Derive and run

A program that is a list of steps looks like a sequence. For pure steps, it is not. It is a graph that someone wrote down in one order.

The glyphs so far: 15, 3 new on this page

• 42Data: a value without a type
• number

  42
  A document: the tab is its type
• number

  42
  A dashed tab: the container declares the type
• add

  number

  1

  +

  number

  2
  Parts: a card inside a card
• kindA kind: a document of type kind
• OKDA rung: own, kind or default
• =3A port: the result of a document
• 0A sealed part: an evaluator cannot read it
• #NAME?An error value, in a red port
• numbera

  3
  A name tag: a binding for later steps
• ref

  namesa
  A reference: a path from a root of the context
• if

  test

  bool

  true

  =true

  then

  text

  “yes”

  =“yes”

  else

  text

  “no”

  not evaluated

  =“yes”
  Not evaluated: a part that its parent did not need
• ▸log

  text

  “hi”

  =“hi”

  =“hi”
  The run side: a magenta edge, and an effectnew
• 1log“hi”An effect: a numbered ticket in the worldnew
• logA capability: an effect that a term can causenew

Try a random order8 of 120 orders give the same result.

As written

1. 1a: Load the priceseffect
2. 2b: Load the rateseffect
3. 3c: Convert a with beffect
4. 4d: Write a log lineeffect
5. 5e: Save ceffect

As a graph

• a step reads a result
• effects keep their order

A list of pure steps is a graph that someone wrote down in one order. Each order that obeys the arrows gives the same result. A step with an effect also keeps its place among the other effects. Mark more effects, and fewer orders stay correct.

Order has a meaning only for effects

A pure step reads values and gives a value. Two orders that obey the arrows give the same result. A step with an effect is different. A log line before a save and a log line after a save are two different histories. Thus each effect keeps its place among the other effects.

Two sides

coeval puts each kind on one of two sides.

	Derive	Run
Effects	None	Yes, with capabilities
Order	Any order that obeys the data	The written order
When	When something reads the value, or when its data changes	When a trigger starts it
Again after a change of data	Yes	No
Examples	Formulas, trees of components, layout rules, queries	Sequences of calls, event handlers

An event handler holds a run as data. The derive side can compute the handler, and the run starts only when its event comes. A change of data does not repeat an effect. The loop on the home page shows both sides.