Expand description
§Typeables: Rust crate of type aliases and struct tuples
Typeables is a Rust crate of semantic types, such as unit types (e.g. metre for length, second for time), content types (e.g. email address, phone number), locale types (e.g. “en” for English, “zh” for Chinese), etc.
- scientific unit
- time unit
- time period
- currency
- email address
- geolocation
- interval
- locale
- localization
- media type
- phone e164
- grammar
- text formats
- content
- date time
§Introduction
Typeables is based on the Rust pattern of “New Type”. This uses a Rust struct tuple as a wrapper of another type (or types), in order to provide encapsulation.
Example:
pub struct SecondAsF64(pub f64); // This is a "New Type" struct tuple.
let s = SecondAsF64(1.0); // One second of time as a floating-point 64-bit.
Typeables helps you write clearer code and stronger code, because you can be more-precise about your variable types and your function inputs and outputs.
Example:
use typeables::metre::MetreAsStructF64;
use typeables::second::SecondAsStructF64;
fn speed(
m: MetreAsStructF64,
s: SecondAsStructF64
) {
println!("Speed is {}", m.0 / s.0)
}
fn main() {
let m = MetreAsStructF64(2.0);
let s = SecondAsStructF64(3.0);
speed(m, s)
}
Typeables helps you create better domain driven design, stronger compile-time checking, and crisper run-time diagnostics.
§What is a struct tuple?
A struct tuple is akin to a wrapper for another type.
A struct tuple can make your code safer because it provides encapsulation.
Example:
pub struct Foo(pub f64);
let x = Foo(1.0);
§What is a type alias?
A type alias is akin to a nickname for another type.
A type alias can make your code clearer because it expresses your intent.
Example:
pub type Foo = f64;
let x: Foo = 1.0;
§How does Typeables provide a struct tuple and type alias?
Typeables provides concepts with a struct tuple and also with a type alias.
Example:
pub struct SecondAsStructF64(f64);
pub type SecondAsTypeF64 = f64;
§How do I refactor code to use Typeables?
Typeables helps you refactor from weaker-type code to stronger-type code.
Step 1. Start with typical code.
fn speed(
m: f64,
s: f64
) {
println!("The speed is {}", m / s)
}
Step 2. Refactor to a Typeables type alias. This is annotation, because you clarify the types that the function expects.
fn speed(
m: MetreAsTypeF64,
s: SecondAsTypeF64
) {
println!("The speed is {}", m / s)
}
Step 3. Refactor to a Typealias struct tuple. This is encapsulation, because you wrap the value in a struct.
fn speed(
m: MetreAsStructF64,
s: SecondAsStructF64
) {
println!("The speed is {}", m.0 / s.0)
}
§Semantics
§Why use semantic names?
When you use semantic names, such as clear descriptions and purposeful naming conventions, then you help developers understand your code, and help compilers provide reliability, and help tools provide inspectability.
Suppose your code has this function:
fn f(year: i16, month: i16) {
println!("Year {} Month {}", year, month)
}
A developer can use your code like this:
let year = 2022;
let month = 12;
f(year, month); // right
// f(month, year); // wrong, yet will compile and be a bug
You can make your code clearer by adding a type alias:
fn f(year: YearAsCommonEraAsTypeI16, month: MonthOfYearAsTypeI16) {
println!("Year {} Month {}", year, month)
}
You can make your code stronger by using a struct tuple:
fn f(year: YearAsCommonEraAsStructI16, month: MonthOfYearAsStructI16) {
println!("Year {} Month {}", year.0, month.0)
}
A developer can use your code like this:
let year = YearAsCommonEraAsStructI16(2022);
let month = MonthOfYearAsStructI16(12);
f(year, month); // right
// f(month, year); // wrong and won't compile
§Why use semantic names, representation names, unit names, and implementation names?
Suppose you’re writing an application for aircraft.
You want to keep track of:
-
Aircraft altitudes.
-
Representation as “Above Ground Level (AGL)” such as the height of the aircraft above the runway during takeoff or landing, or as “Mean Sea Level (MSL)” such as the worldwide height of the aircraft during cruising flight.
-
Unit of measurement as “Metre” which is the international system, or as “Foot” which is the United States system.
-
The implemention as a signed integer 16-bit, because altitude can be negative in some rare areas such as Death Valley California, and your application may need to integrate with legacy code that requires signed integer 16-bit numbers.
You can use this naming convention:
-
Semantic name “Altitude”
-
As representation name “Above Ground Level” or “Mean Sea Level”
-
As unit name “Metre” or “Foot”
-
As primitive name “I16”.
The code looks like this:
pub struct AltitudeAsAboveGroundLevelAsMetreAsStructI16(pub i16);
pub struct AltitudeAsAboveGroundLevelAsFootAsStructI16(pub i16);
pub struct AltitudeAsMeanSeaLevelAsMetreAsStructI16(pub i16);
pub struct AltitudeAsMeanSeaLevelAsFootAsStructI16(pub i16);
Suppose your app also needs to keep track of:
-
Airport elevations.
-
The representation as “Above Ground Level (AGL)” such as the height of an airport building above the airport runway, or as “Mean Sea Level (MSG)” such as the worldwide height of the airporse runway.
-
Etc.
The code looks like this:
pub struct ElevationAsAboveGroundLevelAsMetreAsStructI16(pub i16);
pub struct ElevationAsAboveGroundLevelAsFootAsStructI16(pub i16);
pub struct ElevationAsMeanSeaLevelAsMetreAsStructI16(pub i16);
pub struct ElevationAsMeanSeaLevelAsFootAsStructI16(pub i16);
The naming convention is crystal clear and fully descriptive:
-
Developers can understand your code better, and how to use it.
-
Compilers can provide stronger compile-time guarantees.
-
Debuggers can provide crisper run-time diagnostics.
-
Editors can provide better auto-complete and auto-suggest.
§Use words rather than abbreviations
Examples of semantic names:
-
Use “Latitude” not “Lat”.
-
Use “Longitude” not “Lon”, “Lng”, “Long”.
Examples of representation names:
-
Use “Decimal Degree” not “DD”
-
Use “Degree Minute Second” not “DMS”.
Examples of unit names:
-
Use “Metre” not “M”.
-
Use “Second” not “S”.
Examples of implementation names:
-
Use “TypeString” not “TS”
-
Use “StructString” not “SS”.
§Prefer singular over plural
Examples of representation names:
-
Use “Decimal Degree” not “Decimal Degrees”
-
Use “Degree Minute Second” not “Degrees Minutes Seconds”
Examples of unit names:
-
Use “Metre” not “Metres”.
-
Use “Second” not “Seconds”
§Prefer StemOfScope over ScopeStem
Example:
-
Use “DayOfWeek” not “WeekDay”
-
Use “DayOfMonth” not “MonthDay”
§Naming conventions
Naming convention for struct tuples:
pub struct FooAsStructI8(pub i8);
pub struct FooAsStructI16(pub i16);
pub struct FooAsStructI32(pub i32);
pub struct FooAsStructI64(pub i64);
pub struct FooAsStructI128(pub i128);
pub struct FooAsStructISize(pub isize);
pub struct FooAsStructU8(pub u8);
pub struct FooAsStructU16(pub u16);
pub struct FooAsStructU32(pub u32);
pub struct FooAsStructU64(pub u64);
pub struct FooAsStructU128(pub u128);
pub struct FooAsStructUSize(pub usize);
pub struct FooAsStructF32(pub f32); pub struct FooAsStructF64(pub f64);
pub struct FooAsStructStr(&'static String); pub struct FooAsStructString(pub
String);
Naming convention for type aliass:
pub type FooAsTypeI8 = i8;
pub type FooAsTypeI16 = i16;
pub type FooAsTypeI32 = i32;
pub type FooAsTypeI64 = i64;
pub type FooAsTypeI128 = i128;
pub type FooAsTypeISize = isize;
pub type FooAsTypeU8 = u8;
pub type FooAsTypeU16 = u16;
pub type FooAsTypeU32 = u32;
pub type FooAsTypeU64 = u64;
pub type FooAsTypeU128 = u128;
pub type FooAsTypeUSize = usize;
pub type FooAsTypeF32 = f32;
pub type FooAsTypeF64 = f64;
pub type FooAsTypeStr = str;
pub type FooAsTypeString = String;
§Comparisons
We recommend looking at the Rust crate uom
(unit of measure) and the Rust
book examples of the newtype
pattern.
§Comparison with uom
Broadly speaking:
-
uom favors high-level work, such as automatic normalizations and conversions.
-
Typeables favors low-level work, such as exact representations and primitives.
Quantities v. units v. primitives:
-
uom deliberately favors working with conceptual quantities (length, mass, time, …) rather than measurement units (metre, gram, second, …) and implementation primitives (pub i8, u16, f32, …).
-
Typeables favors working with explicit measurement units and explicit implementation primitives. When you want the concept of “length” and unit “metre” and primitive “f32” then you write “LengthAsMetreAsTypeF32”.
Normalization v. exactness:
-
uom deliberately normalizes values to their base units, such as normalizing 1 gram to 0.001 kilogram, and deliberately trades away representation capabilities (due to inexact conversions) and precision capabilties (due to bit limits).
-
Typeables favors exactness, never normaliziation. When you want the concept of “mass” and unit “gram” and primitive “u128” for 128-bit unsigned integer precision, then you write “GramAsTypeI128”.
§Comparison with Rust “New Type Idiom” a.k.a. “New Type Pattern”
Broadly speaking:
-
The Rust “New Type Idiom” a.k.a. “New Type Pattern” is exactly what Typeables is doing with struct tuples. We like this idiom very much.
-
Typeables additionally provides type aliass. In practice we find this is an important way to help professional developers with larger codebases, because the developers can phase in the type aliass as hints to developers and to tools, then later on can phase in the struct tuples.
Roll your own versus using Typeables crate:
-
You can certainly roll your own new type pattern, and you can use your own type names, or even use the Typeables type names.
-
The Typeables crate is helpful because it provides a bunch of definitions, so you can use the crate, then get all the benefits of the types, plus your tools can use the crate information, such as for editor tool autocomplete and autosuggest.
§Implementation
The type aliases are all for Rust primitives and standards such as strings
(using str
and String
) and numbers (using i64
, u64
, f64
, et al.).
§Overhead
Typeables has zero or near-zero runtime overhead:
-
A type alias is zero runtime overhead because the type alias is replaced at compile time.
-
A struct tuple is near-zero runtime overhead because the struct tuple is a wrapper with a field.
§Typing
Typeables is deliberately verbose.
-
We use editors with autocomplete and autosuggest, so typing is easy and fast.
-
We like long names for low-level clarity.
-
Typeables defines many type aliass and struct tuples. Typically these are fast during development because they’re simple. Typically these are even faster during production because the Rust compiler can optimized these and also eliminate any that are not needed.
§Macros
The Typeables source code does not use macros.
-
We like macros in general.
-
Yet we discovered in practice that macros seem to interfere with some of our tooling.
-
For example, macros do not seem to work with some editors that inspect the Typeables crate in order to do autocomplete and autosuggest.
§Further reading
Modules§
- adjective
- Adjective
- adverb
- Adverb
- altitude
- Altitude
- ampere
- Ampere
- becquerel
- Becquerel
- candela
- Candela
- currency
- Currency
- date
- Date
- date_
time - DateTime
- day
- Day
- day_
of_ month - DayOfMonth
- day_
of_ week - DayOfWeek
- day_
of_ year - DayOfYear
- decimal_
separator - Decimal Separator
- degree_
celcius - DegreeCelcius
- dual_
interval - Dual Interval
- elevation
- Elevation
- email_
address - Email Address
- email_
address_ addr - Email Address Addr
- email_
address_ name - Email Address Name
- farad
- Farad
- global_
location_ number - Global Location Number
- gram
- Gram
- gray
- Gray
- grouping_
separator - Grouping Separator a.k.a. thousands separator for numbers
- henry
- Henry
- hertz
- Hertz
- hour
- Hour
- hour_
of_ day - HourOfDay
- html_
text - HTML Text
- international_
standard_ of_ industrial_ classification_ revision_ 4_ code - International Standard of Industrial Classification, Revision 4, Code
- international_
standard_ of_ industrial_ classification_ revision_ 4_ name - International Standard of Industrial Classification, Revision 4, Name
- joule
- Joule
- json_
text - JSON Text
- katal
- Katal
- kelvin
- Kelvin
- kilogram
- Kilogram
- latitude
- Latitude
- legal_
entity_ identifier_ code - Legal Entity Identifier Code
- litre
- Litre
- locale_
code - Locale Code
- locale_
country_ code - Locale Country Code
- locale_
language_ code - Locale Language Code
- locale_
region_ code - Locale Region Code
- locale_
script_ code - Locale Script Code
- locale_
variant_ code - Locale Variant Code
- longitude
- Longitude
- lumen
- Lumen
- lux
- Lux
- markdown_
text - Markdown Text
- media_
type - Media type
- metre
- Metre
- metre_2
- Metre^2
- metre_3
- Metre^3
- metre_
per_ second - Metre Per Second
- minute
- Minute
- minute_
of_ hour - MinuteOfHour
- mole
- Mole
- month
- Month
- month_
of_ year - MonthOfYear
- noun
- Noun
- ohm
- Ohm
- open_
location_ code - Open Location Code
- pascal
- Pascal
- phone_
e164_ country_ code - Phone E.164 Country Code
- phone_
e164_ group_ identification_ code - Phone E.164 Group Identification Code
- phone_
e164_ national_ destination_ code - Phone E.164 National Destination Code
- phone_
e164_ subscriber_ number - Phone E.164 Subscriber Number
- phone_
e164_ text - Phone E.164 format
- phone_
e164_ trial_ identification_ code - Phone E.164 Trial Identification Code
- pronoun
- Pronoun
- quotation_
start_ delimiter - Quotation Start Delimiter
- quotation_
stop_ delimiter - Quotation Stop Delimiter
- radian
- Radian
- second
- Second
- second_
of_ minute - SecondOfMinute
- siemens
- Siemens
- sievert
- Sievert
- steradian
- Steradian
- tesla
- Tesla
- time
- Time
- time_
offset - Time offset as format “+HH:MM”
- time_
zone - Time zone as abbreviation text
- unit_
interval - Unit Interval
- value_
added_ tax_ identification_ number - Value Added Tax Identification Number
- verb
- Verb
- volt
- Volt
- watt
- Watt
- weber
- Weber
- week
- Week
- week_
of_ month - WeekOfMonth
- week_
of_ year - WeekOfYear
- what_
free_ words_ code - What Free Words Code
- xml_
text - XML Text
- yaml_
text - YAML Text
- year
- Year
- year_
as_ common_ era - YearAsCommonEra