dyn_quantity 0.1.0

Representing physical quantities dynamically (i.e. via values, not via the type system)
Documentation

dyn_quantity

The strong type system of rust allows defining physical quantities as types - see for example the uom crate. This is very useful to evaluate the correctness of calculations at compile time. Sometimes however, the type of a physical quantity is not known until runtime - for example, when parsing a user-provided string. This is where this crate comes into play:

use std::str::FromStr;
use dyn_quantity::DynQuantity;

/*
Parse a string into a physical quantity. The string can contain simple 
mathematical operations as well as scientific notation. If there is no
operand specified between individual components (numbers or physical units),
multiplication is assumed. The resulting value is calculated while parsing.
*/
let quant = DynQuantity::<f64>::from_str("4e2 pi mWb / (2*s^3)^2").expect("valid");

// The SI value of the quantity is "4e2 * pi * 1e-3 / 2^2" => the 1e-3 stems 
// from the prefix "m" of "mWb". This equates to 0.1 * pi or roughly 0.31459.
assert!((quant.value - 0.31459) < 1e-5);

// The SI base units exponents of "Wb / (s^3)^2" are:
assert_eq!(quant.exponents.second, -8);
assert_eq!(quant.exponents.meter, 2);
assert_eq!(quant.exponents.kilogram, 1);
assert_eq!(quant.exponents.ampere, -1);
assert_eq!(quant.exponents.kelvin, 0);
assert_eq!(quant.exponents.mol, 0);
assert_eq!(quant.exponents.candela, 0);

The docstring of the from_str module provides a complete documentation of the available parsing syntax.

Overview

This crate is based around the DynQuantity struct, which represents a physical quantity at runtime via its numerical value and the exponents of the involved SI base units. The latter are fields of the struct UnitExponents, which in turn is a field of DynQuantity.

It is possible to perform basic arithmetic operations on this struct. While some operations such as multiplication, division and exponentiation are infallible, addition and subtraction require the unit exponents of both involved DynQuantity structs to be identical. This is checked at runtime:

use std::str::FromStr;
use dyn_quantity::DynQuantity;

let current = DynQuantity::<f64>::from_str("-1.5 A").expect("valid");
let voltage = DynQuantity::<f64>::from_str("-3 V").expect("valid");
let power = DynQuantity::<f64>::from_str("20 W").expect("valid");

// This works: current times voltage is infallible. The resulting unit is Watt,
// therefore the subtraction succeeds
let diff = power.clone().try_sub(&(current.clone() * voltage.clone())).expect("units are compatible");
assert_eq!(diff.value, 15.5);

// This does not work: current divided by voltage squared is infallible, but the
// resulting units are not compatible to power
let res = power.try_add(&(current / voltage).powi(2));
assert!(res.is_err());

Another special case is root calculation: Since unit exponents can only be integers, the exponents of the radicand ("input") need to be divisible by the degree without remainder:

use dyn_quantity::{DynQuantity, UnitExponents};

// Create a DynQuantity from its components.
let exponents = UnitExponents {
    second: 2,
    meter: -4,
    kilogram: 0,
    ampere: 0,
    kelvin: 0,
    mol: 0,
    candela: 0,
};
let quantity = DynQuantity::new(9.0, exponents);

// Succeeds, since all exponents can be divided by 2 without remainder:
let res = quantity.clone().try_nthroot(2).expect("succeeded");
assert_eq!(res.value, 3.0);

// Fails, since not all exponents can be divided by 4 without remainder: 
assert!(quantity.try_nthroot(4).is_err());

Integration with other crates

uom

The uom integration is gated behind the uom feature flag.

One of the main features of DynQuantity is its capability to bridge the gap between uom's Quantity type (units defined at compile time) and user-provided input where the units are only known at runtime. For example, a user-provided string can fallibly be parsed into a Length. This is a two-step operation, where the string is first parsed into a DynQuantity and then converted into a Length via TryFrom:

use std::str::FromStr;
use uom::si::{f64::{Length, Velocity}, length::meter};
use dyn_quantity::DynQuantity;

let input = "2 mm / s * 0.5 s";
let quantity = DynQuantity::<f64>::from_str(input).expect("valid");
let length: Length = quantity.clone().try_into().expect("valid");
assert_eq!(length.get::<meter>(), 0.001);

// Trying to convert quantity into a Velocity fails because the type does not
// match the unit exponents:
assert!(Velocity::try_from(quantity).is_err());

The reverse conversion from a Quantity to a DynQuantity is always possible via the From implementation.

serde

The serde integration is gated behind the serde feature flag.

A DynQuantity can be deserialized from its "natural" struct representation or directly from a string (by first deserializing into a string and then using the FromStr implementation). In addition, a couple of functions for usage with the deserialize_with field attribute are provided:

use serde::{Deserialize};
use uom::si::{f64::Length, length::meter};
use dyn_quantity::deserialize_quantity;
use indoc::indoc;

#[derive(Deserialize, Debug)]
struct LengthWrapper {
    #[serde(deserialize_with = "deserialize_quantity")]
    length: Length,
}

let ser = indoc! {"
---
length: 1200 mm
"};
let wrapper: LengthWrapper = serde_yaml::from_str(&ser).unwrap();
assert_eq!(wrapper.length.get::<meter>(), 1.2);

The deserialize_with module holds all available functions.

Architecture and building instructions

At its core, this crate offers the following functionality via the DynQuantity struct:

  • Performing simple arithmetic operations on quantities where the units are only known at runtime.
  • Conversion into statically-typed quantities (requires the uom feature to be enabled).
  • Serialization and deserialization, in case of the latter from multiple different representations (requires the serde feature to be enabled).
  • Parsing quantities at runtime from strings.

For the last feature, the logos crate is used within the sub-crate dyn_quantity_lexer. logos creates a high-performance lexer via a procedural macro at compile time. If dyn_quantity_lexer is simply statically compiled into the final binary, each dependent of dyn_quantity needs to compile the lexer anew, leading to very long compile times.

To circumvent this issue, this crate offers the possibility of compiling the parser (of which the lexer is a part of) into a separate static library once. The library is then simply linked into the final binaries of dependents, greatly reducing the required compile time. This is the default compilation strategy realized in build.rs. However, sometimes it might be necessary to directly compile the parser into the final binary. The paragraphs below contain information for each building strategy so the dependent of this crate can choose its preferred model.

Compiling parser and lexer directly into the final binary

This is the "standard" Rust way of handling dependencies: Compile all the source code (including parser and lexer) into one single binary. The big advantage of this strategy is that it is very robust. The disadvantage is the long compile time.

This compilation strategy can be enabled via the feature flag no_static_lib.

Compiling parser and lexer into a separate static library

This strategy works as follows:

  1. Check if there is already a static parser library available at the top repository level (libdyn_quantity_from_str.a). If true, go to step 5. Otherwise, continue with step 2.
  2. Generate a new crate dyn_quantity_from_str using the crate template dyn_quantity_from_str_template and some components of the dyn_quantity crate (namely, lib.rs).
  3. Compile dyn_quantity_from_str to libdyn_quantity_from_str.a using standard cargo invocation (see dyn_quantity_from_str_template/Cargo.toml)
  4. Link to libdyn_quantity_from_str.a in src/from_str/from_str_ext.rs.

The whole procedure is realized in build.rs.

This compilation strategy is used if no_static_lib is disabled.