With the new library(hashtable) and the following code: hashtable_gen(N) :- ht_new(HT), forall( between(1,N,I), ( ht_put(HT,I,I) ) ), assertz(hashtable(HT)). hashtable_lookup(N) :- hashtable(HT), forall( between(1,N,I), ( ht_get(HT,I,I) ) …

[image] swi: EDIT: Is there a predicate to completely remove the hashtable? i.e. the opposite of ht_new/1 ? You shouldn’t need to do that. SWI-Prolog has garbage collection. If you really insist, you could probably do something like: setarg(1, HT, 0), setarg(2, HT, 0), setarg(3, HT, [](_)).…

[image] Boris: ht_new(HT), \+ ( \+ ht_put(HT, a, 1) ). Thanks, that clears it up for me; although I didn’t expect it givent that HTs are mutable structures.

[image] anon95304481: Can library(hashtable) be used together with concurrent_forall/2 ? Like in read-only mode? It is just a Prolog term like all others and thus, when using threads, any thread gets a copy and any subsequent modification in the thread only applies to this specific copy.

[image] swi: although I didn’t expect it givent that HTs are mutable structures. Mutable, but backtrackable. This is the difference between setarg/3 and nb_setarg/3 (“nb” stands for “non-backtrackable”). You should read the whole section probably, “Non-logical operations on terms” . I had mys…

It would be nice if there were a consistent set of predicates for hashtable dict assoc rbtree For example, rbtree lookup has the key as the 2nd arg; the others have it as the 1st. (I prefer it as the 2nd because that plays nicer with predicates like maplist/3). Also, each of these libraries has…

[image] peter.ludemann: benchmarks Database micro-benchmark

[image] jan: It is just a Prolog term like all others Ahh I see, this is what threw me off. I had a mental model that the hashtable was a blob object produced in C and it had to be “released” somehow.

[image] jan: library(yall) translates if the expression is found and translated by goal_expansion/2 and uses copy_term/2 otherwise. So, the semantics when dealing with destructive assignment in terms is rather poorly defined. Hash tables, being terms subject to destructive assignment, should be …

[image] peter.ludemann: Also: any benchmarks on hashtable versus assoc/rbtree? Dear @peter.ludemann , I have updated the micro-benchmark code to support hashtables and these are the results (for version 8.3.4-11-g1db629e24): Database raw and crude micro benchmark …

Interesting. I only compared the performance to the CHR hash tables. Here the result was close time-wise and much better memory wise. Hash tables should also compare positively to trees memory wise. Note that using -O (optimize) may help as hash tables require arithmetic while the others do n…

New hashtable library

Help! Predicate

EricGT July 24, 2020, 7:22pm 19

Database micro-benchmark

Topic		Replies	Views
Database micro-benchmark (Discussion) Discussion	22	1058	April 25, 2020
Performance cost of dictionaries? General	14	422	September 18, 2024
Fast Hash Tables Resources request	10	1529	August 18, 2019
Ann: SWI-Prolog 8.3.4 Releases	19	1240	July 20, 2020
Effective implementation of a set Help!	11	1541	February 26, 2020

New hashtable library

Related topics