Code which previously called `buf.get(x, y)` or `buf.get_mut(x, y)`
should now use index operators, or be transitioned to `buff.cell()` or
`buf.cell_mut()` for safe access that avoids panics by returning
`Option<&Cell>` and `Option<&mut Cell>`.
The new methods accept `Into<Position>` instead of `x` and `y`
coordinates, which makes them more ergonomic to use.
```rust
let mut buffer = Buffer::empty(Rect::new(0, 0, 10, 10));
let cell = buf[(0, 0)];
let cell = buf[Position::new(0, 0)];
let symbol = buf.cell((0, 0)).map(|cell| cell.symbol());
let symbol = buf.cell(Position::new(0, 0)).map(|cell| cell.symbol());
buf[(0, 0)].set_symbol("🐀");
buf[Position::new(0, 0)].set_symbol("🐀");
buf.cell_mut((0, 0)).map(|cell| cell.set_symbol("🐀"));
buf.cell_mut(Position::new(0, 0)).map(|cell| cell.set_symbol("🐀"));
```
The existing `get()` and `get_mut()` methods are marked as deprecated.
These are fairly widely used and we will leave these methods around on
the buffer for a longer time than our normal deprecation approach (2
major release)
Addresses part of: https://github.com/ratatui-org/ratatui/issues/1011
---------
Co-authored-by: EdJoPaTo <rfc-conform-git-commit-email@funny-long-domain-label-everyone-hates-as-it-is-too-long.edjopato.de>
- Simplify `assert_buffer_eq!` logic.
- Deprecate `assert_buffer_eq!`.
- Introduce `TestBackend::assert_buffer_lines`.
Also simplify many tests involving buffer comparisons.
For the deprecation, just use `assert_eq` instead of `assert_buffer_eq`:
```diff
-assert_buffer_eq!(actual, expected);
+assert_eq!(actual, expected);
```
---
I noticed `assert_buffer_eq!` creating no test coverage reports and
looked into this macro. First I simplified it. Then I noticed a bunch of
`assert_eq!(buffer, …)` and other indirect usages of this macro (like
`TestBackend::assert_buffer`).
The good thing here is that it's mainly used in tests so not many
changes to the library code.
`Block::bordered()` is shorter than
`Block::new().borders(Borders::ALL)`, requires one less import
(`Borders`) and in case `Block::default()` was used before can even be
`const`.
Fixes many not yet enabled lints (mostly pedantic) on everything that is
not the lib (examples, benchs, tests). Therefore, this is not containing
anything that can be a breaking change.
Lints are not enabled as that should be the job of #974. I created this
as a separate PR as its mostly independent and would only clutter up the
diff of #974 even more.
Also see
https://github.com/ratatui-org/ratatui/pull/974#discussion_r1506458743
---------
Co-authored-by: Josh McKinney <joshka@users.noreply.github.com>
This PR makes a number of simplifications to the layout and constraint
features that were added after v0.25.0.
For users upgrading from v0.25.0, the net effect of this PR (along with
the other PRs) is the following:
- New `Flex` modes have been added.
- `Flex::Start` (new default)
- `Flex::Center`
- `Flex::End`
- `Flex::SpaceAround`
- `Flex::SpaceBetween`
- `Flex::Legacy` (old default)
- `Min(v)` grows to allocate excess space in all `Flex` modes instead of
shrinking (except in `Flex::Legacy` where it retains old behavior).
- `Fill(1)` grows to allocate excess space, growing equally with
`Min(v)`.
---
The following contains a summary of the changes in this PR and the
motivation behind them.
**`Flex`**
- Removes `Flex::Stretch`
- Renames `Flex::StretchLast` to `Flex::Legacy`
**`Constraint`**
- Removes `Fixed`
- Makes `Min(v)` grow as much as possible everywhere (except
`Flex::Legacy` where it retains the old behavior)
- Makes `Min(v)` grow equally as `Fill(1)` while respecting `Min` lower
bounds. When `Fill` and `Min` are used together, they both fill excess
space equally.
Allowing `Min(v)` to grow still allows users to build the same layouts
as before with `Flex::Start` with no breaking changes to the behavior.
This PR also removes the unstable feature `SegmentSize`.
This is a breaking change to the behavior of constraints. If users want
old behavior, they can use `Flex::Legacy`.
```rust
Layout::vertical([Length(25), Length(25)]).flex(Flex::Legacy)
```
Users that have constraint that exceed the available space will probably
not see any difference or see an improvement in their layouts. Any
layout with `Min` will be identical in `Flex::Start` and `Flex::Legacy`
so any layout with `Min` will not be breaking.
Previously, `Table` used `EvenDistribution` internally by default, but
with that gone the default is now `Flex::Start`. This changes the
behavior of `Table` (for the better in most cases). The only way for
users to get exactly the same as the old behavior is to change their
constraints. I imagine most users will be happier out of the box with
the new Table default.
Resolves https://github.com/ratatui-org/ratatui/issues/843
Thanks to @joshka for the direction
Follow up to https://github.com/ratatui-org/ratatui/pull/783
This PR introduces different priorities for each kind of constraint.
This PR also adds tests that specifies this behavior. This PR resolves a
number of broken tests.
Fixes https://github.com/ratatui-org/ratatui/issues/827
With this PR, the layout algorithm will do the following in order:
1. Ensure that all the segments are within the user provided area and
ensure that all segments and spacers are aligned next to each other
2. if a user provides a `layout.spacing`, it will enforce it.
3. ensure proportional elements are all proportional to each other
4. if a user provides a `Fixed(v)` constraint, it will enforce it.
5. `Min` / `Max` binding inequality constraints
6. `Length`
7. `Percentage`
8. `Ratio`
9. collapse `Min` or collapse `Max`
10. grow `Proportional` as much as possible
11. grow spacers as much as possible
This PR also returns the spacer areas as `Rects` to the user. Users can
then draw into the spacers as they see fit (thanks @joshka for the
idea). Here's a screenshot with the modified flex example:
<img width="569" alt="image"
src="https://github.com/ratatui-org/ratatui/assets/1813121/46c8901d-882c-43b0-ba87-b1d455099d8f">
This PR introduces a `strengths` module that has "default" weights that
give stable solutions as well as predictable behavior.
This uses the new `spacing` feature of the `Layout` struct to allocate
columns spacing in the `Table` widget.
This changes the behavior of the table column layout in the following
ways:
1. Selection width is always allocated.
- if a user does not want a selection width ever they should use
`HighlightSpacing::Never`
2. Column spacing is prioritized over other constraints
- if a user does not want column spacing, they should use
`Table::new(...).column_spacing(0)`
---------
Co-authored-by: Josh McKinney <joshka@users.noreply.github.com>
Any iterator whose item is convertible into `Row` can now be
collected into a `Table`.
Where previously, `Table::new` accepted `IntoIterator<Item = Row>`, it
now accepts `IntoIterator<Item: Into<Row>>`.
BREAKING CHANGE:
The compiler can no longer infer the element type of the container
passed to `Table::new()`. For example, `Table::new(vec![], widths)`
will no longer compile, as the type of `vec![]` can no longer be
inferred.
This prevents creating a table that doesn't actually render anything.
Fixes: https://github.com/ratatui-org/ratatui/issues/537
BREAKING CHANGE: Table::new() now takes an additional widths parameter.
This allows passing an array, slice or Vec of constraints, which is more
ergonomic than requiring this to always be a slice.
The following calls now all succeed:
```rust
Table::new(rows).widths([Constraint::Length(5), Constraint::Length(5)]);
Table::new(rows).widths(&[Constraint::Length(5), Constraint::Length(5)]);
// widths could also be computed at runtime
let widths = vec![Constraint::Length(5), Constraint::Length(5)];
Table::new(rows).widths(widths.clone());
Table::new(rows).widths(&widths);
```
Previously the layout used the floor of the calculated start and width
as the value to use for the split Rects. This resulted in gaps between
the split rects.
This change modifies the layout to round to the nearest column instead
of taking the floor of the start and width. This results in the start
and end of each rect being rounded the same way and being strictly
adjacent without gaps.
Because there is a required constraint that ensures that the last end is
equal to the area end, there is no longer the need to fixup the last
item width when the fill (as e.g. width = x.99 now rounds to x+1 not x).
The colors example has been updated to use Ratio(1, 8) instead of
Percentage(13), as this now renders without gaps for all possible sizes,
whereas previously it would have left odd gaps between columns.
* feat(table): add option to configure selection layout changes
Before this option was available, selecting a row in the table when no row was selected
previously made the tables layout change (the same applies to unselecting) by adding the width
of the "highlight symbol" in the front of the first column, this option allows to configure this
behavior.
* refactor(table): refactor "get_columns_widths" to return (x, width)
and "render" to make use of that
* refactor(table): refactor "get_columns_widths" to take in a selection_width instead of a boolean
also refactor "render" to make use of this change
* fix(table): rename "highlight_set_selection_space" to "highlight_spacing"
* style(table): apply doc-comment suggestions from code review
Co-authored-by: Dheepak Krishnamurthy <me@kdheepak.com>
---------
Co-authored-by: Dheepak Krishnamurthy <me@kdheepak.com>
* refactor: add Line type to replace Spans
`Line` is a significantly better name over `Spans` as the plural causes
confusion and the type really is a representation of a line of text made
up of spans.
This is a backwards compatible version of the approach from
https://github.com/tui-rs-revival/ratatui/pull/175
There is a significant amount of code that uses the Spans type and
methods, so instead of just renaming it, we add a new type and replace
parameters that accepts a `Spans` with a parameter that accepts
`Into<Line>`.
Note that the examples have been intentionally left using `Spans` in
this commit to demonstrate the compiler warnings that will be emitted in
existing code.
Implementation notes:
- moves the Spans code to text::spans and publicly reexports on the text
module. This makes the test in that module only relevant to the Spans
type.
- adds a line module with a copy of the code and tests from Spans with a
single addition: `impl<'a> From<Spans<'a>> for Line<'a>`
- adds tests for `Spans` (created and checked before refactoring)
- adds the same tests for `Line`
- updates all widget methods that accept and store Spans to instead
store `Line` and accept `Into<Line>`
* refactor: move text::Masked to text::masked::Masked
Re-exports the Masked type at text::Masked
* refactor: replace Spans with Line in tests/examples/docs
- control over the style of each cell and its content using the styling capabilities of Text.
- rows with multiple lines.
- fix panics on small areas.
- less generic type parameters.
* Fix percentage and ratio constraints for table to take borders into account
Percentage and ratio constraints don't take borders into account, which causes
the table to not be correctly aligned. This is easily seen when using 50/50
percentage split with bordered table. However fixing this causes the last column
of table to not get printed, so there is probably another problem with columns
width determination.
* Fix rounding of cassowary solved widths to eliminate imprecisions
* Fix formatting to fit convention
Co-authored-by: Kemyt <kemyt4@gmail.com>
