Crate arithmetic_parser

Parser for arithmetic expressions with flexible definition of literals and support of type annotations.

Overall, parsed grammars are similar to Rust syntax, with a few notable differences.

Supported syntax features

• Variables. A variable name is defined similar to Rust and other programming languages, as a sequence of alphanumeric chars and underscores that does not start with a digit.
• Literals. The parser for literals is user-provided, thus allowing to apply the library to different domains (e.g., finite group arithmetic).
• // and /* .. */ comments.
• Basic arithmetic operations: +, - (binary and unary), *, /, ^ (power). The parser outputs AST with nodes organized according to the operation priority.
• Function calls: foo(1.0, x).
• Parentheses which predictably influence operation priority.

The parser supports both complete and streaming (incomplete) modes; the latter is useful for REPLs and similar applications.

Optional syntax features

These features can be switched on or off when defining a Parse impl by declaring the corresponding Features.

• Tuples. A tuple is two or more elements separated by commas, such as (x, y) or (1, 2 * x). Tuples are parsed both as lvalues and rvalues.
• Tuple destructuring. Using a tuple as an lvalue, for example, (x, y, z) = foo. The “rest” syntax is also supported, either named or unnamed: (head, ...tail) = foo, (a, ..., b, c) = foo.
• Function definitions. A definition looks like a closure definition in Rust, e.g., |x| x - 10 or |x, y| { z = max(x, y); (z - x, z - y) }. A definition may be assigned to a variable (which is the way to define named functions).
• Destructuring for function args. Similar to tuple destructuring, it is possible to destructure and group args in function definitions, for example, |(x, y), ...zs| { }.
• Blocks. A block is several ;-delimited statements enclosed in {} braces, e.g, { z = max(x, y); (z - x, z - y) }. The blocks can be used in all contexts instead of a simple expression; for example, min({ z = 5; z - 1 }, 3).
• Objects. Object is a mapping of string fields to values. Objects are defined via object expressions, which look similar to struct initialization in Rust or object initialization in JavaScript; for example, #{ x: 1, y }. (Note the # char at the start of the block; it is used to distinguish object expressions from blocks.)
• Methods. Method call is a function call separated from the receiver with a . char (e.g., foo.bar(2, x)).
• Type annotations. A type annotation in the form var: Type can be present in the lvalues or in the function argument definitions. The parser for type annotations is user-defined.
• Boolean operations: ==, !=, &&, ||, !.
• Order comparisons, that is, >, <, >=, and <= boolean ops.

Differences with Rust

(within shared syntax constructs; of course, Rust is much more expressive)

• No keyword for assigning a variable (i.e., no let / let mut). There are no keywords in general.
• Functions are only defined via the closure syntax.
• There is “rest” destructuting for tuples and function arguments.
• Type annotations are placed within tuple elements, for example, (x: Num, _) = y.
• Object expressions are enclosed in #{ ... }, similarly to Rhai.
• Object field access and method calls accept arbitrary block-enclosed “names” in addition to simple names. E.g., xs.{Array.len}() is a valid method call with xs receiver, Array.len method name and an empty args list.

Crate features

std

(On by default)

Enables support of types from std, such as the Error trait, and propagates to dependencies.

num-complex

(Off by default)

Implements NumLiteral for floating-point complex numbers (Complex32 and Complex64).

num-bigint

(Off by default)

Implements NumLiteral for BigInt and BigUint from the num-bigint crate.

Examples

Using a grammar for arithmetic on real values.

use arithmetic_parser::{
    grammars::{F32Grammar, Parse, Untyped},
    NomResult, Statement, Expr, FnDefinition, LvalueLen,
};

const PROGRAM: &str = "
    // This is a comment.
    x = 1 + 2.5 * 3 + sin(a^3 / b^2 /* another comment */);
    // Function declarations have syntax similar to Rust closures.
    some_function = |a, b| (a + b, a - b);
    other_function = |x| {
        r = min(rand(), 0.5);
        r * x
    };
    // Tuples and blocks are supported and have a similar syntax to Rust.
    (y, z) = some_function({ x = x - 0.5; x }, x);
    other_function(y - z)
";

let block = Untyped::<F32Grammar>::parse_statements(PROGRAM)?;
// First statement is an assignment.
assert_matches!(
    block.statements[0].extra,
    Statement::Assignment { ref lhs, .. } if *lhs.fragment() == "x"
);
// The RHS of the second statement is a function.
let some_function = match &block.statements[1].extra {
    Statement::Assignment { rhs, .. } => &rhs.extra,
    _ => panic!("Unexpected parsing result"),
};
// This function has a single argument and a single statement in the body.
assert_matches!(
    some_function,
    Expr::FnDefinition(FnDefinition { ref args, ref body, .. })
        if args.extra.len() == LvalueLen::Exact(2)
            && body.statements.is_empty()
            && body.return_value.is_some()
);

Modules

• grammars
  Grammar functionality and a collection of standard grammars.

Structs

• Block
  Block of statements.
• Destructure
  Tuple destructuring, such as (a, b, ..., c).
• Error
  Parsing error with a generic code span.
• FnDefinition
  Function definition, e.g., |x, y| x + y.
• LocatedSpan
  Code span together with information related to where it is located in the code.
• ObjectDestructure
  Object destructuring, such as { x, y: new_y }.
• ObjectDestructureField
  Single field in ObjectDestructure, such as x and y: new_y in { x, y: new_y }.
• ObjectExpr
  Object expression, such as #{ x, y: x + 2 }.

Enums

• BinaryOp
  Binary arithmetic operation.
• Context
  Parsing context.
• DestructureRest
  Rest syntax, such as ...rest in (a, ...rest, b).
• ErrorKind
  Parsing error kind.
• Expr
  Arithmetic expression with an abstract types for type annotations and literals.
• ExprType
  Type of an Expr.
• Lvalue
  Assignable value.
• LvalueLen
  Length of an assigned lvalue.
• LvalueType
  Type of an Lvalue.
• Op
  Generic operation, either unary or binary.
• OpPriority
  Priority of an operation.
• Statement
  Statement: an expression or a variable assignment.
• StatementType
  Type of a Statement.
• UnaryOp
  Unary operation.
• UnsupportedType
  Description of a construct not supported by a certain module (e.g., interpreter or type inference).

Functions

• is_valid_variable_name
  Checks if the provided string is a valid variable name.
• with_span
  Wrapper around parsers allowing to capture both their output and the relevant span.

Type Aliases

• InputSpan
  Code span.
• Location
  Value with an associated code location. Unlike Spanned, Location does not retain a reference to the original code span, just its start position and length.
• NomResult
  Parsing outcome generalized by the type returned on success.
• Spanned
  Value with an associated code span.
• SpannedExpr
  Expr with the associated type and code span.
• SpannedLvalue
  Lvalue with the associated code span.
• SpannedStatement
  Statement with the associated code span.