The Chez Scheme implementation of identifier properties and the more limited compile-time values requires the transformer to return a procedure which the expander in turn re-enters with a lookup (rho) procedure to find the compile-time value or the property value. Any helper procedure used by a transformer which has to look up this information has to be passed this procedure explicitly as well.
Meanwhile, the equivalent for compile-time values in Racket, syntax-local-value can simply be imported into a library. Guile has a similar feature, syntax-local-binding. Both are restricted to being called within macro transformers for obvious reasons.
Other than the fact that lookup/rho cannot be used outside a transformer call, and depends on state made available by the expander (which could be done using a parameter or some other dynamic state, for example), is there any reason for the design decision to require explicitly capturing the lookup procedure to access compile-time values and properties, rather than making an identifier-property-ref type procedure which can be in scope anywhere?
(I tried to add the label ‘question’ but couldn’t see how.)
The Chez Scheme implementation of identifier properties and the more limited compile-time values requires the transformer to return a procedure which the expander in turn re-enters with a
lookup(rho) procedure to find the compile-time value or the property value. Any helper procedure used by a transformer which has to look up this information has to be passed this procedure explicitly as well.Meanwhile, the equivalent for compile-time values in Racket,
syntax-local-valuecan simply be imported into a library. Guile has a similar feature,syntax-local-binding. Both are restricted to being called within macro transformers for obvious reasons.Other than the fact that
lookup/rhocannot be used outside a transformer call, and depends on state made available by the expander (which could be done using a parameter or some other dynamic state, for example), is there any reason for the design decision to require explicitly capturing thelookupprocedure to access compile-time values and properties, rather than making anidentifier-property-reftype procedure which can be in scope anywhere?(I tried to add the label ‘question’ but couldn’t see how.)