Textual uses CSS to apply style to widgets. If you have any exposure to web development you will have encountered CSS, but don't worry if you haven't: this chapter will get you up to speed.
!!! tip "VSCode User?"
The official [Textual CSS](https://marketplace.visualstudio.com/items?itemName=Textualize.textual-syntax-highlighter) extension adds syntax highlighting for both external files and inline CSS.
CSS stands for Cascading Stylesheet. A stylesheet is a list of styles and rules about how those styles should be applied to a web page. In the case of Textual, the stylesheet applies styles to widgets, but otherwise it is the same idea.
Let's look at some Textual CSS.
Header {
dock: top;
height: 3;
content-align: center middle;
background: blue;
color: white;
}
This is an example of a CSS rule set. There may be many such sections in any given CSS file.
Let's break this CSS code down a bit.
Header {
dock: top;
height: 3;
content-align: center middle;
background: blue;
color: white;
}
The first line is a selector which tells Textual which widget(s) to modify. In the above example, the styles will be applied to a widget defined by the Python class Header
.
Header {
dock: top;
height: 3;
content-align: center middle;
background: blue;
color: white;
}
The lines inside the curly braces contains CSS rules, which consist of a rule name and rule value separated by a colon and ending in a semicolon. Such rules are typically written one per line, but you could add additional rules as long as they are separated by semicolons.
The first rule in the above example reads "dock: top;"
. The rule name is dock
which tells Textual to place the widget on an edge of the screen. The text after the colon is top
which tells Textual to dock to the top of the screen. Other valid values for dock
are "right", "bottom", or "left"; but "top" is most appropriate for a header.
The DOM, or Document Object Model, is a term borrowed from the web world. Textual doesn't use documents but the term has stuck. In Textual CSS, the DOM is an arrangement of widgets you can visualize as a tree-like structure.
Some widgets contain other widgets: for instance, a list control widget will likely also have item widgets, or a dialog widget may contain button widgets. These child widgets form the branches of the tree.
Let's look at a trivial Textual app.
=== "dom1.py"
```python
--8<-- "docs/examples/guide/dom1.py"
```
=== "Output"
```{.textual path="docs/examples/guide/dom1.py"}
```
This example creates an instance of ExampleApp
, which will implicitly create a Screen
object. In DOM terms, the Screen
is a child of ExampleApp
.
With the above example, the DOM will look like the following:
This doesn't look much like a tree yet. Let's add a header and a footer to this application, which will create more branches of the tree:
=== "dom2.py"
```python hl_lines="7 8"
--8<-- "docs/examples/guide/dom2.py"
```
=== "Output"
```{.textual path="docs/examples/guide/dom2.py"}
```
With a header and a footer widget the DOM looks like this:
!!! note
We've simplified the above example somewhat. Both the Header and Footer widgets contain children of their own. When building an app with pre-built widgets you rarely need to know how they are constructed unless you plan on changing the styles of individual components.
Both Header and Footer are children of the Screen object.
To further explore the DOM, we're going to build a simple dialog with a question and two buttons. To do this we're going to import and use a few more builtin widgets:
textual.layout.Container
For our top-level dialog.textual.layout.Horizontal
To arrange widgets left to right.textual.widgets.Static
For simple content.textual.widgets.Button
For a clickable button.
--8<-- "docs/examples/guide/dom3.py"
We've added a Container to our DOM which (as the name suggests) is a container for other widgets. The container has a number of other widgets passed as positional arguments which will be added as the children of the container. Not all widgets accept child widgets in this way. A Button widget doesn't require any children, for example.
Here's the DOM created by the above code:
Here's the output from this example:
You may recognize some elements in the above screenshot, but it doesn't quite look like a dialog. This is because we haven't added a stylesheet.
To add a stylesheet set the CSS_PATH
classvar to a relative path:
!!! note
Textual CSS files are typically given the extension `.tcss` to differentiate them from browser CSS (`.css`).
--8<-- "docs/examples/guide/dom4.py"
You may have noticed that some constructors have additional keyword arguments: id
and classes
.
These are used by the CSS to identify parts of the DOM. We will cover these in the next section.
Here's the CSS file we are applying:
--8<-- "docs/examples/guide/dom4.tcss"
The CSS contains a number of rule sets with a selector and a list of rules. You can also add comments with text between /*
and */
which will be ignored by Textual. Add comments to leave yourself reminders or to temporarily disable selectors.
With the CSS in place, the output looks very different:
You can also set the CSS_PATH
class variable to a list of paths. Textual will combine the rules from all of the supplied paths.
It is reasonable to ask why use CSS at all? Python is a powerful and expressive language. Wouldn't it be easier to set styles in your .py
files?
A major advantage of CSS is that it separates how your app looks from how it works. Setting styles in Python can generate a lot of spaghetti code which can make it hard to see the important logic in your application.
A second advantage of CSS is that you can customize builtin and third-party widgets just as easily as you can your own app or widgets.
Finally, Textual CSS allows you to live edit the styles in your app. If you run your application with the following command, any changes you make to the CSS file will be instantly updated in the terminal:
textual run my_app.py --dev
Being able to iterate on the design without restarting the application makes it easier and faster to design beautiful interfaces.
A selector is the text which precedes the curly braces in a set of rules. It tells Textual which widgets it should apply the rules to.
Selectors can target a kind of widget or a very specific widget. For instance, you could have a selector that modifies all buttons, or you could target an individual button used in one dialog. This gives you a lot of flexibility in customizing your user interface.
Let's look at the selectors supported by Textual CSS.
The type selector matches the name of the (Python) class. Consider the following widget class:
from textual.widgets import Static
class Alert(Static):
pass
Alert widgets may be styled with the following CSS (to give them a red border):
Alert {
border: solid red;
}
The type selector will also match a widget's base classes.
Consequently, a Static
selector will also style the button because the Alert
(Python) class extends Static
.
Static {
background: blue;
border: rounded green;
}
!!! note
The fact that the type selector matches base classes is a departure from browser CSS which doesn't have the same concept.
You may have noticed that the border
rule exists in both Static
and Alert
.
When this happens, Textual will use the most recently defined sub-class.
So Alert
wins over Static
, and Static
wins over Widget
(the base class of all widgets).
Hence if both rules were in a stylesheet, Alert
widgets would have a "solid red" border and not a "rounded green" border.
Every Widget can have a single id
attribute, which is set via the constructor. The ID should be unique to its container.
Here's an example of a widget with an ID:
yield Button(id="next")
You can match an ID with a selector starting with a hash (#
). Here is how you might draw a red outline around the above button:
#next {
outline: red;
}
A Widget's id
attribute can not be changed after the Widget has been constructed.
Every widget can have a number of class names applied. The term "class" here is borrowed from web CSS, and has a different meaning to a Python class. You can think of a CSS class as a tag of sorts. Widgets with the same tag will share styles.
CSS classes are set via the widget's classes
parameter in the constructor. Here's an example:
yield Button(classes="success")
This button will have a single class called "success"
which we could target via CSS to make the button a particular color.
You may also set multiple classes separated by spaces. For instance, here is a button with both an error
class and a disabled
class:
yield Button(classes="error disabled")
To match a Widget with a given class in CSS you can precede the class name with a dot (.
). Here's a rule with a class selector to match the "success"
class name:
.success {
background: green;
color: white;
}
!!! note
You can apply a class name to any widget, which means that widgets of different types could share classes.
Class name selectors may be chained together by appending another full stop and class name. The selector will match a widget that has all of the class names set. For instance, the following sets a red background on widgets that have both error
and disabled
class names.
.error.disabled {
background: darkred;
}
Unlike the id
attribute, a widget's classes can be changed after the widget was created. Adding and removing CSS classes is the recommended way of changing the display while your app is running. There are a few methods you can use to manage CSS classes.
- [add_class()][textual.dom.DOMNode.add_class] Adds one or more classes to a widget.
- [remove_class()][textual.dom.DOMNode.remove_class] Removes class name(s) from a widget.
- [toggle_class()][textual.dom.DOMNode.toggle_class] Removes a class name if it is present, or adds the name if it's not already present.
- [has_class()][textual.dom.DOMNode.has_class] Checks if one or more classes are set on a widget.
- [classes][textual.dom.DOMNode.classes] Is a frozen set of the class(es) set on a widget.
The universal selector is denoted by an asterisk and will match all widgets.
For example, the following will draw a red outline around all widgets:
* {
outline: solid red;
}
Pseudo classes can be used to match widgets in a particular state. Pseudo classes are set automatically by Textual. For instance, you might want a button to have a green background when the mouse cursor moves over it. We can do this with the :hover
pseudo selector.
Button:hover {
background: green;
}
The background: green
is only applied to the Button underneath the mouse cursor. When you move the cursor away from the button it will return to its previous background color.
Here are some other pseudo classes:
:blur
Matches widgets which do not have input focus.:dark
Matches widgets in dark themes (whereApp.theme.dark == True
).:disabled
Matches widgets which are in a disabled state.:enabled
Matches widgets which are in an enabled state.:even
Matches a widget at an evenly numbered position within its siblings.:first-of-type
Matches a widget that is the first of its type amongst its siblings.:focus-within
Matches widgets with a focused child widget.:focus
Matches widgets which have input focus.:inline
Matches widgets when the app is running in inline mode.:last-of-type
Matches a widget that is the last of its type amongst its siblings.:light
Matches widgets in light themes (whereApp.theme.dark == False
).:odd
Matches a widget at an oddly numbered position within its siblings.
More sophisticated selectors can be created by combining simple selectors. The logic used to combine selectors is know as a combinator.
If you separate two selectors with a space it will match widgets with the second selector that have an ancestor that matches the first selector.
Here's a section of DOM to illustrate this combinator:
Let's say we want to make the text of the buttons in the dialog bold, but we don't want to change the Button in the sidebar. We can do this with the following rule:
#dialog Button {
text-style: bold;
}
The #dialog Button
selector matches all buttons that are below the widget with an ID of "dialog". No other buttons will be matched.
As with all selectors, you can combine as many as you wish. The following will match a Button
that is under a Horizontal
widget and under a widget with an id of "dialog"
:
#dialog Horizontal Button {
text-style: bold;
}
The child combinator is similar to the descendant combinator but will only match an immediate child. To create a child combinator, separate two selectors with a greater than symbol (>
). Any whitespace around the >
will be ignored.
Let's use this to match the Button in the sidebar given the following DOM:
We can use the following CSS to style all buttons which have a parent with an ID of sidebar
:
#sidebar > Button {
text-style: underline;
}
It is possible that several selectors match a given widget. If the same style is applied by more than one selector then Textual needs a way to decide which rule wins. It does this by following these rules:
-
The selector with the most IDs wins. For instance
#next
beats.button
and#dialog #next
beats#next
. If the selectors have the same number of IDs then move to the next rule. -
The selector with the most class names wins. For instance
.button.success
beats.success
. For the purposes of specificity, pseudo classes are treated the same as regular class names, so.button:hover
counts as 2 class names. If the selectors have the same number of class names then move to the next rule. -
The selector with the most types wins. For instance
Container Button
beatsButton
.
The specificity rules are usually enough to fix any conflicts in your stylesheets. There is one last way of resolving conflicting selectors which applies to individual rules. If you add the text !important
to the end of a rule then it will "win" regardless of the specificity.
!!! warning
Use `!important` sparingly (if at all) as it can make it difficult to modify your CSS in the future.
Here's an example that makes buttons blue when hovered over with the mouse, regardless of any other selectors that match Buttons:
Button:hover {
background: blue !important;
}
You can define variables to reduce repetition and encourage consistency in your CSS.
Variables in Textual CSS are prefixed with $
.
Here's an example of how you might define a variable called $border
:
$border: wide green;
With our variable assigned, we can write $border
and it will be substituted with wide green
.
Consider the following snippet:
#foo {
border: $border;
}
This will be translated into:
#foo {
border: wide green;
}
Variables allow us to define reusable styling in a single place. If we decide we want to change some aspect of our design in the future, we only have to update a single variable.
!!! note
Variables can only be used in the _values_ of a CSS declaration. You cannot, for example, refer to a variable inside a selector.
Variables can refer to other variables.
Let's say we define a variable $success: lime;
.
Our $border
variable could then be updated to $border: wide $success;
, which will
be translated to $border: wide lime;
.
All CSS rules support a special value called initial
, which will reset a value back to its default.
Let's look at an example. The following will set the background of a button to green:
Button {
background: green;
}
If we want a specific button (or buttons) to use the default color, we can set the value to initial
.
For instance, if we have a widget with a (CSS) class called dialog
, we could reset the background color of all buttons inside the dialog with the following CSS:
.dialog Button {
background: initial;
}
Note that initial
will set the value back to the value defined in any default css.
If you use initial
within default css, it will treat the rule as completely unstyled.
!!! tip "Added in version 0.47.0"
CSS rule sets may be nested, i.e. they can contain other rule sets. When a rule set occurs within an existing rule set, it inherits the selector from the enclosing rule set.
Let's put this into practical terms. The following example will display two boxes containing the text "Yes" and "No" respectively. These could eventually form the basis for buttons, but for this demonstration we are only interested in the CSS.
=== "nesting01.tcss (no nesting)"
```css
--8<-- "docs/examples/guide/css/nesting01.tcss"
```
=== "nesting01.py"
```python
--8<-- "docs/examples/guide/css/nesting01.py"
```
=== "Output"
```{.textual path="docs/examples/guide/css/nesting01.py"}
```
The CSS is quite straightforward; there is one rule for the container, one for all buttons, and one rule for each of the buttons. However it is easy to imagine this stylesheet growing more rules as we add features.
Nesting allows us to group rule sets which have common selectors.
In the example above, the rules all start with #questions
.
When we see a common prefix on the selectors, this is a good indication that we can use nesting.
The following produces identical results to the previous example, but adds nesting of the rules.
=== "nesting02.tcss (with nesting)"
```css
--8<-- "docs/examples/guide/css/nesting02.tcss"
```
=== "nesting02.py"
```python
--8<-- "docs/examples/guide/css/nesting02.py"
```
=== "Output"
```{.textual path="docs/examples/guide/css/nesting02.py"}
```
!!! tip
Indenting the rule sets is not strictly required, but it does make it easier to understand how the rule sets are related to each other.
In the first example we had a rule set that began with the selector #questions .button
, which would match any widget with a class called "button" that is inside a container with id questions
.
In the second example, the button rule selector is simply .button
, but it is within the rule set with selector #questions
.
The nesting means that the button rule set will inherit the selector from the outer rule set, so it is equivalent to #questions .button
.
The two remaining rules are nested within the button rule, which means they will inherit their selectors from the button rule set and the outer #questions
rule set.
You may have noticed that the rules for the button styles contain a syntax we haven't seen before.
The rule for the Yes button is &.affirmative
.
The ampersand (&
) is known as the nesting selector and it tells Textual that the selector should be combined with the selector from the outer rule set.
So &.affirmative
in the example above, produces the equivalent of #questions .button.affirmative
which selects a widget with both the button
and affirmative
classes.
Without &
it would be equivalent to #questions .button .affirmative
(note the additional space) which would only match a widget with class affirmative
inside a container with class button
.
For reference, lets see those two CSS files side-by-side:
=== "nesting01.tcss"
```css
--8<-- "docs/examples/guide/css/nesting01.tcss"
```
=== "nesting02.tcss"
```sass
--8<-- "docs/examples/guide/css/nesting02.tcss"
```
Note how nesting bundles related rules together. If we were to add other selectors for additional screens or widgets, it would be easier to find the rules which will be applied.
There is no requirement to use nested CSS, but it can help to group related rule sets together (which makes it easier to edit). Nested CSS can also help you avoid some repetition in your selectors, i.e. in the nested CSS we only need to type #questions
once, rather than four times in the non-nested CSS.