metrics_evaluation

metrics_evaluation is a simple text parser that evaluates a given logic against a Resolver (which resolves names to values) resulting in a single bool-result. It is possible to perform simple arithmetic operations on the values before the comparison takes place.

It can be used for example to evaluate named IoT-metrics against a given logic to trigger or not to trigger an action if the result of evaluate results to true. The format of the input is equal to the if-style of rust (no initial braces needed). Arithmetics are limited to + (add), - (sub), * (mul) and / (div) at the moment.

It is possible to evaluate comparisons of variables against fixed values or variables against other variables. Anyway comparisons are limited to always have a variable-name on the left hand at the moment (WIP).

Usage

Add this crate to your cargo.tml

use metrics_evaluation::{parse_tree, solve_tree, Resolver};

Implement a Resolver that can deliver a Value-reference for the give variable-name-lookup (or none if there is no value for this). For the ease of use and pure laziness there is a MapResolver wich is a Resolver and can be formed from HashMap<K: AsRef<str>, V: Into<Value>> (see tests).

Give this Resolver to the evaluate function and call it with a simple text-evaluation as you would do to check if a given value evaluates to true|false in rust. For example: foo + 2 == 42 && bar < 2 || (baz == true && baz + "30sec" >= "42min").

The following operators are supported and behave like in rust:

• >
• <
• >=
• <=
• ==
• !=

The following logics are supported and behave like in rust:

• and, &&
• or, ||

The following arithmetic operators are supported but differ in implementation for different Value-Types:

• + (add)
• - (sub)
• * (mul)
• / (div)

The following [Value]s can be compared:

• Value::Numeric - maps internally to a f64 and has From-implementations ranging from u8 to f64
• Value::String - a string literal which must be always encapsulated by quotation marks.
• Value::Time - maps a NaiveTime and the string-representation must match "%H:%M:%S". Must be always encapsulated by quotation marks.
• Value::Duration a string that is a humantime representing a duration. Must be always encapsulated by quotation marks.

The general form of a comparison is [Name] [Arithmetic] [Comparison-Operator] Value [Arithmetic] [[Logic]...].

Feature-flags

• async - additionally have [AsyncResolver] and [AsyncSolver] over [Resolver] and [Solver] for cases a [Resolver] needs async functionality (async database for example). Use evaluate_async' in this case. MapResolver` is only available in test-configuration here (as it makes no sense to have such in production).

Easy example

See solver tests (end of file) or use it like this:

use std::collections::HashMap;

use anyhow::Result;
use metrics_evaluation::{evaluate, MapResolver, Value};

fn main() -> Result<()> {
    let mut values = HashMap::new();
    values.insert("room1.temp", Value::Numeric(20.0));
    values.insert("room2.temp", Value::Numeric(22.0));
    values.insert("room1.humidity", Value::Numeric(80.0));
    values.insert(
        "worktime",
        Value::Duration(*"1h 5min 42s".parse::<humantime::Duration>().unwrap()),
    );
    let values: MapResolver = values.into();

    assert_eq!(
        evaluate(
            r#"(room1.temp > 1 || room1.humidity <= 80) && worktime > "1h 5min""#,
            &values
        )?,
        true
    );

    assert_eq!(evaluate(r#"worktime + "30m" > "1h""#, &values)?, true);

    assert_eq!(
        evaluate(r#"room2.temp > room1.temp && room2.temp < 25"#, &values)?,
        true
    );

    assert_eq!(evaluate(r#"room1.temp + 2 == room2.temp"#, &values)?, true);

    assert_eq!(
        evaluate(r#"room2.temp - 1  < room1.temp * 2"#, &values)?,
        true
    );

    Ok(())
}

License

metrics_evaluation is distributed under the terms the MIT license.

See LICENSE for details.