Documentation

Observe the demand behavior

• in a given evaluation context,
• of a given unary function,
• called upon a given input,

returning a pair of

• the demand on its output exerted by the evaluation context, and
• the demand on its input this induced

Suppose we want to see how strict reverse is when we evaluate its result to weak-head normal form:

>>> (b, a) = observe1 (`seq` ()) (reverse @Int) [1, 2, 3]
>>> printDemand b  -- output demand
_ : _
>>> printDemand a  -- input demand
_ : _ : _ : _ : []

This tells us that our context did indeed evaluate the result of reverse to force only its first constructor, and that doing so required the entire spine of the list to be evaluated, but did not evaluate any of its elements.

Observe the demand behavior

• in a given evaluation context
• of a given curried n-ary function
• called upon all of its inputs (provided as curried ordinary inputs),

returning a pair of

• the demand on its output exerted by the evaluation context, and
• the demands on its inputs this induced, represented as an NP Demand from Generics.SOP

This function is variadic and curried: it takes n + 2 arguments, where n is the total number of arguments taken by the observed function.

Suppose we want to see how strict zipWith (*) is when we evaluate its result completely (to normal form):

>>> productZip = zipWith ((*) @Int)
>>> (zs, (xs :* ys :* Nil)) = observe normalize productZip [10, 20] [30, 40]
>>> printDemand zs  -- output demand
300 : 800 : []
>>> printDemand xs  -- input demand #1
10 : 20 : []
>>> printDemand ys  -- input demand #2
30 : 40 : _

If you haven't thought very carefully about the strictness behavior of zip, this may be a surprising result; this is part of the fun!

Observe the demand behavior

• in a given evaluation context
• of a given uncurried n-ary function (taking as input an n-ary product of inputs represented as an NP I from Generics.SOP)
• called upon all of its inputs (provided as curried ordinary inputs),

returning a pair of

• the demand on its output exerted by the evaluation context, and
• the demands on its inputs this induced, represented as an NP Demand from Generics.SOP

This is mostly useful for implementing the internals of StrictCheck; observe is more ergonomic for exploration by end-users.