Bio is a Lisp dialect with an interpreter written in Zig

The language is Scheme inspired and features macros, garbage collection, error handling, an object/module facility, and comes with a small standard library.

A description of the project is available at dev.to

Table of Contents

• Building and running
  • Running tests
• Language Reference
  • Naming conventions
  • Intrinsics
    • nil
    • #t and #f
    • #! and #value
    • typename
    • number? symbol? list? bool? callable? error?
    • var and define
    • set! and unset!
    • arithmetic functions
    • equality
    • order
    • env and gc
    • lambda and λ
    • macro
    • self and environment lookups
    • &rest
    • quote
    • quasiquote
    • unquote
    • unquote-splicing
    • eval
    • apply
    • gensym
    • if
    • cond
    • begin
    • try
    • error
    • print
    • as
    • list
    • append
    • range
    • len
    • string
    • import
    • assert
    • exit
    • verbose
    • math.pi and math.e
    • math.floor
    • string.split
  • Standard library
    • car, cdr, caar, cadr, cddr, caddr, last, nth
    • cons
    • nil?
    • atom?
    • bool?
    • relational functions
    • logical functions
    • let macro
    • filter
    • map
    • quicksort
    • while macro
    • each
    • io.read-number
    • typename
    • time.now
    • double-quote
    • inc! and dec!
    • file i/o and stdin/stdout
    • math.odd? math.even? odd-items and even-items
    • math.abs
    • math.pow
    • math.average
    • math.sqrt
    • math.safe-div
    • math.make-random-generator and math.random-list
    • math.fib and math.fact
    • Y
  • Modules
    • Module example

Building and running

Clone the repository and cd to the root directory. You'll need a recent master build of Zig

Build

zig build

Run the REPL:

./bio

Run a Bio source file:

./bio run examples/triangles.lisp

You can also use import to evaluate files from the REPL, e.g. (import "examples/albums.lisp")

Running tests

The test suite in test.lisp can be evaluated with zig test src/main.zig or ./bio run test.lisp

Language Reference

A Bio program consists of one or more s-expressions. An s-expression is recursively defined as being either

1. an atom
2. a list, which may contain lists and atoms

An atom is either a 64-bit floating point number, or a symbol. A symbol is any sequence of utf8 code points that is not a valid number. Symbols serve as strings when the symbol is enclosed in double-quotes, such as "This is a symbol". Bio does not have a separate string type; the term string is simply used to denote a symbol serving as a string.

Naming conventions

• Predicates have a question mark suffix, such as atom?
• Destructive actions have an exclamation point suffix, such as set!
• Sentinels are prefixed with an ampersand, such as &rest
• Symbols with special meaning are prefixed #, such as #t
• Identifiers are kebab-case, while composite types and module variables are PascalCase
• To logically group functions, infix dots are used, such as io.read-number

Intrinsics

Intrinsics are functions, macros, and symbols implemented by the interpreter. They are building blocks for the standard library and user programs.

nil

A symbol representing the absence of a value. Note that nil and '() are considered equal.

#t and #f

Symbols representing true and false, respectively. Note that nil coerces to #f.

#?

A symbol representing the last top-level expression evaluation.

Example:

(+ 2 3)
(* 2 #?)
10

#! and #value

The #value symbol contains the value returned by a tried expression. This usually removes the need to use a temporary variable when you need to both check for errors and use the result.

The #! symbol contains the error after a try expression. If no error occurs, this is set to nil.

(try (math.safe-div (io.read-number) (io.read-number))
    (print "The doubled result is: " (* 2 #value))
    (print "Failed: " #!))

typename

The 'typename' functions returns a string representation of the type. The most specific classification is returned, so (typename #t) is "bool", even though #t is also a symbol.

(typename 'a)
"symbol"

(typename math.pi)
"number"

(typename #t)
"bool"

(typename '(a b c))
"list"

(typename +)
"function"

number? symbol? list? bool? callable? error?

Predicates to determine expression types. There's also an atom? predicate in the standard library.

These all return #t

(number? -5.7)
(number? x)
(symbol? 'x)
(symbol? #t)
(bool? #t)
(bool? #f)
(list? '(abc))
(callable? +)
(error? (math.safe-div 4 0))

var and define

'var' creates a variable binding in the current environment. The binding will fail if it already exists in the current environment. The define function is just an alias to var

(var x 5)
(define y 5)
(var name (io.read-line))
(var double (lambda (val) (* 2 val)))

Local variables:

(var x 2)
(var y 2)
(var z 2)
(var some-function (lambda (x)

    ; Allowed, y is not in the local scope
    (var y 10)

    ; Not allowed, x is a formal in the same scope
    ; (var x 10)

    (print x y z)
))

> (some-function 3)
3 10 2

set! and unset!

Changes the value of an existing binding (values themselves are immutable). The binding is searched from current to root scope.

; Define x, then update it
(var x 5)
(set! x 10)

; Be evil and redefine + to mean -
(set! + -)
(+ 10 2)
8

; Remove binding, allowing it to be defined again
(unset! x)
(var x 'hey)

arithmetic functions

The arithmetic functions work on floating-point numbers. Unlike infix notation, any number of arguments can be given.

(+ -3 5)
2

(/ 2 (* 10 (+ math.pi 2 3 (- 2 3))))
0.028004957675577865

; pi symbol is same as math.pi (tip: this is available with Option+P if you're on a mac)
(define circumference (λ (x) (* 2 π x)))

Additional math functions are available in the standard library.

equality

The = and ~= functions check for exact and approximate equality. The approximate case is only for numeric operands and accepts a third argument to override the tolerance (epsilon is by default 1e-7)

(= x 5)
(= '(1 2 3 4) nums)
(= '(1 3) (odd-items nums))
(~= a b 0.0005)

Exact equality is implemented in terms of the order function, allowing nested lists to be compared.

order

The order function returns 0, -1, 1 do indicate if the first argument is equal, less than or greater than the second argument. All expression types are supported.

Lists are recursively compared. If list lengths differ, the shorter one is considered smaller. Empty lists and nil are considered equal, otherwise nil is always considered smaller.

(order 100 200)
-1

(order 'def 'abc)
1

(order '(1 2 3) '(0 1 2))
1

(order '(1 2) '(0 1 2))
-1

(order 6 (* 2 3))
0

Standard library functions such as < are implemented in terms of order.

env and gc

env prints the current list of environments in creation order. By default, garbage collection is executed before printing to reduce noise. This can be turned off by passing no-gc, i.e (env no-gc)

> (env)
Environment for global: Env@10db4b000
    import = <function>, env *Env@0
    exit = <function>, env *Env@0
    gc = <function>, env *Env@0
    #f = #f, env *Env@0
    #t = #t, env *Env@0
    #? = #t, env *Env@0
    nil = nil, env *Env@0
    ...

> (gc)
Garbage collected 161 items

The garbage collector runs periodically, though the criteria and extent are intentionally left undefined by this language reference.

lambda and λ

The lambda function creates an anonymous function. Lambdas are usually bound to a variable, but can also be invoked directly or passed and returned as values.

> (var square (lambda (x) (* x x)))
> (square 5)
25

Direct application without binding to a variable:

> ((lambda (x) (* x x)) 5)
25

The lambda symbol can be used in place of the lambda identifier

(var doubler (λ (x) (* 2 x)))

A lambda invocation has its own environment, and the parent environment is the one that existed when the lambda was defined. In other words, Bio is lexically scoped.

macro

This function creates a macro expression. Unlike lambdas, arguments are not evaluated when the macro is invoked. Instead, they're evaluated if and when the body does so, usually using quasi quotation.

When executed, the body is evaluated, usually producing a quasiquote expression (which serves as a code template). The last expression is then evaluated again as the result. This causes the quasiquote expansion to be evaluated as code:

(var print-with-label (macro (label &rest values)
	`(print label ": " ,@values)
))

(print-with-label Primes 2 3 5 7)

Primes: 2 3 5 7

A macro invocation has its own environment, and the parent environment is the current one.

self and environment lookups

The self function returns the current environment as an expression. This can then be used as a function to perform lookups in that environment. This enables composite data types with their own functions, as well as modules. The distinction is purely conceptual.

A top-level (self) call will return the root environment as an expression:

> (self)
<env>

> ((self) +)
<function>

> ((self) (+ 1 2))
3

> (((self) +) 1 2)
3

As you can see, when an environment is placed in the first position of a list, it changes which environment the following argument is looked up in.

Use cases of self include modules, composite data types, polymorphic behavior, and enabling duck-typed interfaces/protocols.

See Modules for more information and an example.

&rest

A sentinel symbol causing the rest of the arguments to be delivered as a single list argument. This enables variadic functions and macros.

quote

The quote function and the ' shorthand returns the argument unevaluated.

> (quote a)
a

> 'a
a

> '(a b 1 2)
(a b 1 2)

quasiquote

The quasiquote function and the ` shorthand returns the argument unevaluated. Unlike quote, however, it allows arguments to be selectively evaluated using unquote and unquote-splicing

; Evaluate one of the list items
> (quasiquote (1 2 (unquote (+ 1 2)) 4))
(1 2 3 4)

; Same thing using shorthand notation
> `(1 2 ,(+ 1 2) 4)
(1 2 3 4)

; Use unquote-splicing to make a larger list of primes:
> (var primes '(2 3 5 7 11 13))
> `(,@primes 17 19 23)
(2 3 5 7 11 13 17 19 23)

Quasi quotation is commonly used to make templates in macros, but it has uses in regular functions as well.

unquote

In the context of a quasiquote, evaluate the argument. The shorthand version is ,

unquote-splicing

In the context of a quasiquote, evaluate the elements of the list and place the result in the enclosing list. The shorthand version is ,@

eval

Evaluates all arguments, leaving the last evaluation as the result. If quote and quasiquote expressions are encountered, these are unquoted before evaluation.

(eval '(+ 1 2))
3

(var expr '(* 2 3))

apply

Evaluates the given function with the given argument list. The last argument must be a list argument. Any preceding arguments are prepended to that list. This means that (apply + 1 '(2 3)) is equivalent to (apply + '(1 2 3)).

> (apply + 5 2 1 '(10 20))
38

> (var list-of-numbers '(1 2 3 4))
> (apply * list-of-numbers)
24

> +
<function>

> (apply #? '(5 2))
7

Using apply is mostly useful when arguments are given as a list and the function at hand expects arguments to be passed individually.

gensym

Generates a unique symbol.

> (gensym)
gensym_1

> (gensym)
gensym_2

if

The if expression evaluates the first argument. If true, then the second argument is evaluated, otherwise the third (optional) argument is evaluated. Each branch can have multiple expressions using constructs such as begin or let.

(if (< x 10) 'Yes 'No)

(var res (if (math.odd? x) 'Odd 'Even))

Here's a list of if expressions from examples/fizzbuzz-if.lisp, none of which have an else branch:

(if (= 0 x) (print "Fizz"))
(if (= 0 y) (print "Buzz"))
(if (and (!= 0 x) (!= 0 y)) (print i))

cond

The cond expression is useful when if/else conditions lead to deep nesting. It takes a variable number of predicate/body pairs, and ends with an else clause. The else clause is required and consists only of the body. It must be the last entry in the cond expression.

Here's the cond expression from examples/fizzbuzz-cond.lisp:

(cond
    ((and (= 0 x) (= 0 y)) (print "FizzBuzz" "\n"))
    ((= 0 x) (print "Fizz" "\n"))
    ((= 0 y) (print "Buzz" "\n"))
    ((print i "\n"))
)

begin

Evaluates a list of expressions and returns the last one as the result. This is useful when more than one expression needs to be evaluated, like in the branches of the if function:

(if (< i 10)
    (begin
        (var x 5)
        (set! x (+ x 1))
        (print (* x 2))
    )
)

12

try

The try function evaluates the first argument. If the result is not an error expression, then the second argument is evaluated (the success branch), otherwise the third argument is evaluated (the error branch). The error branch is optional (in which case nil will be returned; add an error branch if you want to propagate the error.)

It's often necessary to know the value of the tried expression if it succeeds. This can be done using an intermediary variable, or by looking up the #value symbol:

(try (math.safe-div 6 2)
    (if (= #value 3)
        (print "As expected!\n")
    )
)

The error expression is available through the #! symbol, used here by the error branch:

(try (math.safe-div 6 2)
    (if (= #value 3)
        (print "As expected!\n")
    )
    (print "Could not divide: " #!)
)

A function that may fail does not have to be used in a try function:

> (var res (math.safe-div x y))
3.5

> (math.safe-div x 0)
Division by zero

You can also use try after the fact:

> (var res (math.safe-div x 0))
> (try (math.safe-div 1 0) #t `(string "Not good: " ,#!))
Not good: Division by zero

error

The error function creates a new error expression.

(var fail-if-ten (lambda (x)
    (if (= x 10)
        (error "10 is not allowed")
        #t
    )
))

(try (fail-if-ten 10)
    "All good"
    (begin
        (print "Something went terribly wrong!\n")
        #!
    )
)

Something went terribly wrong!
10 is not allowed

Notice how the try expression propagates the error by putting #! as the last expression in the error case.

Errors don't have to be symbols, any expression will do.

print

print prints one or more expressions separated by a space. Combine print with string if you need to print verbatim (without spaces between expressions)

(print "What's your name?")
(var name (io.read-line))

(print "What's your age?")
(var age (as number (io.read-line)))

(print "Hi" name (if (> age 80) "... you're quite old" ""))

IO examples work best if you put them in a file and then use (import "thename.lisp") or bio run thename.lisp

See also the io. functions in the standard library.

as

The as function converts an expression from one type to another. If a conversion is not supported, nil is returned.

The target conversion is either number, symbol, or list.

(var age (as number (io.read-line)))
(assert (= '(5) (as list 5)))
(as symbol mynumber)
(set! age (as symbol age))

list

Creates a new list from its arguments. Quoting can be used to create lists without evaluating the expressions.

> (var x 3)
> (list 1 2 x)
(1 2 3)

> '(1 2 x)
(1 2 x)

append

Creates a new list from its arguments. List arguments are spliced into the new list.

> (append '(a b c) '(d e f (g h)))
(a b c d e f (g h))

> (append 'a 'b 1 2 (+ 1 2))
(a b 1 2 3)

range

The range function is the building block for querying lists, and is used to build standard library functions such as car and cdr.

• If called with no arguments, the first expression in a list is returned.
• If called with one argument, a list containing the sublist from start to end-of-list is returned.
• If called with two arguments, a list containing the sublist from start to end (exclusive) is returned.

Negative indices are end-of-list relative. nil is returned if any indices are out of range.

> (var letters '(a b c d e))
(a b c d e)

> (range letters)
a

> (range letters -1)
(e)

> (range (range letters -1))
e

> (range letters 2)
(c d e)

> (range letters 2 4)
(c d)

> (range letters -4 -2)
(c d)

len

The length of a list (in item count) or symbol (in bytes)

> (len '(1 2 3))
3

string

Creates a symbol by concatenating the rendering of its arguments.

> (var message (string "The value is " x))
The value is 5

> (string "An error occurred : " #!)
An error occurred : Division by zero

import

Reads and evaluates the given file. The path can be relative or absolute.

(import "examples/albums.lisp")

assert

Checks if the expression evaluates to #t. If not, an error is printed and the process is terminated. Evaluates to #t if successful.

(assert (= '(a b c 1 2 3) mylist))

exit

Exits the process with an optional exit code (default is 0)

(exit 1)

verbose

Toggles the verbosity flag. When on, some details are printed during evaluation, such as nil results and quasiquote expansions.

math.pi and math.e

The values of π and Euler's number respectively. Note that the math. prefix is just a naming convention.

math.floor

Returns the largest integer less than or equal to the argument.

(floor math.pi)
3

string.split

Given one or more delimiters, tokenizes the input symbol and produces a list of symbols.

(assert (= (string.split "" ",") '()))
(assert (= (string.split "a" ",") '(a)))
(assert (= (string.split "a,b,c" ",") '(a b c)))
(assert (= (string.split "a,b,c," ",") '(a b c)))
(assert (= (string.split "a,b;c," ",;") '(a b c)))
(assert (= (string.split " " ",") (list " ")))

Standard library

The standard library is a file called std.lisp that's loaded and evaluated when the interpreter starts.

car, cdr, caar, cadr, cddr, caddr, last, nth

These functions treat a list as pairs in the classical Lisp sense. car returns the first list item, while cdr returns the rest of the list. last returns the last list item.

(var nums '(1 2 3 4))
(assert (= 1 (car nums)))
(assert (= 2 (cadr nums)))
(assert (= 3 (caddr nums)))
(assert (= 4 (last nums)))
(assert (= 3 (nth 2 nums)))
(assert (nil? (nth 100 nums)))

cons

Prepends an item to a list:

> (cons 'a '(b c))
(a b c)

nil?

True if the argument is nil or an empty list.

atom?

True if the argument is a number or a symbol (in other words, not a list)

bool?

True if the argument is #t or #f

relational functions

The relation functions are <=, <, >, >=, != in addition to the intrinsic =

logical functions

The and and or macros perform the usual shortcut evaluation. The not function checks if the argument is false. If so, the result is then #t, otherwise it's #f

let macro

Local bindings can be created with a lambda expression. If the only reason to create a lambda is to have local variables, then the let macro is more convenient.

(var x 10)
(var y 20)

(let ((x 5) (y 6))
    (+ x y)
)

11

filter

Filters a list.

> (filter (lambda (x) (< x 5)) '(3 9 5 8 2 4 7))
(3 2 4)

map

Applies a function over one or more lists.

; Create a list of sums taking operands from three lists
> (map + '(0 2 5) '(1 2 3) '(1 2 3))
(2 6 11)

; Double every element in a list
> (map (λ (x) (* 2 x)) '(1 2 3))
(2 4 6)

; A list of pairs with the order reversed
> (map (λ (x y) (list y x)) '(a b c) '(1 2 3))
((1 a) (2 b) (3 c))

quicksort

Sorts a list using the supplied comparator function. The following example sorts the same list in ascending and descending order by passing < and > as the comparator functions. In the ascending example, we also filter out negative numbers:

> (filter 
    (quicksort '(5 40 1 -3 2) <) 
    (λ (x) (>= x 0)))
(1 2 5 40)

> (quicksort '(5 40 1 -3 2) >)
(40 5 2 1 -3)

You can also pass a lambda to do your own ordering. See the albums example file for an example of doing this to sort albums.

while macro

Expands to a tail-recursive function running a body while the predicate holds:

(while (< c 10000000)
    (print "Value is now " c "\n")
    (inc! c)
)

list.iterate and each

The list.iterate function allows for convenient iteration of lists. The first argument is a list. The second argument is a function that's called for every item in the list, with the item as an argument. each is an alias to this function.

; Print all numbers
(each lots-of-numbers print)

; Create a new list of numbers, with double the values
(var result '())
(each nums (lambda (item)
    (set! result (append result (* 2 item)))
))

(2 4 6 8 10 12)

io.read-number

Read a number from stdin.

If input is not a number, an error expression is returned.

typename

Returns the type name of its argument

>(define x 5)
>(typename x)
number

>(typename 'x)
symbol

time.now

Returns the current time in milliseconds since unix epoch:

>(time.now)
1618413357184

double-quote

Renders the argument and wraps the result in double-quotes.

> (double-quote name)
"Joanna"

> (double-quote 5)
"5"

> (double-quote (+ 2 3))
"5"

inc! and dec!

Increments and decrements

file i/o and stdin/stdout

A file is opened with io.open-file, which takes a relative or absolute path name as an argument. The file is then closed with io.close-file. Currently, line oriented reading and writing is supported.

(var report (io.open-file "report.csv"))
(io.read-line report)
(io.write-line report "a new line is appended")
(io.close-file report)

io.read-line returns the error expression "EOF" if end of the file is reached.

io.read-line and io.write-line without a file argument reads and writes to stdin and stdout respectively.

io.read-byte reads one byte at a time from a file.

math.odd? math.even? odd-items and even-items

math.odd? and math.even? determine if a number is odd or even, respectively.

odd-items and even-items create a list with odd- and even indexed items respectively. These functions are 1 based.

>(math.odd? 5)
#t

> (odd-items '(a b c d e f g))
(a c e g)

math.abs

Returns the absolute value of the argument.

(math.abs -17)
17

math.pow

Calculates x^y:

>(math.pow 2 32)
4294967296

math.average

The average of a list of numbers

math.sqrt

Calculate the square root using Newton's method

>(math.sqrt 986)
31.40063693621559

math.safe-div

Divides the first argument by the second. If the divisor is zero, the result is an error.

(try (safe-div 6 2)
    (if (= #value 3)
        (print "As expected!\n")
    )
)

math.make-random-generator and math.random-list

Given a seed, math.make-random-generator creates a linear congruent random number generator.

Example using current Unix epoch as seed:

>(var rng (math.make-random-generator 0))

>(rng)
362807296

>(rng)
1965043776

Given a random number generator, math.random-list generates a list of n random numbers:

>(var rng (math.make-random-generator 0))

>(math.random-list rng 10)
(1055406848 752570112 91411200 3016512 1968096889 765038592 339902464 1666232384 1888402176 197119744)

math.fib and math.fact

Recursive Fibonacci and factorial functions (note: as Bio doesn't support arbitrary precision numbers yet, it can only handle relatively small inputs)

Y

The Y fixpoint combinator (technically, the Z combinator as Bio is applicative)

> (var ! (Y (lambda (r) (lambda (x) (if (< x 2) 1 (* x (r (- x 1))))))))
> (! 5)

120

Modules

A Bio module is a module by convention, somewhat similar to classical Javascript modules:

1. A mod-<modulename>.lisp file with the contents wrapped in a lambda call
2. The last expression is (self), making the environment available to the importer
3. The following definitions are available:
   • module-name, a string describing the module
   • module-version, a list of numbers signifying major, major, and patch
   • module-description, an optional description of the module

Module example

Below is mod-point.lisp, which in this example is placed in a subdirectory called modules

((lambda ()
    ; Functions shared between types are actually macros so their parent 
    ; environment is the calling environment; this way x and y will be found
    (var generic-update (macro (new-x new-y)
        (set! x new-x)
        (set! y new-y)
        nil
    ))

    ; A point datatype with regular formatting
    (var new-point (lambda (x y)
        (var update generic-update)
        (var as-string (lambda ()
            (string x " " y)
        ))
        (self)
    ))

    ; A version with longitude and latitude formatting
    (var new-location (lambda (x y)
        (var update generic-update)
        (var as-string (lambda ()
            (string 
                (math.abs x) (if (< x 0) "° S" "° N")
                "  "
                (math.abs y) (if (< y 0) "° W" "° E"))
        ))
        (self)
    ))

    (var module-name "Position Module")
    (var module-version '(1 0 0))
    (self)
))

To use the module:

(var Point (import "modules/mod-point.lisp"))

(var pt (Point new-location 24.5 69.2))

(pt x)
24.5

((pt as-string))
24.5 °N  69.2 °E

; Call update in two different ways
((pt update) 25.4 70.5)
(pt (update 25.4 70.5))