#rules #business #rete #brms

deprecated nightly bin+lib expert

A cache-friendly, typesafe expert/business rules system

1 unstable release

Uses old Rust 2015

0.0.1 Nov 22, 2017

#35 in #business

MIT/Apache

110KB
2K SLoC

expert-rs

A cache friendly and type safe expert/business rules system written in Rust.

Usage

Don't. It's not even Alpha quality, yet.

Eventually you'll be able to add a dependency ala

[dependencies]
expert = "0.10"

And then write code similar to

#[macro_use]
extern crate expert;


// A fact type
#[derive(Debug, Copy, Clone, Eq, PartialEq, Hash, Fact)]
pub struct Fact1 {
    id: usize,
    len: usize,
}


// Another fact type
#[derive(Debug, Copy, Clone, Eq, PartialEq, Hash, Fact)]
pub struct Fact2 {
    id: usize,
    height: usize
}

mod example {
    // Creates a new expert system specifically for Fact1 and Fact2
    expert!(Example; [Fact1, Fact2]);
}

pub fn main() {
    use example::*;
    
    // Create a custom rule builder
    let mut builder = ExampleBuilder::new();
    builder.
        // Add a new rule named "rule"
        rule("rule")
            // When a Fact1 is inserted with self.id == 2 and self.len < 32
            .when<Fact1>().eq("id", 2usize).lt("len", 32usize)
            // And when a Fact2 is inserted with sef.id ==3 and self.height >= 64
            .when<Fact2>().eq("id", 17usize).gte("height", 64usize)
        .then()
            // Add a new Fact1{id: 3, len: 20} to be available for return
            .ret<Fact1>().field("id", 3usize).field("len", 20usize)
        .end();
        
   // Create a knowledge base to hold the read-only, in-memory representation of all of the rules
   let example_base = builder.build();
   
   // Create a knowledge session/rule runtime
   let mut example_session = example_base.new_session();
   
   let a_fact = Fact1{id: 2usize, len: 31usize};
   let another_fact = Fact2{id: 3usize, height: 60usize};
   
   // Insert the facts 
   example_session.insert(a_fact);
   example_session.insert(another_fact);
   
   // Resolve the rule consequences
   example_session.resolve();
   
   let fact1_returns: &[Fact1] = <Fact1 as ExmapleReturn>::get_returns(&example_session);
   assert(fact1_returns.len() == 1);
   let fact1_ret = fact1_returns.get(0).unwrap();
   assert(fact1_ret.id == 3);
   assert(fact1_ret.len == 20);
}

Features

  • The code compiles!
  • Alpha (fact analysis) nodes!
  • Statement Builder!

In-Progress

  • Rule Builder!
  • KnowledgeBase Builder!
  • Alpha, Beta node compilation and layout!
    • Networks represented as a set of arrays!
    • Alpha network ordered from most depended on nodes to least!
    • Beta network ordered from least depended to most!
  • Chaining!
    • Forwards!
    • Backwards! (w/ Iterators)
  • Returns, Insert within rule resolution
  • All of those sweet, sweet macros

Not In-Progress (everything else!)

  • Variables
    • Fact
    • Field
  • Complex logic ((this AND this AND NOT that) OR other)
  • DSL
  • Fast rule compilation times
  • Rule optimizations
  • Rule sanity checks
  • In-place fact modification

Why?

  • Expert/Business Rules Systems are still in use in a variety of industries
  • I wanted to understand how the algorithms worked
  • I wanted to push my software development expertise
  • I wanted to see if I could write it without pointers, fat or otherwise
  • It's hard

License

This project is licensed under either of

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in Serde by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.

Dependencies

~3.5MB
~71K SLoC