Trait rustc_trait_selection::traits::normalize::NormalizeExt

pub trait NormalizeExt<'tcx> {
    // Required methods
    fn normalize<T: TypeFoldable<TyCtxt<'tcx>>>(
        &self,
        value: T
    ) -> InferOk<'tcx, T>;
    fn deeply_normalize<T: TypeFoldable<TyCtxt<'tcx>>>(
        self,
        value: T,
        fulfill_cx: &mut dyn TraitEngine<'tcx>
    ) -> Result<T, Vec<FulfillmentError<'tcx>>>;
}

Required Methods

fn normalize<T: TypeFoldable<TyCtxt<'tcx>>>(&self, value: T) -> InferOk<'tcx, T>

Normalize a value using the AssocTypeNormalizer.

This normalization should be used when the type contains inference variables or the projection may be fallible.

fn deeply_normalize<T: TypeFoldable<TyCtxt<'tcx>>>( self, value: T, fulfill_cx: &mut dyn TraitEngine<'tcx> ) -> Result<T, Vec<FulfillmentError<'tcx>>>

Deeply normalizes value, replacing all aliases which can by normalized in the current environment. In the new solver this errors in case normalization fails or is ambiguous.

In the old solver this simply uses normalizes and adds the nested obligations to the fulfill_cx. This is necessary as we otherwise end up recomputing the same goals in both a temporary and the shared context which negatively impacts performance as these don’t share caching.

FIXME(-Znext-solver): This has the same behavior as traits::fully_normalize in the new solver, but because of performance reasons, we currently reuse an existing fulfillment context in the old solver. Once we also eagerly prove goals with the old solver or have removed the old solver, remove traits::fully_normalize and rename this function to At::fully_normalize.

Object Safety

This trait is not object safe.

Implementors

impl<'tcx> NormalizeExt<'tcx> for At<'_, 'tcx>