Fixes#7378
Includes a regression test that previously failed and now succeeds.
* [x] I have followed the instructions in the PR template
---------
Co-authored-by: Emil Ernerfeldt <emil.ernerfeldt@gmail.com>
Co-authored-by: Lucas Meurer <hi@lucasmerlin.me>
* Follow up to #7146
Previously when galleys were splitted, each exept the last had an extra
empty row that had to be removed when they were concated. This changes
it to remove the `\n` from the layout jobs when splitting.
* [x] I have followed the instructions in the PR template
Splitting this out from the Parley work as requested. This removes
`FontImage` and makes the font atlas use a `ColorImage`. It converts
alpha to coverage at glyph-drawing time, not at delta-upload time.
This doesn't do much now, but will allow for color emoji rendering once
we start using Parley.
I've changed things around so that we pass in `text_alpha_to_coverage`
to the `Fonts` the same way we do with `pixels_per_point` and
`max_texture_side`, reusing the existing code to check if the setting
differs and recreating the font atlas if so. I'm not quite sure why this
wasn't done in the first place.
I've left `ImageData` as an enum for now, in case we want to add support
for more texture pixel formats in the future (which I personally think
would be worthwhile). If you'd like, I can just remove that enum
entirely.
Closes#7077.
This fixes the problem shown in #7077 where clearing a `TextEdit`
wouldn't reset its cursor position. I've fixed that by adding back the
`TextCursorState::range` method, which clamps the selection range to
that of the passed `Galley`, and calling it in the same places where it
was called before #5785.
(/cc @juancampa)
* [x] I have followed the instructions in the PR template
Fixes a regression introduced in https://github.com/emilk/egui/pull/5411
(possibly
d74bee536f)
that breaks `leading_space` handling.
I think this is what the condition should be but I haven't touched this
code in a while.
## What
(written by @emilk)
When editing long text (thousands of line), egui would previously
re-layout the entire text on each edit. This could be slow.
With this PR, we instead split the text into paragraphs (split on `\n`)
and then cache each such paragraph. When editing text then, only the
changed paragraph needs to be laid out again.
Still, there is overhead from splitting the text, hashing each
paragraph, and then joining the results, so the runtime complexity is
still O(N).
In our benchmark, editing a 2000 line string goes from ~8ms to ~300 ms,
a speedup of ~25x.
In the future, we could also consider laying out each paragraph in
parallel, to speed up the initial layout of the text.
## Details
This is an ~~almost complete~~ implementation of the approach described
by emilk [in this
comment](<https://github.com/emilk/egui/issues/3086#issuecomment-1724205777>),
excluding CoW semantics for `LayoutJob` (but including them for `Row`).
It supersedes the previous unsuccessful attempt here:
https://github.com/emilk/egui/pull/4000.
Draft because:
- [X] ~~Currently individual rows will have `ends_with_newline` always
set to false.
This breaks selection with Ctrl+A (and probably many other things)~~
- [X] ~~The whole block for doing the splitting and merging should
probably become a function (I'll do that later).~~
- [X] ~~I haven't run the check script, the tests, and haven't made sure
all of the examples build (although I assume they probably don't rely on
Galley internals).~~
- [x] ~~Layout is sometimes incorrect (missing empty lines, wrapping
sometimes makes text overlap).~~
- A lot of text-related code had to be changed so this needs to be
properly tested to ensure no layout issues were introduced, especially
relating to the now row-relative coordinate system of `Row`s. Also this
requires that we're fine making these very breaking changes.
It does significantly improve the performance of rendering large blocks
of text (if they have many newlines), this is the test program I used to
test it (adapted from <https://github.com/emilk/egui/issues/3086>):
<details>
<summary>code</summary>
```rust
use eframe::egui::{self, CentralPanel, TextEdit};
use std::fmt::Write;
fn main() -> Result<(), eframe::Error> {
let options = eframe::NativeOptions {
..Default::default()
};
eframe::run_native(
"editor big file test",
options,
Box::new(|_cc| Ok(Box::<MyApp>::new(MyApp::new()))),
)
}
struct MyApp {
text: String,
}
impl MyApp {
fn new() -> Self {
let mut string = String::new();
for line_bytes in (0..50000).map(|_| (0u8..50)) {
for byte in line_bytes {
write!(string, " {byte:02x}").unwrap();
}
write!(string, "\n").unwrap();
}
println!("total bytes: {}", string.len());
MyApp { text: string }
}
}
impl eframe::App for MyApp {
fn update(&mut self, ctx: &egui::Context, _frame: &mut eframe::Frame) {
CentralPanel::default().show(ctx, |ui| {
let start = std::time::Instant::now();
egui::ScrollArea::vertical().show(ui, |ui| {
let code_editor = TextEdit::multiline(&mut self.text)
.code_editor()
.desired_width(f32::INFINITY)
.desired_rows(40);
let response = code_editor.show(ui).response;
if response.changed() {
println!("total bytes now: {}", self.text.len());
}
});
let end = std::time::Instant::now();
let time_to_update = end - start;
if time_to_update.as_secs_f32() > 0.5 {
println!("Long update took {:.3}s", time_to_update.as_secs_f32())
}
});
}
}
```
</details>
I think the way to proceed would be to make a new type, something like
`PositionedRow`, that would wrap an `Arc<Row>` but have a separate `pos`
~~and `ends_with_newline`~~ (that would mean `Row` only holds a `size`
instead of a `rect`). This type would of course have getters that would
allow you to easily get a `Rect` from it and probably a `Deref` to the
underlying `Row`.
~~I haven't done this yet because I wanted to get some opinions whether
this would be an acceptable API first.~~ This is now implemented, but of
course I'm still open to discussion about this approach and whether it's
what we want to do.
Breaking changes (currently):
- The `Galley::rows` field has a different type.
- There is now a `PlacedRow` wrapper for `Row`.
- `Row` now uses a coordinate system relative to itself instead of the
`Galley`.
* Closes <https://github.com/emilk/egui/issues/3086>
* [X] I have followed the instructions in the PR template
---------
Co-authored-by: Emil Ernerfeldt <emil.ernerfeldt@gmail.com>
Might want to draw from `interaction.interact_radius` style instead of
hard-coding the margin, but I didn't want to create a breaking change.
If desired, I can follow up with a separate PR to address that concern.
* Closes <https://github.com/emilk/egui/issues/5796>
* [x] I have followed the instructions in the PR template
Enabled the `missing_assert_message` lint
* [x] I have followed the instructions in the PR template
---------
Co-authored-by: Lucas Meurer <lucasmeurer96@gmail.com>
This removes the `expand(1.0)` on text background colors, since it makes
translucent background colors have bad looking bleeding.
There is probably a smarter solution than disabling the highlighting
entirely, but I don't see a way to do that while keeping the area
consumed consistent between translucent/solid colors, or adding a decent
step up in complexity.
Since this makes it impossible to tell if selected text is highlighted,
this also adds a blanket `0.5` gamma multiply to the text selection
background color. If that is undesirable because it's a bad arbitrary
number choice, or if it's too much of an unexpected change and just the
default values should be changed, please let me know.
These changes cause the tests that use screenshots with highlighted text
to fail, though I am not sure how to update those tests to match the
changes.
<details>
<summary>Comparison Images</summary>
Current:
After changes:
</details>
<details>
<summary>Code used to make comparison images</summary>
```rs
fn color_text_format(ui: &Ui, color: Color32) -> TextFormat {
TextFormat { font_id: FontId::monospace(ui.text_style_height(&egui::TextStyle::Monospace)), background: color, ..Default::default() }
}
fn color_sequence_galley(ui: &Ui, text: &str, colors: [Color32; 3]) -> Arc<Galley> {
let mut layout_job = LayoutJob::default();
for color in colors {
layout_job.append(text, 0.0, color_text_format(ui, color));
}
ui.fonts(|f| f.layout_job(layout_job))
}
fn color_sequence_row(ui: &mut Ui, label_text: &str, text: &str, colors: [Color32; 3]) {
ui.label(label_text);
ui.label(color_sequence_galley(ui, text, colors));
ui.end_row();
}
egui::Grid::new("comparison display").show(ui, |ui| {
ui.ctx().set_pixels_per_point(2.0);
let transparent = Color32::TRANSPARENT;
let solid = Color32::RED;
let solid_2 = Color32::GREEN;
let translucent_1 = Color32::GRAY.gamma_multiply(0.5);
let translucent_2 = Color32::GREEN.gamma_multiply(0.5);
color_sequence_row(ui, "Transparent to Solid:", " ", [transparent, solid, transparent]);
color_sequence_row(ui, "Translucent to Transparent:", " ", [transparent, translucent_1, transparent]);
color_sequence_row(ui, "Solid to Transparent:", " ", [solid, solid_2, solid]);
color_sequence_row(ui, "Solid to Solid:", " ", [solid, transparent, solid]);
color_sequence_row(ui, "Solid to Translucent:", " ", [solid, translucent_1, solid]);
color_sequence_row(ui, "Translucent to Translucent:", " ", [translucent_1, translucent_2, translucent_1]);
color_sequence_row(ui, "Transparent to Solid:", "a", [transparent, solid, transparent]);
color_sequence_row(ui, "Translucent to Transparent:", "a", [transparent, translucent_1, transparent]);
color_sequence_row(ui, "Solid to Transparent:", "a", [solid, solid_2, solid]);
color_sequence_row(ui, "Solid to Solid:", "a", [solid, transparent, solid]);
color_sequence_row(ui, "Solid to Translucent:", "a", [solid, translucent_1, solid]);
color_sequence_row(ui, "Translucent to Translucent:", "a", [translucent_1, translucent_2, translucent_1]);
})
```
</details>
* [x] I have followed the instructions in the PR template
---------
Co-authored-by: Emil Ernerfeldt <emil.ernerfeldt@gmail.com>
* [x] I have followed the instructions in the PR template
This PR implements `AsRef<[u8]>` for `FontData`, allowing it to be
passed into `fontdb`'s
[`Source`](https://docs.rs/fontdb/0.16.2/fontdb/enum.Source.html) type.
This would allow `egui` and `cosmic_text` to share font data with
eachother
<!--
Please read the "Making a PR" section of
[`CONTRIBUTING.md`](https://github.com/emilk/egui/blob/master/CONTRIBUTING.md)
before opening a Pull Request!
* Keep your PR:s small and focused.
* The PR title is what ends up in the changelog, so make it descriptive!
* If applicable, add a screenshot or gif.
* If it is a non-trivial addition, consider adding a demo for it to
`egui_demo_lib`, or a new example.
* Do NOT open PR:s from your `master` branch, as that makes it hard for
maintainers to test and add commits to your PR.
* Remember to run `cargo fmt` and `cargo clippy`.
* Open the PR as a draft until you have self-reviewed it and run
`./scripts/check.sh`.
* When you have addressed a PR comment, mark it as resolved.
Please be patient! I will review your PR, but my time is limited!
-->
* Closes N/A, but this is part of
https://github.com/emilk/egui/issues/3378
* [x] I have followed the instructions in the PR template
Other text layout libraries in Rust--namely, Parley and Cosmic
Text--have one canonical text cursor type (Parley's is a byte index,
Cosmic Text's also stores the line index). To prepare for migrating egui
to one of those libraries, it should also have only one text cursor
type. I also think simplifying the API is a good idea in and of
itself--having three different cursor types that you have to convert
between (and a `Cursor` struct which contains all three at once) is
confusing.
After a bit of experimentation, I found that the best cursor type to
coalesce around is `CCursor`. In the few places where we need a
paragraph index or row/column position, we can calculate them as
necessary.
I've removed `CursorRange` and `PCursorRange` (the latter appears to
have never been used), merging the functionality with `CCursorRange`. To
preserve the cursor position when navigating row-by-row, `CCursorRange`
now stores the previous horizontal position of the cursor.
I've also removed `PCursor`, and renamed `RowCursor` to `LayoutCursor`
(since it includes not only the row but the column). I have not renamed
either `CCursorRange` or `CCursor` as those names are used in a lot of
places, and I don't want to clutter this PR with a bunch of renames.
I'll leave it for a later PR.
Finally, I've removed the deprecated methods from `TextEditState`--it
made the refactoring easier, and it should be pretty easy to migrate to
the equivalent `TextCursorState` methods.
I'm not sure how many breaking changes people will actually encounter. A
lot of these APIs were technically public, but I don't think many were
useful. The `TextBuffer` trait now takes `&CCursorRange` instead of
`&CursorRange` in a couple of methods, and I renamed
`CCursorRange::sorted` to `CCursorRange::sorted_cursors` to match
`CursorRange`.
I did encounter a couple of apparent minor bugs when testing out text
cursor behavior, but I checked them against the current version of egui
and they're all pre-existing.
egui never accesses the `FontDefinitions`' member fields mutably, except
in `fonts_tweak_ui` where it cloned the `FontDefinitions` object anyway.
This patch reduces system memory consumption for shared font
definitions.
And also removes some overhead from copying (e.g. for the per
`pixel_per_points` font atlas)
Also it allows to keep a copy of the font definitions outside of egui.
In my App that uses international fonts:
Before:
New:
Note: If `Arc` is not wanted, then it could ofc be abstracted away.
I know this is quite a breaking change API wise, but would like to hear
your opinion.
When using fonts with an average of 50,000 characters,
'epaint texture atlas overflowed!' may be printed and cause problems.
It is necessary to expand the max value related to texture.
* Closes#5256
---------
Co-authored-by: Emil Ernerfeldt <emil.ernerfeldt@gmail.com>
make it easier to add fonts.
For example if I want to add a custom FontFamily or if the user wants to
add a Chinese fallback
* [x] I have followed the instructions in the PR template
---------
Co-authored-by: Emil Ernerfeldt <emil.ernerfeldt@gmail.com>
* Closes https://github.com/emilk/egui/pull/5106
* Closes https://github.com/emilk/egui/issues/5084
Protect against rounding errors in egui layout code.
Say the user asks to wrap at width 200.0.
The text layout wraps, and reports that the final width was 196.0
points.
This than trickles up the `Ui` chain and gets stored as the width for a
tooltip (say).
On the next frame, this is then set as the max width for the tooltip,
and we end up calling the text layout code again, this time with a wrap
width of 196.0.
Except, somewhere in the `Ui` chain with added margins etc, a rounding
error was introduced,
so that we actually set a wrap-width of 195.9997 instead.
Now the text that fit perfectly at 196.0 needs to wrap one word earlier,
and so the text re-wraps and reports a new width of 185.0 points.
And then the cycle continues.
So this PR limits the text wrap-width to be an integer.
Related issues:
* https://github.com/emilk/egui/issues/4927
* https://github.com/emilk/egui/issues/4928
* https://github.com/emilk/egui/issues/5163
---
Pleas test this @rustbasic
* Closes https://github.com/emilk/egui/issues/4976
* Part of #4378
* Implements parts of #843
### Background
Some widgets (like `Grid` and `Table`) needs to know the width of future
elements in order to properly size themselves. For instance, the width
of the first column of a grid may not be known until all rows of the
grid has been added, at which point it is too late. Therefore these
widgets store sizes from the previous frame. This leads to "first-frame
jitter", were the content is placed in the wrong place for one frame,
before being accurately laid out in subsequent frames.
### What
This PR adds the function `ctx.request_discard` which discards the
visual output and does another _pass_, i.e. calls the whole app UI code
once again (in eframe this means calling `App::update` again). This will
thus discard the shapes produced by the wrongly placed widgets, and
replace it with new shapes. Note that only the visual output is
discarded - all other output events are accumulated.
Calling `ctx.request_discard` should only be done in very rare
circumstances, e.g. when a `Grid` is first shown. Calling it every frame
will mean the UI code will become unnecessarily slow.
Two safe-guards are in place:
* `Options::max_passes` is by default 2, meaning egui will never do more
than 2 passes even if `request_discard` is called on every pass
* If multiple passes is done for multiple frames in a row, a warning
will be printed on the screen in debug builds:
### Breaking changes
A bunch of things that had "frame" in the name now has "pass" in them
instead:
* Functions called `begin_frame` and `end_frame` are now called
`begin_pass` and `end_pass`
* `FrameState` is now `PassState`
* etc
### TODO
* [x] Figure out good names for everything (`ctx.request_discard`)
* [x] Add API to query if we're gonna repeat this frame (to early-out
from expensive rendering)
* [x] Clear up naming confusion (pass vs frame) e.g. for `FrameState`
* [x] Figure out when to call this
* [x] Show warning on screen when there are several frames in a row with
multiple passes
* [x] Document
* [x] Default on or off?
* [x] Change `Context::frame_nr` name/docs
* [x] Rename `Context::begin_frame/end_frame` and deprecate the old ones
* [x] Test with Rerun
* [x] Document breaking changes
<!--
Please read the "Making a PR" section of
[`CONTRIBUTING.md`](https://github.com/emilk/egui/blob/master/CONTRIBUTING.md)
before opening a Pull Request!
* Keep your PR:s small and focused.
* The PR title is what ends up in the changelog, so make it descriptive!
* If applicable, add a screenshot or gif.
* If it is a non-trivial addition, consider adding a demo for it to
`egui_demo_lib`, or a new example.
* Do NOT open PR:s from your `master` branch, as that makes it hard for
maintainers to test and add commits to your PR.
* Remember to run `cargo fmt` and `cargo clippy`.
* Open the PR as a draft until you have self-reviewed it and run
`./scripts/check.sh`.
* When you have addressed a PR comment, mark it as resolved.
Please be patient! I will review your PR, but my time is limited!
-->
I removed (I hope so) all wildcard imports I found.
For me on my pc this improved the build time:
- for egui -5s
- for eframe -12s
* [x] I have followed the instructions in the PR template
This allows license checking tools to omit the OFL and UFL licenses when
`default_fonts` are turned off.
There was some discussion of versioning on the original issue; I have
chosen to label this version as `0.28.1` to match the other crates.
Happy to adjust the version as needed.
<!--
Please read the "Making a PR" section of
[`CONTRIBUTING.md`](https://github.com/emilk/egui/blob/master/CONTRIBUTING.md)
before opening a Pull Request!
* Keep your PR:s small and focused.
* The PR title is what ends up in the changelog, so make it descriptive!
* If applicable, add a screenshot or gif.
* If it is a non-trivial addition, consider adding a demo for it to
`egui_demo_lib`, or a new example.
* Do NOT open PR:s from your `master` branch, as that makes it hard for
maintainers to test and add commits to your PR.
* Remember to run `cargo fmt` and `cargo clippy`.
* Open the PR as a draft until you have self-reviewed it and run
`./scripts/check.sh`.
* When you have addressed a PR comment, mark it as resolved.
Please be patient! I will review your PR, but my time is limited!
-->
* Closes <https://github.com/emilk/egui/issues/2321>
* [X] I have followed the instructions in the PR template
---------
Co-authored-by: Alex Pinkus <pinkus@amazon.com>
Previously, many labels had non-integer widths. This lead to rounding
errors.
This was most notable for the new `Area` sizing code:
We would run the initial sizing pass, to measure the size of e.g. a
tooltip.
Say the tooltip contains text that was 100.123 ui points wide. With a
16pt border, that becomes 116.123, which is stored in the `Area` state
as the width. The next frame, we use that stored size as the wrapping
width. With perfect precision, we would then tell the label to wrap to
100.123 pts, which the text would _just_ fit in. However, due to
rounding errors we might end up asking it to wrap to 100.12**2** pts,
meaning the last word would now wrap and end up on the next line.
By rounding label sizes to perfect integers, we avoid such rounding
errors, and most ui elements will now end up on perfect integer point
coordinates (and `f32` can precisely express and do arithmetic on all
integers < 2^24).
Visually this has very little impact. Some labels move by a pixel here
and there, mostly for the better.
