write beautiful state charts
Makes this
from this
initial,
doing: entry/ write unit test
do/ write code
exit/ ...,
# smcat recognizes initial
# and final states by name
# and renders them appropriately
final;
initial => "on backlog" : item adds most value;
"on backlog" => doing : working on it;
doing => testing : built & unit tested;
testing => "on backlog" : test not ok;
testing => final : test ok;
To enable me to make state charts ...
- ... that look good
- ... with the least effort possible
- ... whithout having to interact with drag and drop tools. Entering text is fine, doing my own layout is not.
- ... without having to dive into GraphViz
dot
each time. GraphViz is cool, but is was not designed to write & maintain conceptual documents in (You'll know what I'm talking about if you ever tried to get it to draw nested nodes. Or edges between those. )
A no-frills interpreter on line: state-machine-cat.js.org.
There's an Atom package
with syntax highlighting, a previewer and some export options. You can install
it from within Atom (search for state machine cat in the install section
of the settings screen) or use apm i state-machine-cat-preview
if you're a command line person.
OTOH. if you're a command line person the command line interface might be something for you too:
Just npm install --global state-machine-cat
and run smcat
This is what smcat --help
would get you:
Usage: smcat [options] [infile]
Options:
-V, --version output the version number
-T --output-type <type> svg|dot|smcat|json|ast|html|scxml|scjson|xmi (default: "svg")
-I --input-type <type> smcat|json (default: "smcat")
-E --engine <type> dot|circo|fdp|neato|osage|twopi (default: "dot")
-d --direction <dir> top-down|bottom-top|left-right|right-left (default: "top-down")
-o --output-to <file> File to write to. use - for stdout.
--dot-graph-attrs <string> graph attributes to pass to the dot render engine
--dot-node-attrs <string> node attributes to pass to the dot render engine
--dot-edge-attrs <string> edge attributes to pass to the dot render engine
-l --license Display license and exit
-h, --help output usage information
... so to convert the above chart to sample.svg
bin/smcat docs/sample.smcat
Or, if you'd rather have the native GraphViz dot do that for you:
bin/smcat -T dot docs/sample.smcat -o - | dot -T svg -odoc/sample.svg
Leaving the options at the default settings usually deliver the best results already, so if they bewilder you: don't worry.
The --dot-graph-attrs
(and the node and edge variants thereof) exist in case you want
to override default attributes in the generated picture; e.g. to get a transparent background
and draw edges as line segments instead of splines, use this:
bin/smcat --dot-graph-attrs "bgcolor=transparent splines=line" docs/sample.smcat
state machine cat can emit a valid core constructs scxml document. If you're into that sort of thing you can read all about it in State Machine Cat and SCXML.
If you want to write state charts in state machine cat and use that to make working state machines using a tool that eats XMI and emits code you can do just that - read more in State Machine Cat and XMI
After you npm i
'd state-machine-cat
:
const smcat = require("state-machine-cat");
try {
const lSVGInAString = smcat.render(
`
initial => backlog;
backlog => doing;
doing => test;
`,
{
outputType: "svg"
}
);
console.log(lSVGInAString);
} catch (pError) {
console.error(pError);
}
Read more in docs/api.md
on => off;
- smcat automatically declares the states. You can explicitly declare them if you want them to have more than a name only - see state declarations below.
on => off: switch;
UML prescribes to place conditions after events, to place
conditions within squares and to place actions
after a /
: from => to: event [conditions]/ actions
, e.g. on => off: switch flicked [not an emergency]/ light off;
.
You're free to do so, but smcat doesn't check for it. It internally takes the notation into account, though and if you choose to export to json, scxml or scjson you'll see them nicely split out.
on => off: switch flicked/
light off;
off => on: switch flicked/
light on;
You note that smcat rendered the states in this chart top down instead of left to right. It did that because we told it so. You can do that too with
--direction
on the command line, or in atom bydirection -> left to right
ordirection -> top down
from the chart's submenu.
# this is a note
on => off;
If you need to define activities (e.g. entry
or exit
triggers) on
a state, you can explicitly declare the state and put the activites
after a colon:
# yep, notes get rendered here as well
# multiple notes translate into multiple
# lines in notes in the diagram
doing:
entry/ make a feature branch
exit/ deploy code on production
...;
smcat recognizes the entry/
and exit/
keywords and treats
everything after it on the same line to be the 'body' of the
trigger.
Here too: you're free to use them, but you don't have to. smcat takes them into account in its internal representation and uses them in exports to json, scxml and scjson.
If you want to use a display names that differ from how you name the states (e.g. if the display names are long), you can do so by adding a label to them:
on [label="Lamp aan"],
off [label="Lamp uit"];
off => on: on pushed;
on => off: off pushed;
When initial
or final
, is part of a state's name smcat treats
it as the UML 'pseudo states' initial
and final
respectively:
initial => todo;
todo => doing;
doing => done;
done => final;
smcat recognizes states with history
in their name als history states:
initial,
"power off",
running {
running.history;
washing -> rinsing: 20 min;
rinsing -> spinning: 20 min;
spinning -> final: 20 min;
},
final;
initial => washing;
running => "power off": power out;
"power off" => running.history: restore power;
History states are shallow by default. If you want a history state to
be deep just put that somewhere in the name (e.g. "running deep history"
or running.history.deep
) - and smcat will render them as such.
smcat treats states starting with ^
as UML pseudo state choice. Strictly
speaking 'choice' is a superfluous element of the UML state machine
specification, but it is there and sometimes it makes diagrams easier to read.
^fraud?: transaction fraudulent?;
initial -> reserved;
reserved -> quoted:
quote
requested;
quoted -> ^fraud?: payment;
^fraud? -> ticketed: [no];
^fraud? -> removed: [yes];
ticketed -> final;
removed -> final;
In UML you can fork state transitions into multiple or join them into one
with the fork (one to many) join (many to one) and junction (many to many)
pseudo states. Fork and join are represented by a black bar, junction
by a filled circle.
To make a join, fork or junction pseudo state, start its
name with a ]
.
Here's an example of a join:
a => ]join;
b => ]join;
]join => c;
State machine cat automatically derives which of the three types you meant by counting the number of incoming and the number of outgoing connections:
- one incoming and multiple outgoing: it's a fork
- multiple incoming and one outgoing: it's a join
- all other cases: it's a junction
If you want to defy UML semantics you can do that with explicit type overrides .
a => ]kruising;
b => ]kruising;
]kruising => c;
]kruising => d;
UML has a special pseudo state to indicate your state machine didn't exit properly: terminate. If you want to use it, declare it explicitly:
Aahnohd [type=terminate label="Terminated"];
a => Aahnohd: [hit by meteorite];
For proper exits you'd typically use the final state.
- when you need
;
,,
,{
,[
or spaces as part of a state - place em in quotes"a state"
- Activities have the same restriction, except they allow spaces.
- Labels have the same restriction as activities, except they allow for
,
too. - State declaration precedence is: deep wins from shallow; explicit wins from implicit
- It's possible to declare the same state multiple times on the same level, buts smcat will take the last declaration into account only. For example:
This
# first declaration of "cool state"
"cool state",
"other state",
# second declaration of "cool state"
"cool state": cool down;
results in (/ is equivalent to):
# second declaration of "cool state"
"cool state": cool down,
"other state";
It's possible to have state machines within states. the states stopped, playing and pause can only occur when the tape player is on:
initial,
"tape player off",
"tape player on" {
stopped => playing : play;
playing => stopped : stop;
playing => paused : pause;
paused => playing : pause;
paused => stopped : stop;
};
initial => "tape player off";
"tape player off" => stopped : power;
"tape player on" => "tape player off" : power;
If stuff is happening in parallel within a state you can express that too. Just make sure the state has the word "parallel" in there:
initial,
bla.parallel {
first{
first.thing -> first.thang;
},
second{
second.thing -> second.thang;
second.thang -> second.thing;
};
},
final;
initial -> ]split;
]split -> first.thing;
]split -> second.thing;
first.thang -> ]merge;
second.thang -> ]merge;
]merge -> final;
You can mark one or more states as active by adding active
as an
attribute for that state e.g. to make the do
state an active one
do this:
do [active];
initial -> plan;
plan -> do;
do -> study;
study -> act;
act -> plan;
which will result in
As of version 4.2.0 state-machine-cat has (experimental) support for colors on both states and transitions.
For example, this ...
eat [color="#008800"],
sleep [color="blue" active],
meow [color="red"],
play [color="purple"];
sleep -> meow [color="red"] : wake up;
meow -> meow [color="red"] : no response from human;
meow -> eat [color="#008800"]: human gives food;
meow -> play [color="purple"] : human gives toy;
play -> sleep [color="blue"] : tired or bored;
eat -> sleep [color="blue"] : belly full;
What does 'experimental' mean?
The color attribute is probably here to stay.
However, I haven't found the balance between ease of use and expressiveness yet. Should the text in a state be rendered in the same color? should the background color automatically be rendered as well? In the same color, or in a shade smartly derived? Or should I include a bunch of color attributes (e.g. fillcolor, textcolor, textbgcolor) for ultimate control?
As you read above, state machine cat derives the type of a state from its name.
In some cases that might not be what you want. In those cases, you can
override the type with the type
attribute, like in this example for the
initialized
state.
initial,
starting,
initialized [color="red" type=regular],
running,
final;
initial => starting;
starting => initialized;
initialized => running;
initialized => stopped;
running => stopped;
stopped => final;
The values you can use for type
:
type |
---|
regular |
initial |
final |
terminate |
parallel |
history |
deephistory |
choice |
fork |
join |
forkjoin |
junction |
I made the parser with pegjs - you can find it at src/parse/peg/smcat-parser.pegjs
- Thoroughly tested and good enough for public use.
- Despite this you might bump into the occasional issue - don't hesitate to report it, either on GitLab or on GitHub.
- Runs on latest versions of firefox, safari and chrome and node versions >= 6. Although it might run on other environments, it's not tested there. I will reject issues on these other environments, unless they're accompanied with sufficient enticement to make it worth my while.