Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.lang.javascript > #124407
| From | Mild Shock <janburse@fastmail.fm> |
|---|---|
| Newsgroups | comp.lang.javascript |
| Subject | Jaffar's Unification in Dogelog Player (Re: ANN: Dogelog Player 1.3.6 (Canonical Compare)) |
| Date | 2025-09-23 16:21 +0200 |
| Message-ID | <10auacq$qjhg$2@solani.org> (permalink) |
| References | <10877t8$cm50$3@solani.org> |
We are currently in our third iteration of unification for rational trees. What first begun as a separate routine inside library(math), has now recently become a novel take on the host language routine unify() inside the Dogelog Player Prolog system. We report how we replaced the union find map data structure inside unify() by some Prolog compound pointer swizzling as suggested by Jaxon Jaffar in 1984, arriving at a more competetive rational tree unification realization. Jaxon Jaffar (1984) already proposed Hydra test cases. They take very less memory but are very large when written. We conducted according tests with Dogelog Player and Scryer Prolog. For N = 16'384, 65'536, 261'144 the Rust Prolog has a minor advantage over the JavaScript Prolog, but interestingly losing it for N = 1'048'576. See also: Jaffar's Unification in Dogelog Player https://qiita.com/j4n_bur53/items/b0b754f02d88680d3249 Mild Shock schrieb: > Dear All, > > We are happy to announce a new edition > of the Dogelog Player: > > - Frozen Terms: > We extended our garbage collector marking bits > to Prolog compounds. Unlike Prolog variables, where > setting all bits is used to indicate change set > membership, we use it to indicate frozen objects. > This makes the garbage collector and certain > built-ins such as copy_term/2 etc.. aware of > program sharing (PS). > > - Marking Algorithms: > We now provide native implementations of (==)/2, > copy_term/2, etc.. with sharing and cycle detection. > We didn't deploy additional marking bits and/or > pointers inside the Prolog terms, instead used > additional datastructures, leaving the Prolog terms > untouched. The compare/3 implementation is > not a total order. > > - Canonical Compare: > The library(lists) has experimental predicates > term_decompose/3 and term_canonical/2, still written > in 100% Prolog. We showcase their usage in > library(sequence) and library(aggregate). The end- > user gets by default a structure compare, but can > also have a canonical compare, both being > total and natural orders. > > Have Fun! > > Jan Burse, http://www.herbrand.ai/ , 21.08.2025
Back to comp.lang.javascript | Previous | Next — Previous in thread | Next in thread | Find similar
ANN: Dogelog Player 1.3.6 (Canonical Compare) Mild Shock <janburse@fastmail.fm> - 2025-08-21 15:45 +0200
Jaffar's Unification in Dogelog Player (Re: ANN: Dogelog Player 1.3.6 (Canonical Compare)) Mild Shock <janburse@fastmail.fm> - 2025-09-23 16:21 +0200
Gap Buffers for Dogelog Player (Re: Jaffar's Unification in Dogelog Player) Mild Shock <janburse@fastmail.fm> - 2025-10-02 14:09 +0200
csiph-web