Method eval

hyperopt/pyll/base.py:255–284

Recursively evaluate an expression graph. This method operates directly on the graph of extended inputs to this node, making no attempt to modify or optimize the expression graph. Caveats: * If there are nodes in the graph that do not represent expressio

(self, memo=None)

            scope.define_if_new(**self.define_params)

    def eval(self, memo=None):
        """
        Recursively evaluate an expression graph.

        This method operates directly on the graph of extended inputs to this
        node, making no attempt to modify or optimize the expression graph.

        Caveats:

          * If there are nodes in the graph that do not represent expressions,
            (e.g. nodes that correspond to statement blocks or assertions)
            then it's not clear what this routine should do, and you should
            probably not call it.

          * If there are Lambdas in the graph, this procedure will not evluate
            them -- see rec_eval for that.

        However, for many cases that are pure expression graphs, this
        offers a quick and simple way to evaluate them.
        """
        if memo is None:
            memo = {}
        if id(self) in memo:
            return memo[id(self)]
        else:
            args = [a.eval() for a in self.pos_args]
            kwargs = {n: a.eval() for (n, a) in self.named_args}
            f = scope._impls[self.name]
            memo[id(self)] = rval = f(*args, **kwargs)
            return rval

    def inputs(self):
        # -- this function gets called a lot and it's not 100% safe to cache