A Scheme take on the environment (reader) monad, focusing more on being efficient and convenient than pure.
nameis used for description and may or may not be bound to a value representing the monad. All other parameters are keywords, and with the exception of
fieldssimply provide binding names for the monad operators described below. The
fields:keyword takes a list of field name identifiers known to be used by the monad. This is an optimization hint, as the monad can be used to store and query values for any identifier at runtime. The following keywords obey the definition of a monad: sequence: sequence (>>) - Essentially a semi-colon, this joins two operations together. bind: (>>=) - Runs a normal function. As a syntactic convenience,
bindlooks and behaves like a lambda, but the parameters of the
bindare bound as Scheme variables with the values of the corresponding environment variables. Thus you fetch the values of foo and bar with:
(bind (foo bar) ...)hiding the need for an explicit
ask. If you want to bind the values to some other name, you can use it like a
(bind ((my-foo foo) (my-bar bar)) ...)return: Returns a pure (non-monadic) value. run: Start the monad. The following are specific to the environment monad: ask: Ask the current value of an environment variable. This is not meant to be used directly - use the `bind' syntax to query bindings. local: Shadow the value one or more environment variables, analogous to `let'. In addition, support for optional mutation is provided: local!: (local! (var val) ...) will update the environment with the corresponding variable bindings. In a sequence, successive operations will see the result of the update, unlike with `local'. This is allowed, but not required, to perform mutation. bind-fork:
(bind-fork a b)runs `a' followed by `b', passing `b' the original state before `a' was run.