This project is part of the @thi.ng/umbrella monorepo.
This package provides a single function, a general purpose Finite State Machine transducer, which acts as useful & lightweight mechanism to provide context-sensitive processing capabilities as part of a transducer transformation pipeline.
yarn add @thi.ng/transducers-fsm
For a real world example, the @thi.ng/sax package provides a SAX-like XML parser transducer, built around the FSM provided here.
The following example defines a simple FSM with 3 states:
skip
take
done
The FSM always starts in the skip
state.
The FSM alternates between skipping or consuming (passing through) 5
inputs as long as each input is < 20. Once an input is >= 20, the FSM
switches into the done
state, which has been declared as a terminal
state and once entered will cause processing to terminate (also see API
description further below).
import { fsm } from '@thi.ng/transducers-fsm'
import * as tx from '@thi.ng/transducers'
import { isOdd } from '@thi.ng/checks'
const testFSM = fsm({
// initial state initializer
// (called before processing 1st input)
init: () => ({ state: "skip", count: 0 }),
// terminal state ID
terminate: "done",
// individual state handlers
states: {
// skip state
skip: (state, x) => {
if (x < 20) {
if (++state.count > 5) {
state.state = "take";
state.count = 1;
return [x];
}
} else {
state.state = "done";
}
},
// take state
take: (state, x) => {
if (x < 20) {
if (++state.count > 5) {
state.state = "skip";
state.count = 1;
} else {
return [x];
}
} else {
state.state = "done";
}
},
// terminal state, ignore inputs
done: () => { },
},
});
[...tx.iterator(testFSM, tx.range(100))]
// [ 5, 6, 7, 8, 9, 15, 16, 17, 18, 19 ]
// Use FSM as part of composed transducers...
[...tx.iterator(tx.comp(tx.takeNth(2), testFSM), tx.range(100))]
// [ 10, 12, 14, 16, 18 ]
[
...tx.iterator(
tx.comp(
tx.mapcat((x) => x.split(/[,\s]+/g)),
tx.map((x) => parseInt(x)),
testFSM,
tx.filter(isOdd)
),
["9,8,7,6", "14 1 0 17 15 16", "19,23,12,42,4"]
)
]
// [ 1, 17, 15 ]
Finite State Machine transducer. Takes an FSM configuration object and returns a transducer, which processes inputs using the provided state handler functions, which in turn can produce any number of outputs per consumed input.
Before processing the first input, the FSM state is initialized by
calling the user provided init()
function, which MUST return a state
object with at least a state
key, whose value is used for dynamic
(i.e. stateful) dispatch during input processing. This state object is
passed with each input value to the current state handler, which is
expected to mutate this object, e.g. to cause state changes based on
given inputs.
If a state handler needs to "emit" results for downstream processing, it
can return an array of values. Any such values are passed on
(individually, not as array) to the next reducer in the chain. If a
state handler returns null
or undefined
, further downstream
processing of the current input is skipped.
Regardless of return value, if a state handler has caused a state change
to the configured terminate
state, processing is terminated (by calling
ensureReduced()
) and no further inputs will be consumed.
- Karsten Schmidt
© 2018 Karsten Schmidt // Apache Software License 2.0