@@ 27,6 27,11 @@
;; since a fiber may be re-scheduled on a different thread. If anyone has a low-cost fix
;; for this i'd love to hear about it.
+;; TODO: Evaluate using RRB-trees for the fectors. This would allow efficient appends and splices
+;; at very little cost for guile 3 (which can generate fast code paths). RRB-trees are a
+;; strict superset of the data structure used here, but with relaxed rules about how many
+;; elements in a node. What this means for fectors is: more expensive lookups AFTER a splices
+;; or append has been made.
;; TODO: why is fector-ref so much slower than vlist-ref?
;; have a look at immer pvectors to see if there is anything they are doing except being
@@ 34,10 39,15 @@
;; TODO: Rewrite (cond ... (else (cond ...)).
;; TODO: Should we allow for regular fectors in transient procedures with implicit conversion?
;; It is comfortable, but just mixing it is bad style.
+;; TODO: Make a fector-generator that will provide a nice way to iterate through the fector
+;; that also has a configurable start and end. This could be usable when writing
+;; procedures like fector-copy that allows for subfectors to be copied.
#:use-module (srfi srfi-9)
+ #:use-module ((srfi srfi-9 gnu) #:select (set-record-type-printer!))
#:use-module (ice-9 atomic)
#:use-module (ice-9 match)
+ #:use-module (ice-9 threads)
@@ 685,3 695,13 @@
(if (= -1 index)
(loop (1- index) (+ sum (fector-ref fec index))))))
+(define (display-fector f port)
+ (if (transient-fector? f)
+ (display "#transient-fector")
+ (display "#fector"))
+ (display (fector->list f) port))
+(set-record-type-printer! <fector> display-fector)
+(set-record-type-printer! <transient-fector> display-fector)