Programatic use of Machine / Scryer as library #1880

lucksus · 2023-07-12T10:24:25Z

These changes make it possible to run Prolog queries on a Machine instance in a 3rd-party crate that pulls-in Scryer as a library dependency, interfacing with the machine only with in-memory Strings.

Overview

Machine::run_input_once(&mut self) runs the added toplevel predicate run_input_once/0 which reads the query from user input like repl but just executes it once, returning/printing all results
Machine::load_module_string(&mut self, module_name: &str, program: String) creates a stream from the given string and loads like a file
Machine::set_user_input(&mut self, input: String) and Machine::get_user_output(&self) -> String enabling interfacing with the machine
Machine::run_query(&mut self, query: String) -> QueryResult wraps it all up by setting the query as input and parsing the output string into added Rust type QueryResult.

Example usage

let mut machine = Machine::new_lib();

machine.load_module_string("facts", String::from(r#"
    triple("a", "p1", "b").
    triple("a", "p2", "b").
"#));

let output: QueryResult = machine.run_query(String::from(r#"triple("a",P,"b")."#));

assert_eq!(output, QueryResult::Matches(vec![
    QueryMatch::from(btreemap!{
        "P" => Value::from("p1"),
    }),
    QueryMatch::from(btreemap!{
        "P" => Value::from("p2"),
    }),
]));

Next steps

ensure parsing of complex result values
consult/1 like updating of machine state

…tput() -> String. Make read_term_from_user_input() handle Stream::Byte.

…d runs it

…n_query()

…ults.rs

triska · 2023-07-12T20:10:10Z

Very interesting work and contribution, thank you a lot for working on this!

The only suggestion I have is that it would be awesome, and I think preferable, to generalize the currently existing function run_top_level so that an arbitrary goal can be invoked as the Scryer Prolog toplevel. Currently, '$repl' is hardcoded:

scryer-prolog/src/machine/mod.rs

Line 307 in 112d398

self.run_module_predicate(atom!("$toplevel"), (atom!("$repl"), 1));

Generalizing this function so that an arbitrary goal can be specified could let you avoid the introduction of the new and seemingly rather ad hoc definition of run_input_once in 112d398. Because: What if some other library user would prefer yet another toplevel instead, such as your "run once" with slight variations, or something else entirely?

Personally, I would prefer to not add such additional toplevels to toplevel.pl, but to let every program that needs specialized toplevels to flexibly specify the Prolog code and goal that should act as the toplevel.

lucksus · 2023-07-17T19:58:42Z

Ok, I've generalized run_toplevel() to take module and key of the predicate to run as parameters. But to enable specifying the whole toplevel file by the user of Machine, I also had to generalize load_top_level() and the constructor. For that I added MachineConfig with default implementation and also reduced copy/paste code in Machine::with_test_streams() to use the new config pattern.

I've also extracted the lib query functions and test into their own file. Also added a new lib_toplevel.pl that is only used there. But I ended up copying over almost all of toplevel.pl. Not sure if this is better than just adding a few predicates to toplevel.pl. But this is definitely more generic and more easily applicable to other similar use-cases.

lucksus · 2023-07-17T20:00:40Z

BTW, I was basing these changes off of the v0.9.1 commit because the latest head with updated tokio dependency creates problems when used in my project which also pulls in Deno with specific dependencies which work with that old revision of scryer but not the latest head.

triska · 2023-07-21T16:35:54Z

If user-interaction is not needed, a simpler toplevel could suffice. For instance, here is a very rudimentary toplevel that may suffice for your use case:

toplevel :-
        read_term(Goal, [variable_names(VNs)]),
        Goal,
        write('bindings(['),
        write_bindings(VNs),
        write(']).'),
        nl,
        false.
toplevel :- toplevel.

write_bindings([]).
write_bindings([VN|VNs]) :-
        write_bindings_(VNs, VN).

write_bindings_([], VN) :-
        write_binding(VN).
write_bindings_([VN|VNs], Prev) :-
        write_binding(Prev),
        write(','),
        write_bindings_(VNs, VN).

write_binding(Var=Val) :-
        write(Var),
        write(=),
        write_term(Val, [quoted(true),double_quotes(true)]).

Sample interaction:

?- toplevel.
X=3.
bindings([X=3]).
member(X, "abc").
bindings([X=a]).
bindings([X=b]).
bindings([X=c]).

Add halt/0 or emitting no_more_solutions. as needed!

Such a rudimentary toplevel may be provided by the application itself.

…f write_eq to avoid truncation of results

triska · 2023-10-23T18:16:53Z

If I read the CI logs correctly, only the build for the (comparatively dated) 20.04 version of Ubuntu fails, and only for the 32-bit version? If that is the case, then I do not consider the single failing CI build a justification for any delay with merging this PR: ~~As mentioned, #2126 has the same issue, and~~ all other platforms, including the more recent Ubuntu 22.04, seem to build correctly.

Seconding #1880 (comment), I would recommend to first and foremost get the API right, so that (as mentioned above) answers can be accessed by an iterator, because then also infinite sequences of answers can be processed in programs that use this API. (Think of ?- length(Ls, L). etc.).

Skgland · 2023-10-23T18:33:20Z

This failure and #2126 appear unrelated to me.

Bump rustix from 0.38.14 to 0.38.19 #2126 had a network request fail in the Publish cargo test summary step Request POST /repos/mthom/scryer-prolog/check-runs failed with 403: Forbidden
This failed in the Build wasm step as a compiler process got a sigkill (signal: 9, SIGKILL: kill)

Both of these two steps appear to only run for the ubuntu-20.04 x86_64 job, so other jobs can't fail on these steps.

mthom · 2023-10-25T18:43:31Z

SIGKILL usually happens because the OS has run out of memory. Is anyone able to reproduce the issue on their machine after pulling the branch with wasm-pack build --target web -- --no-default-features?

I measured memory usage of rustc on my machine invoked via wasm-pack and at its peak it used 18.4% of system memory, which on my (64 GB) machine, is around 11 - 12 GBs.

Skgland · 2023-10-25T21:57:21Z

SIGKILL usually happens because the OS has run out of memory. Is anyone able to reproduce the issue on their machine after pulling the branch with wasm-pack build --target web -- --no-default-features?

~~I was not able to reproduce it on my Laptop (16GB).~~

Was on the master branch instead of the PR branch.
With this PR via git fetch upstream pull/1880/head:pr1880 I get a (signal: 15, SIGTERM: termination signal) on my Laptop (16GB).

infogulch · 2023-10-25T22:01:31Z

Notably, the github actions runners have 7GB of memory:

https://docs.github.com/en/actions/using-github-hosted-runners/about-github-hosted-runners/about-github-hosted-runners#supported-runners-and-hardware-resources

rujialiu · 2023-10-26T02:10:12Z

I'm not familiar with this part of rust, but is it possible to reduce the number of inlines during compilation? There seems to be a loooot of them.

infogulch · 2023-10-26T04:37:27Z

fwiw, I tried compiling with -j 1 which compiles serially instead of in parallel when possible, but it didn't change the memory usage (which peaked around 6GB on my system with this methodology), it just took longer to complete.

# I ran this in another terminal during compilation. It's not scientific at all, but I can't be bothered to try harder when the results are so clear already.
while true; do smem -t -P '(cargo|rustc) ' | tail -1 | awk '{print $5}' >> log.txt; sleep 0.1s; done

infogulch · 2023-10-26T05:17:51Z

Ok I made some changes on my fork so the CI gets farther, but there are still some issues which I think are related to platform-specific code that doesn't work in wasm32.

In particular, running cargo test on the wasm target fails:

$ cargo test --target wasm32-unknown-unknown --no-default-features --all --verbose
...
error[E0433]: failed to resolve: use of undeclared crate or module `imp`
  --> /home/joe/.cargo/registry/src/index.crates.io-6f17d22bba15001f/wait-timeout-0.2.0/src/lib.rs:66:9
   |
66 |         imp::wait_timeout(self, dur)
   |         ^^^ use of undeclared crate or module `imp`

My guess is that some tests need to be excluded for the wasm target?

infogulch · 2023-10-26T20:37:45Z

Changes to CI in #2137 may have fixed the build issues here. (Note: a new build will be triggered if you push any commit or amend the last commit and force push it.)

While developing that pr I found that tests are failing for the wasm build and it probably needs some conditional build tags on tests. See this issue for details:

Fix failing tests on wasm target #2138

mthom · 2023-10-26T20:58:26Z

I can also re-run it as the repo owner, which I'm doing now.

Skgland · 2023-10-26T21:02:43Z

I can also re-run it as the repo owner, which I'm doing now.

Wouln't it need to be rebased on master/merged with master first, as CI is testing the state of the PR not the merged state?

triska · 2023-10-26T21:04:52Z

I think the failing build on ubuntu-20.04 nightly is due to f80dff8 not yet being applied in this branch.

infogulch · 2023-10-26T21:15:48Z

I can also re-run it as the repo owner, which I'm doing now.

Wouln't it need to be rebased on master/merged with master first, as CI is testing the state of the PR not the merged state?

A push to the branch is handled differently than manually triggering a rebuild. Specifically, if triggered by the pull_request: trigger then the branch is tested after being merged with the target, but other triggers (including manual) will build the branch as-is. See: actions/checkout#15

mthom · 2023-10-26T21:25:13Z

I can also re-run it as the repo owner, which I'm doing now.

Wouln't it need to be rebased on master/merged with master first, as CI is testing the state of the PR not the merged state?

Right, of course. My enthusiasm for merging this PR got in the way of my sense.

lucksus · 2023-11-02T10:46:34Z

I think the failing build on ubuntu-20.04 nightly is due to f80dff8 not yet being applied in this branch.

I've merge master in, which includes that commit. Still failing...

Skgland · 2023-11-02T11:15:58Z

I think the failing build on ubuntu-20.04 nightly is due to f80dff8 not yet being applied in this branch.

I've merge master in, which includes that commit. Still failing...

Different Error though, now in step Build release binary:

error[E0599]: no function or associated item named `new_multi_thread` found for struct `tokio::runtime::Builder` in the current scope
  --> src/bin/scryer-prolog.rs:19:44
   |
19 |     let runtime = tokio::runtime::Builder::new_multi_thread()
   |                                            ^^^^^^^^^^^^^^^^
   |                                            |
   |                                            function or associated item not found in `Builder`
   |                                            help: there is an associated function with a similar name: `new_current_thread`

For more information about this error, try `rustc --explain E0599`.
warning: `scryer-prolog` (bin "scryer-prolog") generated 1 warning
error: could not compile `scryer-prolog` (bin "scryer-prolog") due to previous error; 1 warning emitted

instead of target_os = “wasi”

mthom · 2023-11-02T19:46:38Z

Great! If there are no further comments I'll merge and we'll have a new release out in time for the meetup.

lucksus · 2023-11-02T22:35:46Z

Great! If there are no further comments I'll merge and we'll have a new release out in time for the meetup.

🎉 🥳 🚀

(I didn't get to changing run_query() to return an iterator, but don't think that should block a merge - can be added in a separate, small PR)

bakaq · 2023-11-03T01:53:51Z

(I didn't get to changing run_query() to return an iterator, but don't think that should block a merge - can be added in a separate, small PR)

Keep in mind what I said about breaking changes before. If the current version gets on crates.io, which is the next logical step, it will have to follow the Rust variant of semver. Changing the signature of run_query() is a breaking change, so if the published version is 0.9.3, after this change it will have to be 0.10.0 instead of 0.9.4. A way to avoid this is to introduce a different run_query_iter() function instead, which would only require a minor version bump, and can also be used to define run_query() if it seems appropriate. Maybe bumping the major version is not such a big deal, but ideally the versions of the binary and the library are synchronized (it's literally the same codebase), so bumping the major version of the library would mean also bumping the version of the binary, which would probably be expected by users to be a big update instead of some random "internal" interface change.

infogulch · 2023-11-04T22:16:43Z

I think this could be a good way to do benchmarking with scryer-prolog. See discussion: #1782 With iai ¹, which measures performance using Cachegrind by counting instructions and cache misses, this could be viable to set up in CI.

Thoughts?

Example:

benches/edges.rs:

use iai::{black_box, main};

fn iai_benchmark_edges(n: u64) -> u64 {
    let mut machine = Machine::new_lib();
    
    machine.load_module_string("facts", String::from(r#"
      :- use_module(library(clpb)).
      :- use_module(library(assoc)).
      :- use_module(library(lists)).
      
      /* - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
         Contiguous United States and DC as they appear in SGB:
         http://www-cs-faculty.stanford.edu/~uno/sgb.html
      - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - */
      
      edge(al, fl).
      edge(al, ga).
      edge(al, ms).
      edge(al, tn).
      edge(ar, la).
      edge(ar, mo).
      edge(ar, ms).
      edge(ar, ok).
      edge(ar, tn).
      edge(ar, tx).
      edge(az, ca).
      edge(az, nm).
      edge(az, nv).
      edge(az, ut).
      edge(ca, nv).
      edge(ca, or).
      edge(co, ks).
      edge(co, ne).
      edge(co, nm).
      edge(co, ok).
      edge(co, ut).
      edge(co, wy).
      edge(ct, ma).
      edge(ct, ny).
      edge(ct, ri).
      edge(dc, md).
      edge(dc, va).
      edge(de, md).
      edge(de, nj).
      edge(de, pa).
      edge(fl, ga).
      edge(ga, nc).
      edge(ga, sc).
      edge(ga, tn).
      edge(ia, il).
      edge(ia, mn).
      edge(ia, mo).
      edge(ia, ne).
      edge(ia, sd).
      edge(ia, wi).
      edge(id, mt).
      edge(id, nv).
      edge(id, or).
      edge(id, ut).
      edge(id, wa).
      edge(id, wy).
      edge(il, in).
      edge(il, ky).
      edge(il, mo).
      edge(il, wi).
      edge(in, ky).
      edge(in, mi).
      edge(in, oh).
      edge(ks, mo).
      edge(ks, ne).
      edge(ks, ok).
      edge(ky, mo).
      edge(ky, oh).
      edge(ky, tn).
      edge(ky, va).
      edge(ky, wv).
      edge(la, ms).
      edge(la, tx).
      edge(ma, nh).
      edge(ma, ny).
      edge(ma, ri).
      edge(ma, vt).
      edge(md, pa).
      edge(md, va).
      edge(md, wv).
      edge(me, nh).
      edge(mi, oh).
      edge(mi, wi).
      edge(mn, nd).
      edge(mn, sd).
      edge(mn, wi).
      edge(mo, ne).
      edge(mo, ok).
      edge(mo, tn).
      edge(ms, tn).
      edge(mt, nd).
      edge(mt, sd).
      edge(mt, wy).
      edge(nc, sc).
      edge(nc, tn).
      edge(nc, va).
      edge(nd, sd).
      edge(ne, sd).
      edge(ne, wy).
      edge(nh, vt).
      edge(nj, ny).
      edge(nj, pa).
      edge(nm, ok).
      edge(nm, tx).
      edge(nv, or).
      edge(nv, ut).
      edge(ny, pa).
      edge(ny, vt).
      edge(oh, pa).
      edge(oh, wv).
      edge(ok, tx).
      edge(or, wa).
      edge(pa, wv).
      edge(sd, wy).
      edge(tn, va).
      edge(ut, wy).
      edge(va, wv).
      
      pairs_keys_values([], [], []).
      pairs_keys_values([A-B|ABs], [A|As], [B|Bs]) :-
              pairs_keys_values(ABs, As, Bs).
      
      independent_set(*(NBs)) :-
              findall(U-V, edge(U, V), Edges),
              setof(U, V^(member(U-V, Edges);member(V-U, Edges)), Nodes),
              pairs_keys_values(Pairs, Nodes, _),
              list_to_assoc(Pairs, Assoc),
              maplist(not_both(Assoc), Edges, NBs).
      
      not_both(Assoc, U-V, ~BU + ~BV) :-
              get_assoc(U, Assoc, BU),
              get_assoc(V, Assoc, BV).
    "#));
    
    let output: QueryResult = machine.run_query(String::from(r#"independent_set(Sat), sat_count(Sat, Count)."#));
    
    assert_eq!(output, QueryResult::Matches(vec![
        QueryMatch::from(btreemap!{
            "Count" => Value::from("217968"),
        }),
    ]));
}

iai::main!(iai_benchmark_edges);

https://bheisler.github.io/criterion.rs/book/iai/iai.html ↩

alexpetros · 2023-11-05T22:07:59Z

I think I just discovered this too early before you've had the chance to write it, but as far as I can tell there's no reference to the possibility of using this as an embedded library on the README or the documentation (or this PR). Want to just drop a note of in support of promoting this use-case because it's exactly what I came here looking for, and I think it's a killer feature!

lucksus added 9 commits July 11, 2023 14:22

Add Machine::set_user_input(&mut self, input: String) and get_user_ou…

95b3114

…tput() -> String. Make read_term_from_user_input() handle Stream::Byte.

Add Machine::run_input_once() which reads one goal from user input an…

112d398

…d runs it

Add convenience methods Machine::load_module_string() and Machine::ru…

2f45f0c

…n_query()

Make run_input_once/0 match and print all results

703efdb

Parsed QueryResult

568abef

Remove debug println!s

3347f83

Refactor result parsing to idiomatic Rust and extract into parsed_res…

f324c95

…ults.rs

Fix build warnings

f656758

Add test for programatic queries

c0dd94c

triska mentioned this pull request Jul 12, 2023

scryper-prolog as a library #225

Closed

lucksus added 5 commits July 17, 2023 21:34

WIP: refactor to generalize Machine::run_top_level()

5f8cc3c

fmt machine/parsed_results.rs

e7f1e32

Use lib constructor in lib tests

3947390

Add back all needed predicates to lib_toplevel.pl

bb95ed3

Add back newline at end of toplevel.pl

644559b

lucksus added 5 commits July 20, 2023 21:34

Don't panic when parsing results fails

836f6c1

Error handling

cae32d6

type QueryResult = Result<QueryResolution, String>

7c93450

Add missing write_goal/3 to lib_toplevel.pl

0b833bd

Fix result parsing for complex string results

9e85be1

lucksus added 2 commits July 22, 2023 00:31

Add missing list_last_item to lib_toplevel.pl and increase MaxDepth o…

d8a9475

…f write_eq to avoid truncation of results

Dedupe machine results

9fd6e18

mthom force-pushed the master branch from 5bc3ca4 to bff48e7 Compare July 26, 2023 15:33

lucksus force-pushed the library-use-case branch from 625c2f0 to 9fd6e18 Compare July 27, 2023 08:57

lucksus added 2 commits July 27, 2023 11:41

Add consult that works with streams / strings in library use-case

2f99bb0

Add special case when parsing

0d28404

Merge branch 'master' into library-use-case

ee1bd9e

Check for target_arch = “wasm32”

59264c0

instead of target_os = “wasi”

mthom merged commit 575245c into mthom:master Nov 2, 2023
11 checks passed

lucksus mentioned this pull request Nov 7, 2023

Fixing run_query() #2150

Closed

infogulch mentioned this pull request Nov 9, 2023

Add benchmarks #2153

Merged

lucksus mentioned this pull request Jan 29, 2024

Iron-out edge cases for library use-case, adding extensive real-world test assertions #2309

Merged

bakaq mentioned this pull request Apr 26, 2024

How to use run_query in streaming mode/limiting number of matches? #2394

Open

This was referenced Aug 5, 2024

ISSUE-2464: Make scryer prolog usable as a shared library for other programs #2464

Open

Add run_query_iter() #2472

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Programatic use of Machine / Scryer as library #1880

Programatic use of Machine / Scryer as library #1880

lucksus commented Jul 12, 2023 •

edited

Loading

triska commented Jul 12, 2023

lucksus commented Jul 17, 2023

lucksus commented Jul 17, 2023

triska commented Jul 21, 2023 •

edited

Loading

triska commented Oct 23, 2023 •

edited

Loading

Skgland commented Oct 23, 2023 •

edited

Loading

mthom commented Oct 25, 2023

Skgland commented Oct 25, 2023 •

edited

Loading

infogulch commented Oct 25, 2023

rujialiu commented Oct 26, 2023

infogulch commented Oct 26, 2023

infogulch commented Oct 26, 2023

infogulch commented Oct 26, 2023 •

edited

Loading

mthom commented Oct 26, 2023

Skgland commented Oct 26, 2023

triska commented Oct 26, 2023

infogulch commented Oct 26, 2023

mthom commented Oct 26, 2023

lucksus commented Nov 2, 2023

Skgland commented Nov 2, 2023 •

edited

Loading

mthom commented Nov 2, 2023 •

edited

Loading

lucksus commented Nov 2, 2023

bakaq commented Nov 3, 2023

infogulch commented Nov 4, 2023 •

edited

Loading

alexpetros commented Nov 5, 2023

Programatic use of Machine / Scryer as library #1880

Programatic use of Machine / Scryer as library #1880

Conversation

lucksus commented Jul 12, 2023 • edited Loading

Overview

Example usage

Next steps

triska commented Jul 12, 2023

lucksus commented Jul 17, 2023

lucksus commented Jul 17, 2023

triska commented Jul 21, 2023 • edited Loading

triska commented Oct 23, 2023 • edited Loading

Skgland commented Oct 23, 2023 • edited Loading

mthom commented Oct 25, 2023

Skgland commented Oct 25, 2023 • edited Loading

infogulch commented Oct 25, 2023

rujialiu commented Oct 26, 2023

infogulch commented Oct 26, 2023

infogulch commented Oct 26, 2023

infogulch commented Oct 26, 2023 • edited Loading

mthom commented Oct 26, 2023

Skgland commented Oct 26, 2023

triska commented Oct 26, 2023

infogulch commented Oct 26, 2023

mthom commented Oct 26, 2023

lucksus commented Nov 2, 2023

Skgland commented Nov 2, 2023 • edited Loading

mthom commented Nov 2, 2023 • edited Loading

lucksus commented Nov 2, 2023

bakaq commented Nov 3, 2023

infogulch commented Nov 4, 2023 • edited Loading

Footnotes

alexpetros commented Nov 5, 2023

lucksus commented Jul 12, 2023 •

edited

Loading

triska commented Jul 21, 2023 •

edited

Loading

triska commented Oct 23, 2023 •

edited

Loading

Skgland commented Oct 23, 2023 •

edited

Loading

Skgland commented Oct 25, 2023 •

edited

Loading

infogulch commented Oct 26, 2023 •

edited

Loading

Skgland commented Nov 2, 2023 •

edited

Loading

mthom commented Nov 2, 2023 •

edited

Loading

infogulch commented Nov 4, 2023 •

edited

Loading