Marco Antoniotti wrote:
I would like to provide an ITERATE extension to leverage the ENUMERATION instances. I went through the manual but I am a little unclear about how to user DEFMACRO-DRIVER etc. Could anybody help me?
Sure, send this to the list.
I admit that I'm also confused by the proliferation of terms and lack of definitions (or hard to find definitions in flowtext), e.g. defclause defclause-driver defmacro-driver defmacro-clause defclause-sequence
What's a clause? what's driver? etc.
My finding so far is that a driver comes into play when FOR can be replaced by GENERATE. defmacro-* is best suited for users writing extensions, drawing from the macro concept, i.e. the body expands to other iterate clauses and CL forms.
C.f. iterate-test.lisp
I'll make a note that the manual should provide a glossary, or at least make the index work -- where does a TeX .ind file come from?
Also unclear to me is top-level vs. anywhere forms. Why are some clauses (mostly FOR, but not all of them) restricted to the top-level?
As an exercise, I just looked at iterate-pg.lisp. It appears that it cannot currently be written using only names exported from the ITERATE package. The only thing lacking is an interface to add-loop-body-wrapper for the sake of with-pg-cursor. Other than that, it could be written like
(defmacro-driver (FOR var-spec IN-RELATION query ON-CONNECTION conn) (add-loop-body-wrapper `(with-pg-cursor ,cursor ,conn ,query)) `(for ,var-spec = (first (or (pg-result #) (finish)))) or `(for/generate ,var-spec next (first (or (pg-result #) (terminate)))) )
Alternatively, with-pg-cursor might be "inlined" into the expansion using iterate:finally-protected and with clauses. That would also prevent double evaluation of CONN.
I recommend that you stick to the manual as long as possible and write a macro-clause or -driver. Don't look at the internals, don't use return-driver-code, make-binding etc.
Actually, it may be a worthwhile exercise to rewrite some part of Iterate in a system construction manner, building drivers on top of others. It could make code shorter and possibly easier to maintain. E.g. FOR ... IN could be written as follows: (defmacro-driver (FOR var IN list &optional BY (step ''cdr)) (let ((for/gen (if generate 'generate 'FOR))) `(,for/gen (,var) ON ,list BY ,step))) and the expansion is only slightly worse than now (one temporary let)
Regards, Jorg Hohle