Runtime Type System • interface

The interface package provides a system for defining and implementing interfaces in R, with runtime type checking, bringing some of the benefits of statically-typed languages to R with zero dependencies.

interface provides:

Interfaces: Define and implement interfaces with type checking. Interfaces can be extended and nested.
Typed Functions: Define functions with strict type constraints.
Typed Frames: Choose between a data.frame or data.table with column type constraints and row validation.
Enums: Define and use enumerated types for stricter type safety.

Installation

Install the package from CRAN:

install.packages("interface")

Or install the latest development version from GitHub:

# Install the package from the source
remotes::install_github("dereckmezquita/interface")

Getting started

Import the package functions.

box::use(interface[ interface, type.frame, fun, enum ])

Define an interface and implement it:

# Define an interface
Person <- interface(
    name = character,
    age = numeric,
    email = character
)

# Implement the interface
john <- Person(
    name = "John Doe",
    age = 30,
    email = "john@example.com"
)

print(john)
#> Object implementing interface:
#>   name: John Doe
#>   age: 30
#>   email: john@example.com
#> Validation on access: Disabled

# interfaces are lists
print(john$name)
#> [1] "John Doe"

# Modify the object
john$age <- 10
print(john$age)
#> [1] 10

# Invalid assignment (throws error)
try(john$age <- "thirty")
#> Error : Property 'age' must be of type numeric

Extending Interfaces and Nested Interfaces

Create nested and extended interfaces:

# Define nested interfaces
Address <- interface(
    street = character,
    city = character,
    postal_code = character
)

Scholarship <- interface(
    amount = numeric,
    status = logical
)

# Extend interfaces
Student <- interface(
    extends = c(Address, Person),
    student_id = character,
    scores = data.table::data.table,
    scholarship = Scholarship
)

# Implement the extended interface
john_student <- Student(
    name = "John Doe",
    age = 30,
    email = "john@example.com",
    street = "123 Main St",
    city = "Small town",
    postal_code = "12345",
    student_id = "123456",
    scores = data.table::data.table(
        subject = c("Math", "Science"),
        score = c(95, 88)
    ),
    scholarship = Scholarship(
        amount = 5000,
        status = TRUE
    )
)

print(john_student)
#> Object implementing interface:
#>   student_id: 123456
#>   scores: Math
#>    scores: Science
#>    scores: 95
#>    scores: 88
#>   scholarship: <environment: 0x12ad94c40>
#>   street: 123 Main St
#>   city: Small town
#>   postal_code: 12345
#>   name: John Doe
#>   age: 30
#>   email: john@example.com
#> Validation on access: Disabled

Custom Validation Functions

Interfaces can have custom validation functions:

is_valid_email <- function(x) {
    grepl("[a-z|0-9]+\\@[a-z|0-9]+\\.[a-z|0-9]+", x)
}

UserProfile <- interface(
    username = character,
    email = is_valid_email,
    age = function(x) is.numeric(x) && x >= 18
)

# Implement with valid data
valid_user <- UserProfile(
    username = "john_doe",
    email = "john@example.com",
    age = 25
)

print(valid_user)
#> Object implementing interface:
#>   username: john_doe
#>   email: john@example.com
#>   age: 25
#> Validation on access: Disabled

# Invalid implementation (throws error)
try(UserProfile(
    username = "jane_doe",
    email = "not_an_email",
    age = "30"
))
#> Error : Errors occurred during interface creation:
#>   - Invalid value for property 'email': FALSE
#>   - Invalid value for property 'age': FALSE

Typed Functions

Define functions with strict type constraints:

typed_fun <- fun(
    x = numeric,
    y = numeric,
    return = numeric,
    impl = function(x, y) {
        return(x + y)
    }
)

print(typed_fun(1, 2))  # [1] 3
#> [1] 3
try(typed_fun("a", 2))  # Invalid call
#> Error : Property 'x' must be of type numeric

Functions with multiple possible return types:

typed_fun2 <- fun(
    x = c(numeric, character),
    y = numeric,
    return = c(numeric, character),
    impl = function(x, y) {
        if (is.numeric(x)) {
            return(x + y)
        } else {
            return(paste(x, y))
        }
    }
)

print(typed_fun2(1, 2))  # [1] 3
#> [1] 3
print(typed_fun2("a", 2))  # [1] "a 2"
#> [1] "a 2"

Typed `data.frame`s and `data.table`s

Create data.frames with column type constraints and row validation:

PersonFrame <- type.frame(
    frame = data.frame, 
    col_types = list(
        id = integer,
        name = character,
        age = numeric,
        is_student = logical
    )
)

# Create a data frame
persons <- PersonFrame(
    id = 1:3,
    name = c("Alice", "Bob", "Charlie"),
    age = c(25, 30, 35),
    is_student = c(TRUE, FALSE, TRUE)
)

print(persons)
#> Typed Data Frame Summary:
#> Base Frame Type: data.frame
#> Dimensions: 3 rows x 4 columns
#> 
#> Column Specifications:
#>   id         : integer
#>   name       : character
#>   age        : numeric
#>   is_student : logical
#> 
#> Frame Properties:
#>   Freeze columns : Yes
#>   Allow NA       : Yes
#>   On violation   : error
#> 
#> Data Preview:
#>   id    name age is_student
#> 1  1   Alice  25       TRUE
#> 2  2     Bob  30      FALSE
#> 3  3 Charlie  35       TRUE

# Invalid modification (throws error)
try(persons$id <- letters[1:3])
#> Error : Property 'id' must be of type integer

Additional options for data.frame validation:

PersonFrame <- type.frame(
    frame = data.frame,
    col_types = list(
        id = integer,
        name = character,
        age = numeric,
        is_student = logical,
        gender = enum("M", "F"),
        email = function(x) all(grepl("^[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}$", x))
    ),
    freeze_n_cols = FALSE,
    row_callback = function(row) {
        if (row$age >= 40) {
            return(sprintf("Age must be less than 40 (got %d)", row$age))
        }
        if (row$name == "Yanice") {
            return("Name cannot be 'Yanice'")
        }
        return(TRUE)
    },
    allow_na = FALSE,
    on_violation = "error"
)

df <- PersonFrame(
    id = 1:3,
    name = c("Alice", "Bob", "Charlie"),
    age = c(25, 35, 35),
    is_student = c(TRUE, FALSE, TRUE),
    gender = c("F", "M", "M"),
    email = c("alice@test.com", "bob_no_valid@test.com", "charlie@example.com")
)

print(df)
#> Typed Data Frame Summary:
#> Base Frame Type: data.frame
#> Dimensions: 3 rows x 6 columns
#> 
#> Column Specifications:
#>   id         : integer
#>   name       : character
#>   age        : numeric
#>   is_student : logical
#>   gender     : Enum(M, F)
#>   email      : custom function
#> 
#> Frame Properties:
#>   Freeze columns : No
#>   Allow NA       : No
#>   On violation   : error
#> 
#> Data Preview:
#>   id name age is_student gender email
#> 1  1 TRUE   1       TRUE   TRUE  TRUE
#> 2  1 TRUE   1       TRUE   TRUE  TRUE
#> 3  1 TRUE   1       TRUE   TRUE  TRUE
summary(df)
#>        id        name                age    is_student    
#>  Min.   :1   Length:3           Min.   :1   Mode:logical  
#>  1st Qu.:1   Class :character   1st Qu.:1   TRUE:3        
#>  Median :1   Mode  :character   Median :1                 
#>  Mean   :1                      Mean   :1                 
#>  3rd Qu.:1                      3rd Qu.:1                 
#>  Max.   :1                      Max.   :1                 
#>  gender.Length  gender.Class  gender.Mode    email          
#>  1        -none-   logical                Length:3          
#>  1        -none-   logical                Class :character  
#>  1        -none-   logical                Mode  :character  
#>                                                             
#>                                                             
#> 

# Invalid row addition (throws error)
try(rbind(df, data.frame(
    id = 4,
    name = "David",
    age = 50,
    is_student = TRUE,
    email = "d@test.com"
)))
#> Error in rbind(deparse.level, ...) : Number of columns must match

Enums

Define enums for categorical variables; these are safe to use to protect a value from being modified to invalid options. The enum function creates a generator which is then used to create the enum object. This can be used standalone or as part of an interface.

Colour <- enum("red", "green", "blue")

# Create an enum object
colour <- Colour("red")
print(colour)
#> Enum: red

colour$value <- "green"
print(colour)
#> Enum: green

# Invalid modification (throws error)
try(colour$value <- "yellow")
#> Error in `$<-.enum`(`*tmp*`, value, value = "yellow") : 
#>   Invalid value. Must be one of: red, green, blue

# Use in an interface
Car <- interface(
    make = enum("Toyota", "Ford", "Chevrolet"),
    model = character,
    colour = Colour
)

# Implement the interface
car1 <- Car(
    make = "Toyota",
    model = "Corolla",
    colour = "red"
)

print(car1)
#> Object implementing interface:
#>   make: Toyota
#>   model: Corolla
#>   colour: red
#> Validation on access: Disabled

# Invalid implementation (throws error)
try(Car(
    make = "Honda",
    model = "Civic",
    colour = "yellow"
))
#> Error : Errors occurred during interface creation:
#>   - Invalid enum value for property 'make': Invalid value. Must be one of: Toyota, Ford, Chevrolet
#>   - Invalid enum value for property 'colour': Invalid value. Must be one of: red, green, blue

# Invalid modification (throws error)
try(car1$colour$value <- "yellow")
#> Error in `$<-.enum`(`*tmp*`, value, value = "yellow") : 
#>   Invalid value. Must be one of: red, green, blue
try(car1$make$value <- "Honda")
#> Error in `$<-.enum`(`*tmp*`, value, value = "Honda") : 
#>   Invalid value. Must be one of: Toyota, Ford, Chevrolet

Conclusion

The interface package provides powerful tools for ensuring type safety and validation in R. By defining interfaces, typed functions, and typed data.frames, you can create robust and reliable data structures and functions with strict type constraints. For more details, refer to the package documentation.

License

This package is licensed under the MIT License.

Citation

If you use this package in your research or work, please cite it as:

Mezquita, D. (2024). interface: A Runtime Type System. R package version 0.1.2. https://github.com/dereckmezquita/interface

interface