* Added ability to set title alignment, added tests, modified blocks example to show the feature
* Added test for inner with title in block
* Updated block example to show center alignment
* Formatting fixed
* Updated tests to use lamdas and be more concise. Updated title alignmnet code to be more straightforward and have correct behavior when placing title in the center without left border
- 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.
* gauge now uses unicode blocks for more descriptive progress
* removed unnecessary if
* changed function name to better reflect unicode
* standardized block symbols, added no unicode option, added tests
* formatting
* improved readability
* gauge tests now check color
* formatted
* 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>
Before this change, the style of the points drawn in the graph area could reused to draw the
title of the axis and the legend. Now the style of these components put on top of the graph area
is solely based on the widget style.
* Fix barchart incorrectly showing chart filled more than actual
Determination of how filled the bar should be was incorrectly taking the
entire chart height into account, when in fact it should take height-1, because
of extra line with label. Because of that the chart appeared fuller than it
should and was full before reaching maximum value.
* Add a test case for checking if barchart fills correctly up to max value
Co-authored-by: Kemyt <kemyt4@gmail.com>
This allows for clearer colors than using Dot, especially when
decreasing the size of the terminal font in order to increase the
resolution of the canvas
* Support more style modifiers on Windows
* Change Crossterm backend to write directly to buffer instead of String
Crossterm might actually do WinAPI calls instead of writing ANSI excape
codes so writing to an intermediate String may cause issues on older
versions of Windows. It also fails to compile with Crossterm 0.17.8 due
to Crossterm now expecting the writer to support `flush`, which String
doesn't.
Fixes#373
Both termion and crossterm backends were not moving the cursor if the first diff to draw was on the
second cell. The condition triggering the cursor move has been updated to fix this. In addition, two
tests have been added to avoid future regressions.
There was now way to avoid the autoresize behavior of `Terminal`. While it was fine for most users,
it made the testing experience painful as it was impossible to avoid the calls to `Backend::size()`.
Indeed they trigger the following error: "Inappropriate ioctl for device" since we are not running
the tests in a real terminal (at least in the CI).
This commit introduces a new api to create a `Terminal` with a fixed viewport.
- merge `Style` and `StyleDiff` together. `Style` now is used to activate or deactivate certain
style rules not to overidden all of them.
- update all impacted widgets, examples and tests.
* `Paragraph:scroll` takes a tuple of offsets instead of a single vertical offset.
* `LineTruncator` takes this new horizontal offset into account to let the paragraph scroll horizontally.
* Remove custom Debug implementation of Buffer
* Add `TestBackend::assert_buffer` to compare buffers in integration tests. When
the assertion fails, the output now show the list of differences in addition
of the views of the computed and expected buffers. This effectively replaces
the table of debug code for colors and modifiers as it is easier to read.
- Remove deny warnings in lib.rs. This allows easier iteration when developing
new features. The warnings will make the CI fails anyway on the clippy CI
stage.
- Run clippy on all targets (including tests and examples) and all features.
- Fail CI on clippy warnings.
The code that outputs the list elements uses the full inner width of its
block, without taking the width of the highlight symbol into
consideration. This allows the elements to overflow the box and draw
over the block's border. To fix that, we need to reduce the target width
for the list elements.
Add `BarChart::bar_set` and `Sparkline::bar_set` methods to customize
the set of symbols used to display the data. The new set should give
a better looking output on terminal that do not support a wide range
of unicode symbols.
* Update the `Shape` trait. Instead of returning an iterator of point, all
shapes are now aware of the surface they will be drawn to through a `Painter`.
In order to draw themselves, they paint points of the "braille grid".
* Rewrite how lines are drawn using a common line drawing algorithm (Bresenham).
* Clean redundant generics params in Table
It is possible to use associated types the same way as generics parameters. In fact associated types are nothing more than better organized generics params. For example, there is no need to introduce another param to constraint iterator item to be Display, you can just say `where I::Iterator, I::Item: Display`. This allows to drop type params for Table from 5 to 2.
Most widgets can be drawn directly based on the input parameters. However, some
features may require some kind of associated state to be implemented.
For example, the `List` widget can highlight the item currently selected. This
can be translated in an offset, which is the number of elements to skip in
order to have the selected item within the viewport currently allocated to this
widget. The widget can therefore only provide the following behavior: whenever
the selected item is out of the viewport scroll to a predefined position (make
the selected item the last viewable item or the one in the middle).
Nonetheless, if the widget has access to the last computed offset then it can
implement a natural scrolling experience where the last offset is reused until
the selected item is out of the viewport.
To allow such behavior within the widgets, this commit introduces the following
changes:
- Add a `StatefulWidget` trait with an associated `State` type. Widgets that
can take advantage of having a "memory" between two draw calls needs to
implement this trait.
- Add a `render_stateful_widget` method on `Frame` where the associated
state is given as a parameter.
The chosen approach is thus to let the developers manage their widgets' states
themselves as they are already responsible for the lifecycle of the wigets
(given that the crate exposes an immediate mode api).
The following changes were also introduced:
- `Widget::render` has been deleted. Developers should use `Frame::render_widget`
instead.
- `Widget::background` has been deleted. Developers should use `Buffer::set_background`
instead.
- `SelectableList` has been deleted. Developers can directly use `List` where
`SelectableList` features have been back-ported.
* removed flush calls because execute already calls flush under the hood.
* moved some static functions into From traits
* removed useless clone in demo
* upgrade to crossterm 0.13
* map all errors
This allows table column widths to be adapted more and scale with the
UI.
The constraints are solved using the Cassowary solver. An added
constraint for fitting them all in the width is added.
* Remove compilation warnings
* Fix rendering artifacts in the crossterm demo. In particular, the bold modifier
was leaking on most of the terminal screen because the old logic was not
properly unsetting the bold modifier after use (took inspiration of the termion
backend implementation)
Coordinates returned by Buffer::pos_of were interpreted as local coordinates
while they were global. This was resulting in panics due to out of bounds
accesses. Interpreting the coordinates as global and using correct offsets
when computing the new index within the buffer for each cell fix the issue.
It basically never makes sense to render without syncing the size.
Without resizing, if shrinking, we get artefacts. If growing, we may get
panics (before this change the Rustbox sample (the only one which didn't
handle resizing on its own) panicked because the widget would get an
updated size, while the terminal would not).
