achieved by Box'ing relatively less frequently
used enum variants. The kitty image data variant is particularly
large, and the Window variant is also pretty heavy.
Remove the faulty cache key stuff and hang references to the cached
data directly off the underlying Line. That makes the association
between the the Line and the data O(1) plus some basic cache
invalidation checks.
Adjust the shape cache portion of this to use the same ID for the
shape, so that things that invalidate just the quads (such as cursor
movement when not composing, and selection) only need to recompute
the quads without re-shaping.
Adds Pane::for_each_logical_line_in_stable_range_mut and
Pane::with_lines_mut which allow iterating mutably over lines.
The idea is that this will allow the renderer to directly cache
data in the Line via its appdata without having to build cumbersome
external caching logic and managing cache keys.
This commit just swaps the implementation around for localpane
and sanity checks that the renderer functions.
Various overlays and the mux client don't properly implement these
yet and current warn at compile time and panic at runtime.
To follow is the logic to cache the data and make sure that it
works the way that I think before converting the other Pane
implementations.
I'd forgotten that this was in here.
The recentish fixes to flush for various terminal queries
likely need this to be 100% good. Surprised this hasn't
come up as a bug already :-/
This makes those fields usable in `wezterm cli list --format json`.
This doesn't change the ABI of the mux protocol, but prior to
this commit, those fields were always 0.
refs: #2319
Just clamp it to the current physical, non-scrollback portion
of the screen.
This avoids a panic but doesn't address the screen size mismatch
in the associated issue.
refs: https://github.com/wez/wezterm/issues/2133
The recent work on the scrollback made it easier to constrain the
search region, so expose those parameters to the Pane::search
interface and to the mux protocol.
Use those new parameters to constrain quickselect search to
1000 rows above and below the current viewport by default, and
add a new parameter to QuickSelectArgs that allows overriding that
range.
A follow-up commit could make the search/copy overlay issue a series
of searches in chunks so that it avoids blocking the UI when
searching very large scrollback.
refs: https://github.com/wez/wezterm/pull/1317
This reduces the resident memory by another ~10% because it avoids
keeping as many runs of whitespace.
Runtime for `time cat enwiki8.wiki` is still ~11-12s, resident: 530K
refs: https://github.com/wez/wezterm/issues/1626
The previous commit added the option to convert the storage to
the cluster format. That saves memory as rows are moved to scrollback,
but makes scrolling back more expensive due to that conversion.
This commit adds a fast(ish) path for the common case of simply
appending text to a new line before it gets scrolled: the default
format for lines in the screen is now the cluster format and,
provided that the cursor moves from left to right as text is
emitted, can simply append to the cluster storage in-place
and avoids a conversion when the line is moved to scrollback.
This improves the throughput of `time cat enwiki8.wiki` so
that the runtime is typically around 11-12s (compared to 9-10s
before introducing cluster storage). However, this is often
a deal of variance in the measured time and I believe that
that is due to the renderer triggering conversions back to
the vec storage and introducing slowdowns from the render side.
That's what I'll investigate next.
This commit causes lines to be "compressed" (really, just translated
to the new clustered line storage variant) as they move into scrollback.
The memory savings are significant for large scrollback:
`wezterm -n --config scrollback_lines=1000000`
`time cat enwiki8.wiki`
before: ~9s, Resident: 2.1G
after: ~15s, Resident: 620K (!)
The performance impact is non-trivial, and I will dig into that
next.
refs: https://github.com/wez/wezterm/issues/1626
Previously, the mux layer had no internal understanding of titles other
than the Pane::get_title method to return state from a pane.
Users have asked for ways to explicitly set titles on windows and tabs,
so this commit is a step towards that.
The mux window and tab objects now store a title string.
The terminal layer now emits Alert::WindowTitleChanged when the window
title is changed via eg: OSC 0 or OSC 2.
The mux layer will respond to Alert::WindowTitleChanged by resolving the
window that corresponds to the source pane and amending its title.
The MuxWindow and MuxTab objects now provide accessor methods for the
title.
TabInformation (as used by format-tab-title and format-window-title) now
exposes the underlying window_id as well as tab_title and window_title.
The tab title can be changed via the lua MuxTab type, but there is not
currently an escape sequence associated with this.
The defaults for format-tab-title and format-window-title don't
currently consider these new title strings.
refs: https://github.com/wez/wezterm/issues/1598
This enables tentative support for https://sw.kovidgoyal.net/kitty/keyboard-protocol
It's only been lightly tested with the notcurses-input program and
eyeballed against a few random keypresses in kitty running
`printf "\x1b[=11u" ; od -c`
I tried with neovim, but it doesn't seem like the version available
in Fedora 36 supports this yet.
refs: https://github.com/wez/wezterm/issues/1141
Main thing to note here is that the open crate has deprecated
open::that_in_background, but made open::that non-blocking.
I think this is OK, but I'm a little cagey about what will
happen with this on Windows. We may need to spawn our own
thread for this if things go awry.
Avoid using serde for mapping between Lua and Rust for the `Config`
struct.
This improves the build speed of the config crate by 2x; it goes down
from 30 seconds to 9 seconds on my 5950x.
I want to make some changes to the scrollback structure that
first require routing some operations through an accessor
method, so this commit does that.
It should have no functional difference.
I've bundled this into termwiz's UnicodeVersion type as that is
a similar concept that is already routed through to the appropriate
function.
refs: https://github.com/wez/wezterm/issues/1888
I'm not totally sure this is all there is to it, but in #1850
the response is showing up after vim was exited, so it seems like
it didn't fully receive these responses.
refs: #1850
We're now capable of remembering an alpha value for everything
in the palette and the main window theme, but not the tab bar theme.
Whether the alpha is correctly respected at render time is a different
story!
refs: #1835
The bug here was that each paint call in a window would update
the focus state of its panes to reflect the one that had focus.
However, it didn't account for the actual window focus; it would
just assume that it was focused.
The result was that the perceived focus would alternate between each of
the windows in the wezterm process, and if you were running an
application that had enabled focus tracking, those events could cause a
repaint and drive up the CPU utilization.
This commit addresses that by gating the focus update to only occur
when we have the focus, and for extra safety, avoid generating focus
events at the terminal layer if the new state matches the current state.
refs: https://github.com/wez/wezterm/issues/1838
Go directly to the underlying env_logger crate, as pretty_env_logger
hasn't been updated in some time, and I'd like to be able to redirect
the log output to a file more directly, and that feature is in a newer
version of the env logger than pretty_env_logger was pulling in.
This persuades vim to set ttymouse=sgr by default.
We're carefully sitting below 279 which triggers
some queries for things that we don't respond to yet.
refs: https://github.com/wez/wezterm/issues/1825
I think this looks less jarring than the yellowish selection color,
and the translucency looks nicer than changing the fg color completely
black as it did previously.
* gui: improve mouse text selection
* implement mouse press capture between the terminal and UI, so when you
start selecting text from the terminal the tabs won't activate and
vice-versa
* selecting from the top and bottom lines won't scroll the viewport
anymore, it will only scroll if the mouse is moved out of line bounds
* change cell selection so that it behaves like text selection usually
does in other popular software
refs: https://github.com/wez/wezterm/issues/1199
refs: https://github.com/wez/wezterm/issues/1386
refs: https://github.com/wez/wezterm/issues/354
Previously, we'd just default the resize quirk on for all programs on
Windows.
That pessimizes the otherwise fine behavior of `wezterm ssh` or mux
connections to unixy platforms.
This commit moves the quirk out of the config trait and makes it
a runtime property of the terminal, and then arranges for it
to be set when we know that we set up a conpty for the terminal.
refs: #1265
* #217 DECSDM, and commented pseudocode for improved handling of transparent image cells
* oops forgot cargo fmt --all
* #217 set initial palette colors to match VT340
* #217 cleanup color_map initialization, fix sixel scrolling
* #217 remove large comment - transparency discussion for another time
This change also allows removing the dep on the palette crate,
which I found to be difficult to use (API changed often, and relied
on a lot of `.into` that was hard to follow and reconcile across
upgrades). We already pulled in the csscolorparse crate as an indirect
dep of colorgrad, so we can replace the color conversion we need for
sixel with that crate while we're in here.
refs: #1615
Previously overlooked because we didn't render or otherwise
change behavior: after running vttest and running ls, some
lines still had double width/height set.
This ensures that we remove that attribute when clearing
the screen.
This was used by the mux client when the session is tardy.
I've heard feedback that the greying out is distracting and not
especially useful, so I don't think this is any great loss.
This commit allows the following configuration:
```
wezterm -n --config 'colors = { selection_fg = "clear", selection_bg = "rgba:50% 50% 50% 50%" }'
```
which sets the selection_bg to fully transparent, and selection_bg to
50% transparent gray.
When selection_fg is fully transparent we'll use the normal fg color.
When selection_bg is partially (or fully!) transparent, it will be
alpha blended over the current cell background color.
To support this, the config file will now accept rgba colors specified
as 4 whitespace delimited numeric values. If a value ends with `%` it
is interpreted as a number in the range 0-100. Otherwise, it is
interpreted as a number in the range 0-255. The 4 values are
red, green, blue, alpha.
At this time, only the selection_fg and selection_bg settings accept
alpha values.
refs: #1615
Resolves a little bit of the awkward duplication of color types
between some of the crates by factoring them a little bit better.
This is prep for allowing specifying alpha for some colors
in the config.
This commit refines bidi property handling:
* experimental_bidi has been split into two new configuration settings;
`bidi_enabled` (which controls whether the terminal performs implicit
bidi processing) and `bidi_direction` which specifies the base
direction and whether auto detection is enabled.
* The `Line` type can now store those bidi properties (they are actually
split across 3 bits representing enabled, auto-detection and
direction)
* The terminal now has a concept of active bidi properties and default
bidi properties
* The default properties are pulled from the wezterm configuration
* active bidi properties are potentially set via escape sequences,
BDSM (which sets bidi_enabled) and SCP (which sets bidi_direction).
We don't support the 2501 temporary dec private mode suggested by
the BIDI recommendation doc at this time.
* When creating new `Line`'s or clearing from the start of a `Line`, the
effective bidi properties are computed (from the active props,
falling back to default propr) and applied to the `Line`.
* When rendering the line, we now look at its bidi properties instead
of just the global config.
The default bidi properties are `bidi_enabled: false` and
`bidi_direction: LeftToRight` which corresponds to the typical
bidi-unaware mode of most terminals.
It is possible to live reload the config to change the effective
defaults, but note that they apply, by design, to new lines being
processed through the terminal. That means existing output is
left unaffected by a config reload, but subsequently printed lines
will respect it. Pressing CTRL-L or otherwise contriving to have
the running application refresh its display should cause the
refreshed display to update and apply the new bidi mode.
refs: #784
This setting now allows specifying what the canonical format is;
previously it was a boolean that meant "don't change anything"
if false, and rewrite to CRLF if true.
It's now an enum with None, CR, LF, and CRLF variants that express
more control.
The config accepts true (maps to CRLF) and false (maps to None)
for backwards compatibility.
The default behavior is unchanged by this commit, however, I did
uncover a bug in the canonicalizer for inputs like `\r\r\n`.
refs: #1575
This commit enables the following config to work for local (not mux yet!)
panes:
```lua
local wezterm = require 'wezterm'
wezterm.on("format-tab-title", function(tab, tabs, panes, config, hover, max_width)
if tab.is_active then
return {
{Background={Color="blue"}},
{Text=" " .. tab.active_pane.title .. " "},
}
end
local has_unseen_output = false
for _, pane in ipairs(tab.panes) do
if pane.has_unseen_output then
has_unseen_output = true
break;
end
end
if has_unseen_output then
return {
{Background={Color="Orange"}},
{Text=" " .. tab.active_pane.title .. " "},
}
end
return tab.active_pane.title
end)
return {
}
```
refs: https://github.com/wez/wezterm/discussions/796
This commit decomposes the main get_semantic_zones method into two
parts:
* A per-line portion, where the line ranges are cached (invalidated on
change)
* The overall screen portion, where the line ranges are merged
This changes the overall complexity of computing zones from
O(width * scrollback-height)
To an incremental:
O((width * number of changed lines since last query) + scrollback-height)
You can see some samples of elapsed time below; those show the times for
running both the old and the new implementation on the same data. The
number of lines/zones in the scrollback increases with each call and you
can see that the new implementation is a bit faster anyway at low
volumes but is significantly faster as the number of lines/zones
increases, because the amount of work is reduced.
```
get_semantic_zones: 71.708µs
get_semantic_zones_new: 59.041µs
get_semantic_zones: 71.166µs
get_semantic_zones_new: 9.166µs
get_semantic_zones: 44.291µs
get_semantic_zones_new: 4.208µs
get_semantic_zones: 69.791µs
get_semantic_zones_new: 10.291µs
get_semantic_zones: 59.375µs
get_semantic_zones_new: 7.958µs
get_semantic_zones: 52.5µs
get_semantic_zones_new: 4.5µs
get_semantic_zones: 91.791µs
get_semantic_zones_new: 20.916µs
get_semantic_zones: 229.916µs
get_semantic_zones_new: 109.208µs
get_semantic_zones: 224.125µs
get_semantic_zones_new: 15.208µs
get_semantic_zones: 291.791µs
get_semantic_zones_new: 11.833µs
get_semantic_zones: 238.875µs
get_semantic_zones_new: 12.625µs
get_semantic_zones: 468.458µs
get_semantic_zones_new: 126.583µs
get_semantic_zones: 460.5µs
get_semantic_zones_new: 25.666µs
get_semantic_zones: 358.291µs
get_semantic_zones_new: 19.541µs
get_semantic_zones: 436.833µs
get_semantic_zones_new: 17.875µs
get_semantic_zones: 313.166µs
get_semantic_zones_new: 15.25µs
get_semantic_zones: 333.958µs
get_semantic_zones_new: 16.541µs
get_semantic_zones: 364.666µs
get_semantic_zones_new: 14.041µs
```
* #1457 Initial commit compiles but doesn't yet work
* #1457 Fix cell offsets
* #1457 refactor, ClickPosition
* fix nits
* #1457 filter per-pixel motion when not in SGR-Pixels mode
* Much cleaner match condition for filtering mouse events based on encoding protocol
It's a little ambiguous whether we should cancel the mode when moving
the cursor explicitly; I opted to do it when we move it due to newline
processing, or when its y position is changed.
refs: #1539
user vars were stubbed out. This commit adds storage for them
in the mux client and adds a new notification that publishes each
var as it is changed. That differential stream is applied to the
storage in the mux client when it is received.
```lua
local wezterm = require 'wezterm'
wezterm.on("update-right-status", function(window, pane)
local woot = pane:get_user_vars().woot
window:set_right_status(tostring(woot))
end);
return {
unix_domains = {
{name="unix"},
},
}
```
then running:
* `wezterm connect unix`
* in that session: `printf "\033]1337;SetUserVar=%s=%s\007" woot `echo -n nice | base64``
causes `nice` to show in the status area.
refs: #1528
We need 100% of the info for it to work correctly, so this commit:
* Exposes the keyboard encoding mode via the Pane trait
* Adds the scan code to the RawKeyEvent
* Has the GUI perform the encoding if the keyboard is set that way
* Removes the basic encoder from termwiz in favor of the gui level one
The net result is that we bypass the Pane::key_up/Pane::key_down methods
in almost all cases when the encoding mode is set to win32-input-mode.
There is now a config option: allow_win32_input_mode that can be
used to prevent using this mode.
refs: #1509
This commit causes the terminal to emit win32-input-mode encoded key up
and down events for a limited subset of keys When win32-input-mode is
enabled.
We limit them to keys where we know the VK key code equivalent,
and where those keys are either not representable (eg: modifier
only key events), or may generate ambiguous output (eg: CTRL-SPACE
in different keyboard layouts).
However, in my experiments, modifier only key presses confuse powershell
and cause it to emit `@`, so I've disabled that in the code for now.
refs: https://github.com/wez/wezterm/issues/318
refs: https://github.com/wez/wezterm/issues/1509
refs: https://github.com/wez/wezterm/issues/1510
Nothing generates them right now, and the mux client has no
way to represent them on the wire.
I'm considering constraining them to just win32 for now.
refs: https://github.com/wez/wezterm/issues/1509
This is a baby step that formalizes the different encoding schemes into
an enum, and hooks up the decset sequence for win32-input-mode.
It doesn't change any of the actual encoding at this time.
refs: #1509
