The State Monad

Eucalypt provides an optional state monad library for threading block-valued state through a sequence of operations. Import it with:

{ import: "state.eu" }

The Problem

Eucalypt blocks are immutable. To "update" a value deep in a structure, you must reconstruct the entire surrounding structure. When several updates depend on each other, manually threading the modified state through each step is verbose and error-prone — the same problem that { :random ... } solves for PRNG streams.

s0: { count: 0, name: "init" }
s1: merge(s0, {count: 1})
s2: merge(s1, {name: "updated"})

s0:
  count: 0
  name: init
s1:
  count: 1
  name: init
s2:
  count: 1
  name: updated

The state monad automates this threading.

State Actions

A state action is a function from a state block to a block with value and state fields — the same pattern as the random monad's value/rest blocks:

# An action that reads the count and increments it
count-and-inc(s): { value: s.count, state: s merge({count: s.count + 1}) }

The state monad provides primitives for building these actions without manually deconstructing and reconstructing blocks.

Monadic Blocks with `{ :state ... }`

Tag a block with :state to get automatic bind threading — each declaration becomes a monadic step:

{ import: "state.eu" }

action: { :state
  n: state.query(_.count)
  _: state.put(:count, n + 1)
  _: state.put(:name, "step-2")
}

result: state.run(action, {count: 0, name: "init"}).state

result:
  count: 1
  name: step-2

Each step sees the state left by the previous step:

state.get returns the whole state as the value — then .count navigates it
state.put(:count, n + 1) sets a key in the state
The implicit return synthesises {n: n} from the non-underscore bindings (or use .(expr) for an explicit return)

Run the action by passing an initial state to state.run (returns a value/state block) or state.eval (value only); the final state alone is state.run(action, s).state.

Primitives

Function	Description
`state.get`	Action: return the whole state as the value
`state.put(k, v)`	Action: set key `k` to value `v`
`state.lift(f)`	Action: apply `f` to transform the state, return null
`state.query(f)`	Action: apply `f` to read from the state, state unchanged
`state.modify(k, f)`	Action: update key `k` by applying `f` to its value

Running Actions

Function	Description
`state.run(action, s)`	Run action from state `s`, return block with `value` and `state` fields
`state.eval(action, s)`	Run action, return only the value
`state.exec(action, s)`	Deprecated — use `state.run(action, s).state`. Run action, return only the final state

Using Existing Pipeline Functions

The lift and query primitives let you reuse existing eucalypt functions as state actions:

{ import: "state.eu" }

action: { :state
  _: :count %! (+ 1)
  n: state.query(_.count)
}.(n)

result: state.eval(action, {count: 0})

result: 1

With the lens library imported, you can also use state.lift with lens functions like over(at(:key), f) for deep updates. Since state.eu imports lens.eu internally, the lens functions are available without a separate import.

Lens Operators

The =! and %! operators provide concise syntax for lens-based state updates. When the left operand is a symbol, it is automatically wrapped in at():

{ import: "state.eu" }

action: { :state
  _: :count =! 0
  _: :count %! (+ 1)
  _: :count %! (+ 1)
}

result: state.run(action, {count: 99}).state.count

result: 2

Operator	Meaning	Equivalent to
`lens =! val`	Set at lens	`state.lift(over(lens, -> val))`
`lens %! fn`	Modify at lens	`state.lift(over(lens, fn))`

Comparison with the Random Monad

The state monad follows the same pattern as { :random ... }:

	Random	State
State type	PRNG stream	Block
Import	Built-in (prelude)	`{ import: "state.eu" }`
Block tag	`{ :random ... }`	`{ :state ... }`
Run	`random.run(action, stream)`	`state.run(action, initial)`
Read state	`random.float`, `random.int(n)`	`state.get`, `state.query(f)`
Modify state	Automatic (stream advances)	`state.put`, `state.modify`, `=!`, `%!`

Example: Accumulating Counts

{ import: "state.eu" }

count-item(item): { :state
  current: state.query(~ item)
  _: state.put(item, current fnil(+ 1, 0))
}

count-items(s, item): state.exec(count-item(item), s)

result: [:a, :b, :a, :c, :a, :b] foldl(count-items, {})

result:
  a: 3
  b: 2
  c: 1

How It Works

Under the hood, state-bind and state-ret are:

state-ret(v, s): { value: v, state: s }

state-bind(action, f, s): {
  r: action(s)
  run: f(r.value)(r.state)
  value: run.value
  state: run.state
}

The state namespace is registered with monad: true metadata, so the desugarer recognises { :state ... } blocks and chains declarations through state.bind / state.return automatically.