nonstrict

Nonstrict, in the context of programming languages and semantics, refers to evaluation behavior where function arguments are not necessarily evaluated before the function runs. In nonstrict (or lazy) languages, an expression may be kept unevaluated until its value is actually needed to produce a result. This contrasts with strict evaluation, where arguments are computed before the function body executes.

The nonstrict approach is typically implemented through evaluation strategies such as call-by-name or call-by-need, often using thunks (delayed computations) to represent unevaluated expressions. Call-by-need, a common form of nonstrict evaluation, shares results so repeated uses do not re-compute the same expression. Nonstrictness allows programs to manipulate infinite data structures and define functions that can be used partially, taking only as much input as required for a given computation.

Nonstrict languages enable certain powerful patterns but also introduce complexity. Reasoning about termination and performance can be more challenging because evaluation order is not always explicit. Side effects compound this challenge: in nonstrict contexts, the timing and manner of effects may vary depending on what is demanded by the program. To manage this, many nonstrict languages enforce purity or provide controlled mechanisms for effects.

Haskell is the most widely cited example of a nonstrict, lazy language, where values are computed only as needed and infinite lists can be handled in practical ways. Other languages may support nonstrict features locally or provide optional lazy evaluation through libraries or language extensions.

In summary, nonstrictness describes a semantic and practical approach to evaluation that defers computation until necessary, enabling certain abstractions and efficiencies while demanding careful attention to termination and side effects.