infradig
commented
2 years ago Yes it needs a call-in API. Will look into it...
On Tue, 27 Sept 2022, 15:41 guregu, @.***> wrote:
I'm working on the last major feature for the WASM ports: returning one query result at a time instead of findall'ing the query. Ideally we could just use stdin and wait for some input to continue the query, but unfortunately the majority of WASM runtimes don't implement stdin as a stream, so this approach doesn't work. I was considering doing something clever with history_getch as an exported host function but I think that'll just lead it to block forever in the browser.
Is there a way to control the redo process for queries with the current API?
I am imagining an API kind of like this:
// works like pl_eval, except returns a handle to the query instead of waiting for toplevel input query pl_query(prolog, const char expr);// redo a given query, returning false if there are no more choice pointsbool pl_redo(prolog, *query);
And the WASM host would call it like so:
class Prolog { ptr; // *prolog ptr
// example: javascript generator function that returns one result at a tinme *query(expr) { // eval once and grab ptr to query handle const query = this.pl_query(this.ptr, expr); do { const answer = /* read from stdin */; yield answer; } while (this.pl_redo(this.ptr, query)) }}// usageconst query = pl.query("member(X, [1,2,3]).");for (const answer of query) { console.log(answer) // X=1, then X=2 next iteration, etc}
Ideally these would be silent, not dumping vars or printing prompts so it's easy to parse the output. We can use Prolog code to marshal the answers to an easy-to-parse format, so we don't need a C results API, just a way to run redos.
— Reply to this email directly, view it on GitHub https://github.com/trealla-prolog/trealla/issues/46, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFNKSEU7GIBYFRJCWUXNMOTWAKCJLANCNFSM6AAAAAAQWNSJMY . You are receiving this because you are subscribed to this thread.Message ID: @.***>
guregu
I'm working on the last major feature for the WASM ports: returning one query result at a time instead of findall'ing the query. Ideally we could just use stdin and wait for some input to continue the query, but unfortunately the majority of WASM runtimes don't implement stdin as a stream, so this approach doesn't work. I was considering doing something clever with
history_getchas an exported host function but I think that'll just lead it to block forever in the browser.Is there a way to control the redo process for queries with the current API?
I am imagining an API kind of like this:
And the WASM host would call it like so:
Ideally these would be silent, not dumping vars or printing prompts so it's easy to parse the output. We can use Prolog code to marshal the answers to an easy-to-parse format, so we don't need a C results API, just a way to run redos.