ratatui-widgets

We are excited to announce the new version of ratatui - a Rust library that's all about cooking up TUIs 👨‍🍳🐀

✨ Release highlights: https://ratatui.rs/highlights/v0302/

⚠️ List of breaking changes can be found here.

Features

• 90639c1 (uncategorized) Add Termina backend by @joshka in #2561

  Summary

  • add the ratatui-termina backend crate using the published termina
    crate
  • expose the backend through the termina feature and Ratatui
    prelude/backend re-exports
  • add a small Termina event-loop example and wire the backend into CI,
    xtask, README generation, and docs

  Refs #1784

  Validation

  • cargo +nightly fmt
  • cargo check -p ratatui-termina --all-features --all-targets
  • cargo check -p ratatui --no-default-features --features termina
  • cargo check -p xtask
  • cargo check -p release-header
  • cargo xtask check-backend termina
  • cargo xtask test-backend termina
  • cargo xtask rdme --check
  • markdownlint-cli2 ARCHITECTURE.md ratatui-termina/README.md .github/ISSUE_TEMPLATE/bug_report.md

  ---

Bug Fixes

• fce3c80 (widgets) Require thread-safe shadow effects by @joshka in #2584

  Summary

  • require custom shadow effects to preserve the auto traits expected by
    Block-backed widgets
  • document the CellEffect auto-trait contract
  • add a public widget regression test for the affected ratatui::widgets
    re-exports

  Fixes #2583

  ---

• e306ce6 (buffer) Create updates for "uncovered" cells by @benjajaja in #2587

  When a wide cell from the previous buffer is replaced by a short/normal
  cell, the trailing cell does not get an update if its content does not
  change. But if the wide cell has a background (or other) style, the
  terminal did render the trailing cell with that style.

  Force trailing cells to update if background, underline, or modifiers
  are different than the wide cell. We can ignore foreground.

  Fixes #2585 (see that for the detailed visual reports)

• 81e667f (scrollbar) Keep a large thumb within the track at the end by @satyakwok in #2594

  Closes #2582.

  Problem

  When the content is shorter than the viewport, the thumb is large
  relative to the track. With the position at the end, part_lengths
  clamped thumb_start to track_length - 1 while thumb_length was
  clamped independently to [1, track_length], so thumb_start + thumb_length could exceed track_length.

  bar_symbols lays out begin + track_start + thumb + track_end + end
  and zips it against the cells of the area. When the thumb overruns the
  track, track_end saturates to 0 but the thumb still emits more cells
  than the track can hold, so the trailing end symbol is pushed past the
  end of the area. The last visible cell ends up being a thumb (█) where
  the end arrow (▼) should be.

  Concretely, for the issue's repro (VerticalRight, content_length = 9, position = 8, height 24): track is 22, thumb_length = 17,
  thumb_start = 6, and 6 + 17 = 23 > 22.

  This is a regression from v0.30.0, where thumb_length was derived as
  thumb_end - thumb_start and therefore always fit within the track.

  Fix

  Clamp thumb_start to track_length - thumb_length (instead of
  track_length - 1) so the thumb always fits within the track and the
  end symbol is preserved.

  Test

  Two regression tests, both fail on main and pass with the fix:

  • thumb_stays_within_track_for_large_thumb_at_end checks
    part_lengths directly with the issue's parameters — asserts
    thumb_start + thumb_length <= track_length and that the parts sum to
    the track length.
  • render_scrollbar_keeps_end_symbol_for_large_thumb renders the #2582
    case (both arrows, large thumb at the end) and asserts the end symbol is
    drawn rather than overwritten by a thumb cell.

  All existing scrollbar tests still pass.

Miscellaneous Tasks

• c75d778 (ci) Add cargo-udeps dependency check by @joshka in #2599

  Adds cargo xtask udeps and runs it from CI as a required job.

  This complements cargo-machete rather than replacing it. cargo-machete
  is a fast static source scan, which is why it missed the package-level
  unused deps fixed in #2598 when the same dependency names were still
  referenced by example crates. cargo-udeps compiles the workspace and
  checks rustc dep-info, so it can catch unused dependency declarations
  for the package being checked.

  To make the new job pass, this also removes the remaining true-positive
  unused dev-deps and records explicit cargo-udeps ignores for current
  false positives / intentional cases: ratatui-core critical-section,
  ratatui-crossterm's duplicate crossterm version feature shape, and
  ratatui-termwiz's doc-example-only ratatui dev-dependency.

  I searched existing issues and PRs for udeps / cargo-udeps / "cargo
  udeps". I did not find prior ratatui discussion about adopting
  cargo-udeps; the only hits were Dependabot PR bodies for
  taiki-e/install-action release notes mentioning cargo-udeps version
  updates, for example #1971, #2095, #2194, and #2522.

  Validation:- cargo xtask udeps

  • cargo xtask format --check

  ---

• 4a63d41 (uncategorized) Remove unused dependencies by @KikiKian in #2598

  Audit removes these dependencies that are not used:

  ratatui/Cargo.toml — Removed from [dev-dependencies]:

  • futures
  • rand_chacha
  • tokio
  • tracing
  • tracing-appender
  • tracing-subscriber

  ratatui-core/Cargo.toml — Moved from [dependencies] →
  [dev-dependencies]:

  • indoc

  ---

Continuous Integration

• 36854ef (uncategorized) Add auto-merge required gate by @joshka in #2596

  Summary

  This makes GitHub auto-merge usable for Ratatui PRs once maintainers are
  happy with the change but CI is still running.

  The workflow change adds a single aggregate required job to the main
  CI workflow. The repository now has auto-merge
  enabled
  and an ensure checks pass
  ruleset
  that requires that required status context on main.

  Why

  Without a required status context, GitHub's auto-merge button is not
  useful for the maintainer flow we want. The goal is to let a maintainer
  review a PR, decide it is ready, click auto-merge, and move on without
  coming back later just to check whether the remaining jobs finished.

  This does not relax the merge policy. GitHub's own auto-merge behavior
  is to merge only after all required reviews and required status checks
  are satisfied. This change gives GitHub a stable required status to wait
  on automatically.

  Precedent

  I have been using this same auto-merge pattern in
  ratatui/tui-widgets, where
  it has worked well for the intended maintainer flow: once a PR looks
  ready, I can enable auto-merge and let GitHub merge it after the
  remaining checks and review requirements are satisfied.

  How it works

  The new required job depends on the main CI jobs in
  .github/workflows/ci.yml and always runs after them. It fails if any
  required dependency fails, is cancelled, or is skipped.

  The repository ruleset requires only this aggregate required context
  instead of requiring every individual matrix job separately. That gives
  GitHub one stable status to wait on while preserving the existing CI
  coverage.

  Things to know

  • Auto-merge is opt-in per PR. Maintainers still choose when to click
    it.
  • It does not skip review requirements, status checks, labels, or any
    other protection rule.
  • A PR with auto-merge enabled can still show as blocked while checks or
    required reviews are pending. That is expected.
  • If something needs to merge normally, maintainers can still use the
    regular merge path or an allowed ruleset bypass. This is a convenience
    path, not a hard blocker.
  • Existing open PRs may need a rebase or synchronize event after this
    lands so they pick up the new required workflow job.
  • If a new required CI job is added later, it should be added to the
    required.needs list or it will not be represented by the aggregate
    gate.
  • Jobs that are intentionally allowed to fail should be handled
    carefully before adding them to required.needs, because skipped,
    cancelled, and failed dependencies make the aggregate fail.

  Current PR state

  Auto-merge is already enabled on this PR. If you approve it and the
  required checks pass, GitHub will squash-merge it automatically;
  approving it is enough to let the PR merge once the remaining
  requirements are satisfied.

  GitHub docs

  • Automatically merging a pull
    request
  • Managing auto-merge for pull requests in your
    repository
  • About
    rulesets
  • Require status checks to pass before
    merging
  • About status
    checks
  • Troubleshooting required status
    checks

  Validation

  • ruby -e 'require "yaml"; YAML.load_file(".github/workflows/ci.yml"); puts "ok"'
  • actionlint .github/workflows/ci.yml
  • Verified ratatui/ratatui has allow_auto_merge: true
  • Verified the active ensure checks pass ruleset requires status
    context required
  • Verified this PR has squash auto-merge enabled and is blocked pending
    checks/review

New Contributors

• @satyakwok made their first contribution in #2594
• @KikiKian made their first contribution in #2598

Full Changelog: ratatui-v0.30.1...ratatui-v0.30.2

"Rats, we're rats; we're the rats." – Rat Movie

We are excited to announce the new version of ratatui - a Rust library that's all about cooking up TUIs 👨‍🍳🐀

✨ Release highlights: https://ratatui.rs/highlights/v0301/

⚠️ List of breaking changes can be found here.

Features

• 74d6a84 (block) Support shadows by orhun in #2481

  Introduce Block::shadow(...) with a new Shadow type that supports:

  • presets: overlay, block, light_shade, medium_shade,
    dark_shade
  • custom symbols via Shadow::symbol(...)
  • custom effects via Shadow::custom(...)

  use ratatui::layout::Offset;
  use ratatui::style::Stylize;
  use ratatui::widgets::{Block, Shadow};
  
  let popup = Block::bordered().title("Popup").shadow(
      Shadow::dark_shade()
          .black()
          .on_white()
          .offset(Offset::new(2, 1)),
  );

  Results in:

  ┌Popup─────┐
  │content   │▒
  └──────────┘▒
    ▒▒▒▒▒▒▒▒▒▒▒
  

  fixes #1892

  ---

• 4d30420 (buffer) Add CellDiffOption::AlwaysUpdate to force cell updates by sxyazi in #2480

  When this option is used, the cells are updated even when content is identical.

  Follow-up for #1605

  Trying to resolve #1116

• 39c32c6 (buffer) Add cell diff options by benjajaja in #1605

  Problem:Escape sequences always cause a cell to count as "multiwidth",
  even when it doesn't render wider than one cell, or not as wide as the
  escape sequence would be computed as.

  Solution:Convert skip:bool to enum. Add enum option ForceWidth to
  force a cell width for diffing.

  When using the option, this also fixes some bug where diffing is not
  idempotent and causes a diff operation for (symbol.len() - 1) times.

  There are three new specific test cases:

  1. Rendering hyperlinks by squeezing the escape sequence into the first
     cell and forcing the width to the unicode width of the text part.
     This is much easier to implement for a Link widget, as it would only
     need to get the unicode-width once and not iterate over graphemes
     like Spans must do.
  2. Rendering hyperlinks by squeezing the opening sequence into the first
     cell with the first grapheme and forcing the width to that of the
     first grapheme. Then rendering each grapheme as usual. Then squeezing
     the closing sequence into the last cell with the last grapheme and
     forcing the width to that of the last grapheme.
     This is harder to implement for a Link widget, as it would have to
     iterate over graphemes with their width like Spans do.
  3. Kitty image sequence with utf-8 placeholders, similar to 2 but with
     known constant grapheme widths.

  Link widget that leverages this

  https://github.com/benjajaja/tui-link

  It would be cooler if we could just add something like .link(url) to
  Spans, because it would much simpler to insert some link and leverage
  all the Line/Text/Paragraph wrapping and whatnot. With a custom widget
  you need to take care of the Area where you'd want to render it, so
  it's not that clean. But we could iterate on this later, if even
  possible.

• 6faaddb (core) Implement from slice for line and text by NoOPeEKS in #2371

  This PR adds the following implementations of the From trait for Line
  and Text structs:

  • Implements From<&[T]> where T is Into<Span> for Line, allowing using
    of slices to construct Lines.
  • Implements From<&[T]> where T is Into<Line> for Text, allowing using
    of slices of various types to construct Texts.

  closes #2279

• 5fc6ab8 (core) Support layout-cache in no_std environments by junkdog in #2399

  this enables "layout-cache" for no_std builds; it's meant for embedded
  environments, where the layout engine otherwise consumes all CPU,
  capping the framerate at around ~10fps. the same app can refresh 300-500
  times per second with layout cache enabled.

  i had to add layout-cache = ["dep:critical-section"] to all builds -
  it's pretty tiny and shouldn't leave a trace in std-builds. the
  alternative is to add an extra layer of features for layout-cache with
  std and no_std, but it pollutes the feature space.

  ---

• ee4b7a9 (crossterm) Add the missing hidden modifier by sxyazi in #2413

  Fixes sxyazi/yazi#3724, see
  sxyazi/yazi#3724 (comment) for a
  reproducer.

  This PR adds the missing Modifier::HIDDEN style and introduces a
  queue_modifier_diff to test ModifierDiff::queue().

  It also fixes a bug where CrosstermAttribute::Bold and
  CrosstermAttribute::Dim would be emitted twice when resetting
  intensity. For example:

  #[case(Modifier::DIM, Modifier::BOLD, &[CrosstermAttribute::NormalIntensity, CrosstermAttribute::Bold])]

  would become:

  #[case(Modifier::DIM, Modifier::BOLD, &[CrosstermAttribute::NormalIntensity, CrosstermAttribute::Bold, CrosstermAttribute::Bold])]

  ---

• 9d9239a (examples) Add volatility-surface 3D visualization by floor-licker in #2322

  A design demonstration of a 3D volatility surface rendering using
  Braille canvas with interactive rotation and zoom controls. I built this
  for myself for an app I'm currently building but just wanted to share it
  with the community as well to inspire more 3D perspective terminal
  widgets in the future.

  Final Demo

  Description

  Adds a new example demonstrating 3D visualization techniques in the
  terminal. There aren't many examples showing how to represent 3D objects
  in 2D terminal space so my goal is just to demonstrate more advanced
  Canvas and Braille rendering techniques for 3D graphics. The example
  visualizes an implied volatility surface which is a common financial
  visualization using interactive rotation and zoom controls. You can run
  the interactive demo for yourself with cargo run -p volatility-surface

  demo.mov

  <img width="659" height="432" alt="image"

  src="https://github.com/user-attachments/assets/68698cb0-c5d5-4b41-a3c3-65ec8fff12f5"
  />

  Technical Highlights

  • Demonstrates how to implement perspective projection in a terminal
  • Shows advanced use of Canvas widget with Marker::Braille
  • Example of smooth animation patterns and state management
  • Self-contained with synthetic data generation (no external APIs)

  ---

• ae975c7 (examples) Allow overlap spacing in explorer by joshka in #2316

  Store spacing as i16 so negative values map to Spacing::Overlap, and
  show overlap in the axis label.

• 1e0ab0c (ratatui-crossterm) Add IntoCrossterm for Style by 0xferrous in #2323

• 101a63e (render) Add function for applying buffer by musjj in #2566

  Add a public API for applying and flushing the terminal buffer.

  A minimal usage will look something like this:

  use ratatui::Terminal;
  use ratatui::backend::CrosstermBackend;
  use ratatui::buffer::Buffer;
  use ratatui::widgets::Widget;
  
  let backend = CrosstermBackend::new(io::stdout());
  let mut terminal = Terminal::new(backend)?;
  
  terminal.autoresize()?;
  
  let mut custom_buffer = Buffer::default();
  custom_buffer.resize(terminal.get_frame().area());
  custom_buffer.reset();
  
  "Hello World!".render(custom_buffer.area, &mut custom_buffer);
  
  terminal.current_buffer_mut().merge(&custom_buffer);
  terminal.apply_buffer()?;

  My primary motivation for this PR is to improve the ECS ergonomics in
  bevy_ratatui. But this
  should be useful for anyone who wants to commit incremental writes to
  the buffer without having to do everything in one monolithic
  Terminal::draw
  closure.

  ---

• 09a3027 (symbol) Add custom marker by BenFradet in #2356

• f9d066f (table) Let Cells span multiple columns by karkhaz in #2150

  Add a 'column_span' field to table cells. The default value
  is 1; larger values will cause cells to span over multiple columns,
  being rendered over all columns plus the spaces between them.

  Fixes #1568.

• a5b08d6 (widgets) Add Fill widget by Metbcy in #2520

  Adds a new Fill widget that paints every cell within its area with a
  single repeated symbol and style. Integrates with Stylize so the whole
  chain works as expected:

  use ratatui::widgets::{Fill, Widget};
  use ratatui::style::Stylize;
  
  Fill::new("X").blue().bold().render(area, buf);

  Implements Widget (for both Fill and &Fill) and Styled, accepts
  anything that converts into a Cow<'a, str>, and degrades gracefully on
  empty / multi-grapheme symbols.

  ---

• 9094fd2 (widgets) Add line shape with filled area for Canvas and Chart by bananaofhappiness in #2426

  This commit adds filled area-chart rendering for both Canvas and Chart.

  You can now render a line and fill the area between that line and a baseline Y value, which helps highlight magnitude/volume trends.

  • In Canvas, use FilledLine.
  • In Chart, use GraphType::Area and set the baseline via Dataset::fill_to_y(f64).

  The f64 baseline is now configured on Dataset (not in GraphType), so GraphType stays a simple enum variant.

  Under the hood, line rasterization in canvas was refactored to share a reusable Bresenham point iterator, and FilledLine builds on that to paint vertical spans from each line point to fill_to_y.

  Some screenshots with and without this new type:
  
  

  // In Canvas
  use ratatui::widgets::canvas::FilledLine;
  
  Canvas::default()
      .paint(|ctx| {
          ctx.draw(&FilledLine::new(0.0, 0.0, 10.0, 5.0, 0.0, Color::Red));
      });
  
  // In Chart
  let dataset = Dataset::default()
      .data(&data)
      .graph_type(GraphType::Area)
      .fill_to_y(0.0); // fill to y = 0
  
  Chart::new(vec![dataset]);

  ---

• 0a87882 (uncategorized) Add impl From<u16> for Padding and Margin by JayanAXHF in #2438

• 556cc7b (uncategorized) Add comment for inner area to popup example by Its-Just-Nans in #2309

• 01a15f9 (uncategorized) Add AsRef impls for widget types by joshka in #2297

Bug Fixes

• d12bb83 (barchart) Handle empty horizontal charts by fallintoplace in #2553

  Fixes #2552

  This makes the BarChart constructors ignore empty groups, matching the
  existing .data(...) builder behavior. Without this,

  BarChart::horizontal(Vec::<Bar>::new()) stores one empty group,
  proceeds into horizontal rendering, skips the bar loop, and then
  underflows when computing the group label row from bar_y - self.bar_gap.

  The fix normalizes constructor input through a shared non_empty_groups
  helper for new, horizontal, and grouped. Empty horizontal charts
  now render nothing instead of panicking, and constructor behavior is
  consistent with .data(...).

  ---

• 6396b1c (block) Saturate block edge arithmetic by joshka in #2488

  Motivation

  • Block border and title layout used unchecked u16 arithmetic in
    several places.
  • In debug builds that can panic on tiny or edge-case geometry; in
    release builds the same arithmetic wraps.
  • The original report came from merge-border rendering on tiny areas,
    but the same pattern appeared in title layout and spacing helpers as
    well.

  Description

  • Use saturating arithmetic in Block::inner, render_sides,
    render_corners, titles_area, and vertical_space.
  • Clamp rendered title widths to u16 for layout arithmetic.
  • Replace unchecked title-width accumulation and cursor-advance math
    with bounded arithmetic.
  • Add debug-only regression tests covering empty areas, maximal padding,
    title-area edge cases, and very large title widths.

  Testing

  • Ran cargo test -p ratatui-widgets block::tests.

  ---

  Codex
  Task

• 9143b83 (buffer) Diff for trailing cells when only style changes by gcavelier in #2308

  this PR closes #2307 by preventing unnecessary diff updates for trailing
  cells when only style changes.

  This PR was generated by Claude, and validated by me.

  Summary

  This PR fixes a visual artifact bug where block borders would appear
  offset when rendered over a widget that had a foreground color style
  applied to the entire area.

  The Fix

  - if !next_trailing.skip && prev_trailing != next_trailing {
  + // Only emit update if the SYMBOL changed, not just the style.
  + // The style of hidden trailing cells is not visible, so style
  + // differences alone should not trigger updates that can cause
  + // cursor positioning issues on some terminals.
  + if !next_trailing.skip && prev_trailing.symbol() != next_trailing.symbol() {

  This aligns the code with the documented intent: only emit updates when
  the symbol (visible content) changes, not when only the style
  changes.

  Changes

  File	Change
  ratatui-core/src/buffer/buffer.rs:526-530	Compare only symbol, not
  full cell
  ratatui-core/src/buffer/buffer.rs:1376-1425	Add regression test

  Test Added

  #[test]
  fn diff_ignores_style_only_changes_in_trailing_cells() {
      // Verifies that trailing cells with same symbol but different style
      // do NOT generate diff updates
  }

  Why This Is Safe

  1. Trailing cells are hidden - they are visually covered by the wide
     character
  2. Style is invisible - the fg/bg color of a hidden cell has no
     visual effect
  3. Symbol changes still trigger updates - if the symbol changes
     (e.g., from " " to "x"), the update is still emitted
  4. Aligns with documented intent - the original comment says
     "non-blank content", not "different style"

  Related

  • The existing test diff_clears_trailing_cell_for_wide_grapheme
    verifies that symbol changes DO trigger updates
  • This fix complements that behavior by ensuring style-only changes do
    NOT trigger updates

• e6b71f2 (build) Correct rust-toolchain->rust-version on cargo-deny-action by sermuns in #2471

  closes #2470

• 65c5202 (changelog) Fix typo by joshka in #2300

• 43bbaae (clippy) Fix beta clippy errors by Logan-Ruf in #2433

  Noticed these errors on my other PR and figured I could just fix them
  real quick.

  closes #2432

• 957fbb0 (core) Use correct width for halfwidth dakuten/handakuten by orhun in #2499

  unicode-width reports U+FF9E/U+FF9F as zero-width, but terminals render
  them as 1 cell.
  Adjusts CellWidth trait accordingly for fixing this behavior.

  fixes #2188

• d7646c7 (core) Avoid overflow in BufferDiff forced-width advance by joshka in #2487

  Motivation

  • Prevent arithmetic overflow when advancing self.pos for
    CellDiffOption::ForcedWidth(NonZeroU16) in
    ratatui-core/src/buffer/diff.rs, which could panic in debug or wrap in
    release and cause an iterator hang/DoS.

  Description

  • Replace the unchecked self.pos += width.get().saturating_sub(1) with
    a saturating addition via self.pos = self.pos.saturating_add(width.get().saturating_sub(1) as usize) to
    avoid overflow while preserving existing iterator semantics.

  Testing

  • Ran cargo test -p ratatui-core buffer::diff --lib and the buffer
    diff tests completed successfully (10 passed, 0 failed).

  ---

  Codex
  Task

• 77f8006 (core) Avoid cursor position queries during resize by orhun in #2485

  Terminal::resize() now clears without calling get_cursor_position(),
  so that CPR (Cursor position report) calls does not interfere with
  stdin.

  Fixes #2483

  ---

• 18aa467 (examples) Make line-gauge example compatible with macos sequoia's terminal.app by lazo4 in #2474

  Part of the fix for #1972

  Summary

  This fix makes the line-gauge example compatible with the macos
  sequoia Terminal.app which doesn't support truecolor. It reuses the
  is_true_color_supported introduced in #2211 by ffex and uses a color
  theme instead of hardcoding the colors.

  Result on macos sequoia

  Before:
  
  After:
  

  Notes

  This is my first open source contribution, thanks to @ffex for letting
  me help on this issue

• ce2c228 (examples) Change flex example colors for MacOS default terminal by ffex in #2211

  Part of the fix for #1972

  Summary

  This fix introduces a function to check if we are in a terminal without
  truecolor(24-bit) and changes the default colors to appear fine of the
  flex example.

  Notes

  The function "is_true_color_supported” is an old problem and there is no
  common way to determine if a terminal supports or not the truecolor.

  This is the main reason why the function detects specifically the
  Terminal.app version before the Tahoe. If there are other known
  terminals with this problem, we can add it to this function.

  ---

• ef72dba (examples) Fix import for widget examples by orhun in #2422

  closes #2299

• 88441cf (terminal) Fix inline viewport resizing issues by clearing the screen by wyvernbw in #2355

  adds a check to the autoresize function to clear the entire screen and
  move the inline viewport to the top when the window shrinks horizontally
  in order to avoid line wrapping issues.

  Other libraries like ink purge the history as well, but the
  backend::ClearType type does not support that. Without this if the
  user scrolls up they will see previous broken renders. This should work
  well with all terminal emulators and multiplexers.

  fixes #2086

• 91b6fb7 (tests) Use the correct type for the cell diff test by orhun in #2472

  fixes the CI!

• 4493742 (widgets) Handle single y-axis label by fallintoplace in #2550

  Fixes #2549.

  This prevents Chart from panicking when the Y axis is configured with
  exactly one label. The X-axis rendering path already skips label
  placement when fewer than two labels are provided; this applies the same
  guard to Y-axis labels before the spacing calculation divides by
  labels_len - 1.

  This also updates the Axis::labels docs so they describe the new

  behavior:fewer than two labels are not rendered instead of causing a
  panic.

• 0bdebd6 (widgets) Prevent chart scaling overflow by fallintoplace in #2546

  Summary

  Fixes #2545.

  This changes BarChart and Sparkline scaling to use a u128 intermediate
  before division, then caps the scaled ticks at the drawable area. That
  prevents debug-build panics and release-build wrapping when public u64
  chart values are large.

  Validation

  • cargo test -p ratatui-widgets barchart::tests
  • cargo test -p ratatui-widgets sparkline::tests
  • cargo check -p ratatui-widgets --all-features
  • cargo clippy -p ratatui-widgets --all-targets --all-features -- -D warnings
  • cargo test -p ratatui-widgets

• e27a22a (widgets) Inherit the text alignment for Paragraph by 7Bpencil in #2369

  Paragraph didn't take into account alignment of the text it was created
  from:

  let lines = vec![
      Line::from("one"),
      Line::from("double"),
      Line::from("quadruple"),
  ];
  let text = Text::from(lines).centered();
  
  // used to be rendered left-aligned, now centered
  let paragraph = Paragraph::new(text).block(block);

  Now the Paragraph inherits the text alignment.

• b5c0831 (widgets) Avoid panic if Clear area is outside of buffer by 7Bpencil in #2368

  If Clear area is at least partially outside of buffer, panic "index
  outside of buffer" happens on Widget::render

  Demo source code

  use crossterm::event::{self, Event, KeyModifiers};
  use ratatui::{
      layout::Rect,
      text::Line,
      widgets::{Block, Borders, Clear, Paragraph},
      DefaultTerminal, Frame,
  };
  use std::iter;
  
  fn main() {
      ratatui::run(app);
  }
  
  fn app(terminal: &mut DefaultTerminal) {
      loop {
          if let Event::Key(key_event) = event::read().expect("failed to read event") {
              if key_event.kind.is_press()
                  && key_event.modifiers.contains(KeyModifiers::CONTROL)
                  && key_event.code.is_char('c')
              {
                  break;
              }
          }
          terminal.draw(render).expect("failed to draw frame");
      }
  }
  
  fn render(frame: &mut Frame) {
      {
          let width = frame.area().width;
          let area = Rect::new(0, 0, width, 10);
          let line = Line::from("W".repeat(area.width as usize));
          let lines: Vec<Line> = iter::repeat_n(line, area.height as usize).collect();
          frame.render_widget(Paragraph::new(lines), area);
      }
      {
          let area = Rect::new(50, 2, 20, 5);
          let block = Block::default()
              .title_top(Line::from("Popup-with-Clear").centered())
              .borders(Borders::ALL);
          let lines = vec![
              Line::from("one"),
              Line::from

Note truncated.

Features

• Add experimental test backend

Breaking Changes

• Replace Item by a generic and flexible Text that can be used in both Paragraph and List widgets.
• Remove unnecessary borrows on Style.

Features

• Add a basic Crossterm backend

Breaking Changes

• Remove Group and introduce Layout in its place
  • Terminal is no longer required to compute a layout
  • Size has been renamed Constraint
• Widgets are rendered on a Frame instead of a Terminal in order to avoid mixing draw and render calls
• draw on Terminal expects a closure where the UI is built by rendering widgets on the given Frame
• Update Widget trait
  • draw takes area by value
  • render takes a Frame instead of a Terminal
• All widgets use the consumable builder pattern
• SelectableList can have no selected item and the highlight symbol is hidden in this case
• Remove markup language inside Paragraph. Paragraph now expects an iterator of Text items

Features

• Terminal implements Debug

Breaking Changes

• Use FnOnce instead of FnMut in Group::render

Features

• Add AlternateScreenBackend in termion backend
• Add TermionBackend::with_stdout in order to let an user of the library provides its own termion struct
• Add tests and documentation for Buffer::pos_of
• Remove leading whitespaces when wrapping text

Bug Fixes

• Fix debug_assert in Buffer::pos_of
• Pass the style of SelectableList to the underlying List
• Fix missing character when wrapping text
• Fix panic when specifying layout constraints

Features

• Add MouseBackend in termion backend to handle scroll and mouse events
• Add generic Item for items in a List
• Drop log4rs as a dev-dependencies in favor of stderrlog

Breaking Changes

• Rename TermionBackend to RawBackend (to distinguish it from the MouseBackend)
• Generic parameters for List to allow passing iterators as items
• Generic parameters for Table to allow using iterators as rows and header
• Generic parameters for Tabs
• Rename border bitflags to Borders