Basic tgui backend code consists of the following vars and procs:
ui_interact(mob/user, datum/tgui/ui)
ui_data(mob/user)
ui_act(action, params)
ui_state()
src_object
- The atom, which UI corresponds to in the game world.ui_interact
- The proc where you will handle a request to open an interface. Typically, you would update an existing UI (if it exists), or set up a new instance of UI by calling theSStgui
subsystem.ui_data
- In this proc you munges whatever complex data yoursrc_object
has into an associative list, which will then be sent to UI as a JSON string.ui_act
- This proc receives user actions and reacts to them by changing the state of the game.ui_state
- This proc dictates under what conditions a UI may be interacted with. This may be the standard checks that check if you are in range and conscious, or more.
Once backend is complete, you create an new interface component on the frontend, which will receive this JSON data and render it on screen.
States are easy to write and extend, and what make tgui interactions so powerful. Because states can be overridden from other procs, you can build powerful interactions for embedded objects or remote access.
Let's start with a very basic hello world.
/obj/machinery/my_machine/ui_interact(mob/user, datum/tgui/ui)
ui = SStgui.try_update_ui(user, src, ui)
if(!ui)
ui = new(user, src, "MyMachine")
ui.open()
This is the proc that defines our interface. There's a bit going on here, so
let's break it down. First, we override the ui_interact proc on our object. This
will be called by interact
for you, which is in turn called by attack_hand
(or attack_self
for items). ui_interact
is also called to update a UI (hence
the try_update_ui
), so we accept an existing UI to update.
Inside the if(!ui)
block (which means we are creating a new UI), we choose our
template, title, and size; we can also set various options like style
(for
themes), or autoupdate. These options will be elaborated on later (as will
ui_state
s).
After ui_interact
, we need to define ui_data
. This just returns a list of
data for our object to use. Let's imagine our object has a few vars:
/obj/machinery/my_machine/ui_data(mob/user)
var/list/data = list()
data["health"] = health
data["color"] = color
return data
The ui_data
proc is what people often find the hardest about tgui, but its
really quite simple! You just need to represent your object as numbers, strings,
and lists, instead of atoms and datums.
Finally, the ui_act
proc is called by the interface whenever the user used an
input. The input's action
and params
are passed to the proc.
/obj/machinery/my_machine/ui_act(action, params)
if(..())
return
if(action == "change_color")
var/new_color = params["color"]
if(!(color in allowed_coors))
return FALSE
color = new_color
. = TRUE
update_icon()
The ..()
(parent call) is very important here, as it is how we check that the
user is allowed to use this interface (to avoid so-called href exploits). It is
also very important to clamp and sanitize all input here. Always assume the user
is attempting to exploit the game.
Also note the use of . = TRUE
(or FALSE
), which is used to notify the UI
that this input caused an update. This is especially important for UIs that do
not auto-update, as otherwise the user will never see their change.
Finally, let's make a React Component for your interface. This is also a source of confusion for new developers. If you got some basic javascript and HTML knowledge, that should ease the learning process, although we recommend getting yourself introduced to
A React component is not a regular HTML template. A component is a
javascript function, which accepts a props
object (that contains
properties passed to a component) and a context
object (which is
necessary to access UI data) as arguments, and outputs an HTML-like
structure.
So let's create our first React Component. Create a file with a name
SampleInterface.js
(or any other name you want), and copy this code
snippet (make sure component name matches the file name):
import { useBackend } from '../backend';
import { Button, LabeledList, Section } from '../components';
import { Window } from '../layouts';
export const SampleInterface = (props, context) => {
const { act, data } = useBackend(context);
// Extract `health` and `color` variables from the `data` object.
const {
health,
color,
} = data;
return (
<Window resizable>
<Window.Content scrollable>
<Section title="Health status">
<LabeledList>
<LabeledList.Item label="Health">
{health}
</LabeledList.Item>
<LabeledList.Item label="Color">
{color}
</LabeledList.Item>
<LabeledList.Item label="Button">
<Button
content="Dispatch a 'test' action"
onClick={() => act('test')} />
</LabeledList.Item>
</LabeledList>
</Section>
</Window.Content>
</Window>
);
};
Here are the key variables you get from a useBackend(context)
function:
config
is part of core tgui. It contains meta-information about the interface and who uses it, BYOND refs to various objects, and so forth. You are rarely going to use it, but sometimes it can be used to your advantage when doing complex UIs.data
is the data returned fromui_data
andui_static_data
procs in your DM code. Pretty straight forward.- Note, that javascript doesn't have associative arrays, so when you
return an associative list from DM, it will be available in
data
as a javascript object instead of an array. You can use it normally like so:object.key
, so it's not a problem if it's representing a data structure, but commonArray
methods, such asarray.map(item => ...)
, are not available on it. Always prefer returning clean arrays from your code, since arrays are easier to work with in javascript!
- Note, that javascript doesn't have associative arrays, so when you
return an associative list from DM, it will be available in
act(name, params)
is a function, which you can call to dispatch an action to your DM code. It will be processed inui_act
proc. Action name will be available inparams["action"]
, mixed together with the rest of parameters you have passed inparams
object.
Let's talk about the syntax.
The syntax you're seeing here is called JSX - a very simple extension of the core javascript language. It's basically a pre-processor, that takes expressions that look like html, and turns them into function calls.
Take a look at this example:
<div className={'color-' + status}>
You are in {status} condition!
</div>
After compiling the code above, this is what it becomes:
createElement('div',
{ className: 'color-' + status },
'You are in ', status, ' condition!');
It is very important to remember, that JSX is just a javascript expression
made out of createElement
function calls. Naturally, this allows doing
all sorts of stuff on these expressions, just like you would with anything
else in javascript.
Take a look at these examples:
Render an element inside of another element if showProgress
is true.
This example uses the &&
operator (the logical AND). It returns
the first operand if it evaluates to false
, and returns the second operand
if it evaluates to true
.
If showProgress
is true
, the whole expression evaluates
to a <ProgressBar />
element. If showProgress
is false
, the whole
expression evaluates to false
, and false
is not rendered by React.
<Box>
{showProgress && (
<ProgressBar value={progress} />
)}
</Box>
You can also use the ||
operator (the logical OR), which works the same way,
except it will return the second operand on false
instead of true
.
Loop over the array to map every item to a corresponding React element.
Array.map()
is a method, that calls a function on every item in the array,
and builds a new array based on what was returned by that function.
<LabeledList>
{items.map(item => (
<LabeledList.Item
key={item.id}
label={item.label}>
{item.content}
</LabeledList.Item>
))}
</LabeledList>
If you need more examples of what you can do with React, see the interface conversion guide.
You interface will eventually get really, really big. The easiest thing you can do in this situation, is divide and conquer. Grab a chunk of your JSX code, and wrap it into a second, smaller React component:
import { useBackend } from '../backend';
import { Button, LabeledList, Section } from '../components';
import { Window } from '../layouts';
export const SampleInterface = (props, context) => {
return (
<Window resizable>
<Window.Content scrollable>
<HealthStatus user="Jerry" />
</Window.Content>
</Window>
);
};
const HealthStatus = (props, context) => {
const { act, data } = useBackend(context);
const {
user,
} = props;
const {
health,
color,
} = data;
return (
<Section title={"Health status of: " + user}>
<LabeledList>
<LabeledList.Item label="Health">
{health}
</LabeledList.Item>
<LabeledList.Item label="Color">
{color}
</LabeledList.Item>
</LabeledList>
</Section>
);
};
We all do it, even the best of us. If you just want to make a tgui fast, here's what you need (note that you'll probably be forced to clean your shit up upon code review):
/obj/copypasta/ui_interact(mob/user, datum/tgui/ui)
ui = SStgui.try_update_ui(user, src, ui)
if(!ui)
ui = new(user, src, "copypasta")
ui.open()
/obj/copypasta/ui_data(mob/user)
var/list/data = list()
data["var"] = var
return data
/obj/copypasta/ui_act(action, params)
if(..())
return
switch(action)
if("copypasta")
var/newvar = params["var"]
// A demo of proper input sanitation.
var = CLAMP(newvar, min_val, max_val)
. = TRUE
update_icon() // Not applicable to all objects.
And the template:
import { useBackend } from '../backend';
import { Button, LabeledList, Section } from '../components';
import { Window } from '../layouts';
export const SampleInterface = (props, context) => {
const { act, data } = useBackend(context);
// Extract `health` and `color` variables from the `data` object.
const {
health,
color,
} = data;
return (
<Window resizable>
<Window.Content scrollable>
<Section title="Health status">
<LabeledList>
<LabeledList.Item label="Health">
{health}
</LabeledList.Item>
<LabeledList.Item label="Color">
{color}
</LabeledList.Item>
<LabeledList.Item label="Button">
<Button
content="Dispatch a 'test' action"
onClick={() => act('test')} />
</LabeledList.Item>
</LabeledList>
</Section>
</Window.Content>
</Window>
);
};