Finish parser
This commit is contained in:
parent
d553a05f45
commit
13a66f2d16
6 changed files with 772 additions and 58 deletions
107
src/debug/expr_show.ts
Normal file
107
src/debug/expr_show.ts
Normal file
|
|
@ -0,0 +1,107 @@
|
|||
import { Expr, Pattern, ProductPattern, Literal, FieldAssignment, FieldPattern } from '../value';
|
||||
|
||||
export function exprToString(expr: Expr): string {
|
||||
switch (expr.tag) {
|
||||
case "literal":
|
||||
return literalToString(expr.literal);
|
||||
|
||||
case "var_use":
|
||||
return expr.name;
|
||||
|
||||
case "call":
|
||||
return `${expr.name}(${expr.args.map(exprToString).join(", ")})`;
|
||||
|
||||
case "tuple":
|
||||
return `(${expr.exprs.map(exprToString).join(", ")})`;
|
||||
|
||||
case "record": {
|
||||
const fields = expr.fields.map(fieldAssignmentToString).join(", ");
|
||||
return `{ ${fields} }`;
|
||||
}
|
||||
|
||||
case "tag":
|
||||
return `#${expr.tag_name}`;
|
||||
|
||||
case "tagged": {
|
||||
// We wrap the payload in parens if it's complex to be safe,
|
||||
// but strictly speaking #foo 1 is valid.
|
||||
// Let's just output space + expr.
|
||||
const payload = exprToString(expr.expr);
|
||||
// Heuristic: if payload starts with '(', '{', or is simple, we might not need parens,
|
||||
// but for "tagged" expression precedence, it's often safer to wrap unless atomic.
|
||||
return `#${expr.tag_name} ${payload}`;
|
||||
}
|
||||
|
||||
case "let": {
|
||||
const bindings = expr.bindings
|
||||
.map(b => `${productPatternToString(b.pattern)} = ${exprToString(b.expr)}`)
|
||||
.join(", ");
|
||||
return `let { ${bindings} . ${exprToString(expr.body)} }`;
|
||||
}
|
||||
|
||||
case "lambda": {
|
||||
const params = expr.parameters.map(productPatternToString).join(", ");
|
||||
return `fn { ${params} . ${exprToString(expr.body)} }`;
|
||||
}
|
||||
|
||||
case "apply": {
|
||||
const args = expr.args.map(exprToString).join(", ");
|
||||
return `apply(${exprToString(expr.callee)} ! ${args})`;
|
||||
}
|
||||
|
||||
case "match": {
|
||||
const branches = expr.branches
|
||||
.map(b => `${patternToString(b.pattern)} . ${exprToString(b.body)}`)
|
||||
.join(" | ");
|
||||
return `match ${exprToString(expr.arg)} { ${branches} }`;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// === Helpers ===
|
||||
|
||||
function literalToString(lit: Literal): string {
|
||||
switch (lit.tag) {
|
||||
case "number": return lit.value.toString();
|
||||
case "string": return `"${lit.value}"`; // simplistic string escaping
|
||||
}
|
||||
}
|
||||
|
||||
function fieldAssignmentToString(f: FieldAssignment): string {
|
||||
return `${f.name} = ${exprToString(f.expr)}`;
|
||||
}
|
||||
|
||||
// === Pattern Printers ===
|
||||
|
||||
export function patternToString(pat: Pattern): string {
|
||||
switch (pat.tag) {
|
||||
case "tag":
|
||||
return `#${pat.tag_name}`;
|
||||
case "tagged":
|
||||
return `#${pat.tag_name} ${patternToString(pat.pattern)}`;
|
||||
// If it's a product pattern (any, tuple, record)
|
||||
default:
|
||||
return productPatternToString(pat);
|
||||
}
|
||||
}
|
||||
|
||||
export function productPatternToString(pat: ProductPattern): string {
|
||||
switch (pat.tag) {
|
||||
case "any":
|
||||
return pat.name;
|
||||
|
||||
case "tuple":
|
||||
return `(${pat.patterns.map(productPatternToString).join(", ")})`;
|
||||
|
||||
case "record":
|
||||
return `{ ${pat.fields.map(fieldPatternToString).join(", ")} }`;
|
||||
}
|
||||
}
|
||||
|
||||
function fieldPatternToString(f: FieldPattern): string {
|
||||
// Check for punning: if pattern is "any" and name matches fieldName
|
||||
if (f.pattern.tag === "any" && f.pattern.name === f.fieldName) {
|
||||
return f.fieldName;
|
||||
}
|
||||
return `${f.fieldName} = ${productPatternToString(f.pattern)}`;
|
||||
}
|
||||
76
src/debug/repl.ts
Normal file
76
src/debug/repl.ts
Normal file
|
|
@ -0,0 +1,76 @@
|
|||
import * as readline from 'readline';
|
||||
import { parse } from '../parser/parser';
|
||||
import { exprToString } from '../debug/expr_show';
|
||||
import { Result } from '../result';
|
||||
|
||||
// Helper to calculate line/col from an absolute offset
|
||||
function getLineCol(text: string, offset: number) {
|
||||
let line = 1;
|
||||
let col = 1;
|
||||
for (let i = 0; i < offset && i < text.length; i++) {
|
||||
if (text[i] === '\n') {
|
||||
line++;
|
||||
col = 1;
|
||||
} else {
|
||||
col++;
|
||||
}
|
||||
}
|
||||
return { line, col };
|
||||
}
|
||||
|
||||
const rl = readline.createInterface({
|
||||
input: process.stdin,
|
||||
output: process.stdout,
|
||||
prompt: 'expr> '
|
||||
});
|
||||
|
||||
console.log("=== Quick & Dirty Parser REPL ===");
|
||||
console.log("Type an expression to parse and verify round-trip stringification.");
|
||||
console.log("Ctrl+C to exit.\n");
|
||||
|
||||
rl.prompt();
|
||||
|
||||
rl.on('line', (lineInput) => {
|
||||
const trimmed = lineInput.trim();
|
||||
if (trimmed) {
|
||||
try {
|
||||
const result = parse(trimmed);
|
||||
|
||||
if (result.tag === "ok") {
|
||||
const ast = result.value;
|
||||
const reconstructed = exprToString(ast);
|
||||
|
||||
console.log(`\n✅ Parsed Successfully:`);
|
||||
console.log(` ${reconstructed}`);
|
||||
|
||||
} else {
|
||||
const err = result.error;
|
||||
// FIX: Calculate line/col manually using the input string
|
||||
const { line, col } = getLineCol(trimmed, err.span.start);
|
||||
const loc = `${line}:${col}`;
|
||||
|
||||
console.log(`\n❌ Parse Error [${err.tag}] at ${loc}`);
|
||||
|
||||
if ('expected' in err) {
|
||||
console.log(` Expected: ${(err as any).expected}`);
|
||||
}
|
||||
if ('reason' in err) {
|
||||
console.log(` Reason: ${(err as any).reason}`);
|
||||
}
|
||||
if ('received' in err) {
|
||||
// Useful to see what char we actually got (print char code if needed)
|
||||
console.log(` Received CodePoint: ${(err as any).received}`);
|
||||
}
|
||||
}
|
||||
} catch (e) {
|
||||
console.log(`\n🔥 CRASH (Uncaught Exception):`);
|
||||
console.log(e);
|
||||
}
|
||||
}
|
||||
|
||||
console.log();
|
||||
rl.prompt();
|
||||
}).on('close', () => {
|
||||
console.log('Bye!');
|
||||
process.exit(0);
|
||||
});
|
||||
435
src/parser/parser.ts
Normal file
435
src/parser/parser.ts
Normal file
|
|
@ -0,0 +1,435 @@
|
|||
import { Expr, ExprBinding, FieldAssignment, FieldPattern, MatchBranch, Pattern, ProductPattern } from '../value';
|
||||
import { Cursor } from './cursor';
|
||||
import { ExprScanError, exprStart, ExprStartToken, identifier, isNextTokenExprStart, isNextTokenProductPatternStart, patternStart, PatternStartToken, skipWhitespaceAndComments } from './scanner';
|
||||
import { char, CodePoint, SourceText, Span } from './source_text';
|
||||
import { Result } from '../result';
|
||||
|
||||
// CONVENTION: Every parser is responsible to consume whitespace/comments at the end.
|
||||
// Every parser is not responsible for cleaning up whitespace/comments at the start - only the final `parse` that's exposed to the public.
|
||||
|
||||
const DELIMITER_COMMA = char(',');
|
||||
const DELIMITER_PIPE = char('|');
|
||||
const TERMINATOR_CLOSE_PAREN = char(')');
|
||||
const TERMINATOR_CLOSE_BRACE = char('}');
|
||||
const TERMINATOR_DOT = char('.');
|
||||
|
||||
// TODO: Errors have to have better contextual information
|
||||
export type ParseError =
|
||||
| ExprScanError
|
||||
| { tag: "UnexpectedToken", expected: string, span: Span }
|
||||
| { tag: "UnexpectedTokenWhileParsingSequence", reason: "ExpectedDelimiterOrTerminator", received: CodePoint, expectedDelimiter: CodePoint, expectedTerminator: CodePoint, span: Span }
|
||||
|
||||
// === Specific Context Errors ===
|
||||
| { tag: "ExpectedExpression", span: Span } // Expected start of expr (e.g. hit EOF or keyword)
|
||||
| { tag: "ExpectedFieldAssignmentSymbol", span: Span } // Expected '=' in field assignment
|
||||
| { tag: "ExpectedPatternAssignmentSymbol", span: Span } // Expected '=' in pattern assignment
|
||||
| { tag: "ExpectedPatternBindingSymbol", span: Span } // Expected '.' in pattern binding
|
||||
| { tag: "ExpectedFunctionCallStart", span: Span } // Expected '(' after function name
|
||||
| { tag: "ExpectedLetBlockOpen", span: Span } // Expected '{' after 'let'
|
||||
| { tag: "ExpectedLetBlockClose", span: Span } // Expected '}' at end of 'let' expression
|
||||
| { tag: "ExpectedMatchBlockOpen", span: Span } // Expected '{' after 'match'
|
||||
| { tag: "ExpectedMatchBlockClose", span: Span } // Expected '}' at end of 'match' expression
|
||||
| { tag: "ExpectedLambdaBlockOpen", span: Span } // Expected '{' after `fn`
|
||||
| { tag: "ExpectedLambdaBlockClose", span: Span } // Expected '}' at end of `fn` expression
|
||||
| { tag: "ExpectedApplyStart", span: Span } // Expected '(' after 'apply'
|
||||
| { tag: "ExpectedApplySeparator", span: Span } // Expected '!' inside 'apply'
|
||||
| { tag: "UnexpectedTagPattern", span: Span } // Found #tag where product pattern expected
|
||||
| { tag: "ExpectedPattern", span: Span } // EOF or invalid start of pattern
|
||||
| { tag: "ExpectedRecordField", span: Span }; // Expected identifier in record pattern
|
||||
|
||||
// TODO: Delete?
|
||||
export type Expectation =
|
||||
| "ExpectedExpression"
|
||||
| "ExpectedFieldAssignmentSymbol"
|
||||
| "ExpectedPatternAssignmentSymbol"
|
||||
| "ExpectedPatternBindingSymbol"
|
||||
| "ExpectedFunctionCallStart"
|
||||
| "ExpectedLetBlockOpen"
|
||||
| "ExpectedLetBlockClose"
|
||||
| "ExpectedMatchBlockOpen"
|
||||
| "ExpectedMatchBlockClose"
|
||||
| "ExpectedApplyStart"
|
||||
| "ExpectedApplySeparator"
|
||||
| "UnexpectedTagPattern"
|
||||
| "ExpectedPattern"
|
||||
| "ExpectedRecordField"
|
||||
|
||||
export type Parser<T> = (cursor: Cursor) => T
|
||||
|
||||
// === Helpers ===
|
||||
// Don't use `cursor.next()` directly - for most of the cases you also want to consume the WS/comments after.
|
||||
function nextWS(cursor: Cursor): CodePoint | undefined {
|
||||
const c = cursor.next();
|
||||
if (c === undefined) {
|
||||
return undefined;
|
||||
} else {
|
||||
skipWhitespaceAndComments(cursor);
|
||||
return c;
|
||||
}
|
||||
}
|
||||
|
||||
// Returns true if the char was consumed.
|
||||
function tryConsume(cursor: Cursor, expected: CodePoint): boolean {
|
||||
if (cursor.peek() === expected) {
|
||||
nextWS(cursor);
|
||||
return true;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
// TODO: Perhaps when it comes to terminators, we should allow the user of this function to create better contextual errors?
|
||||
// Parses a delimited sequence of `p` that's terminated by a special character.
|
||||
function delimitedTerminalSequence<A>(cursor: Cursor, delimiter: CodePoint, terminator: CodePoint, p: Parser<A>): A[] {
|
||||
// Let's say the terminator is `.` and delimiter is `,`. This is the grammar for valid sequences
|
||||
// list(p) :=
|
||||
// | p
|
||||
// | p , p
|
||||
// | p , p , p
|
||||
// | ...
|
||||
// delimitedTerminalSequence(p) :=
|
||||
// | .
|
||||
// | optional(,) list(p) optional(,) .
|
||||
//
|
||||
// All our use-cases always have a well-defined terminator character:
|
||||
// tuples: ( a, b, c ) -> `)`
|
||||
// records: { f0 = e0, f1 = e1 } -> `}`
|
||||
// function call: f(a, b, c) -> `)`
|
||||
// let-binding: let { p = e . body } -> `.`
|
||||
// fn-asbtraction: fn { p0, p1 . body } -> `.`
|
||||
// fn-application: apply(e ! e0, e1, e2) -> `)`
|
||||
const items: A[] = [];
|
||||
|
||||
if (cursor.peek() === terminator) {
|
||||
nextWS(cursor);
|
||||
return items;
|
||||
}
|
||||
|
||||
// optional delimiter
|
||||
if (cursor.peek() === delimiter) {
|
||||
nextWS(cursor);
|
||||
}
|
||||
|
||||
while (true) {
|
||||
const item = p(cursor); // `p` should be responsible for getting rid of whitespace after it has done its work
|
||||
items.push(item);
|
||||
|
||||
const nextChar = cursor.peek();
|
||||
|
||||
if (nextChar === terminator) {
|
||||
nextWS(cursor);
|
||||
|
||||
return items;
|
||||
} else if (nextChar === delimiter) {
|
||||
nextWS(cursor);
|
||||
|
||||
if (cursor.peek() === terminator) {
|
||||
nextWS(cursor);
|
||||
// the previous delimiter was trailing
|
||||
return items;
|
||||
}
|
||||
// the previous delimiter was a separator
|
||||
continue;
|
||||
} else {
|
||||
throw {
|
||||
tag: "UnexpectedTokenWhileParsingSequence",
|
||||
received: nextChar,
|
||||
reason: "ExpectedDelimiterOrTerminator",
|
||||
expectedDelimiter: delimiter,
|
||||
expectedTerminator: terminator,
|
||||
span: cursor.makeSpan(cursor.currentLocation()),
|
||||
} as ParseError;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
function exprStartToken(cursor: Cursor): ExprStartToken {
|
||||
const token = exprStart(cursor);
|
||||
skipWhitespaceAndComments(cursor);
|
||||
return token;
|
||||
}
|
||||
|
||||
function patternStartToken(cursor: Cursor): PatternStartToken {
|
||||
const token = patternStart(cursor);
|
||||
skipWhitespaceAndComments(cursor);
|
||||
return token;
|
||||
}
|
||||
|
||||
// === Expression Parsers ===
|
||||
|
||||
function expr(cursor: Cursor): Expr {
|
||||
const token = exprStartToken(cursor);
|
||||
// TODO: You need to include the spans and perhaps other meta-info.
|
||||
switch (token.tag) {
|
||||
case "EOF":
|
||||
throw {
|
||||
tag: "UnexpectedToken",
|
||||
expected: "Expression",
|
||||
span: token.span
|
||||
} as ParseError;
|
||||
case "number":
|
||||
return Expr.number(token.value);
|
||||
case "string":
|
||||
return Expr.string(token.text);
|
||||
case "variable_use":
|
||||
return Expr.var_use(token.name);
|
||||
case "tag":
|
||||
// #true
|
||||
// #foo e
|
||||
if (isNextTokenExprStart(cursor)) {
|
||||
const value = expr(cursor);
|
||||
return Expr.tagged(token.name, value);
|
||||
} else {
|
||||
return Expr.tag(token.name);
|
||||
}
|
||||
case "tuple_start":
|
||||
// e.g. (a, b, c)
|
||||
const items = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_CLOSE_PAREN, expr);
|
||||
return Expr.tuple(items);
|
||||
case "record_start":
|
||||
// e.g. { x = 1, y = 2 }
|
||||
const fields = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_CLOSE_BRACE, fieldAssignment);
|
||||
return Expr.record(fields);
|
||||
case "function_name":
|
||||
// e.g. my_func(arg1, arg2)
|
||||
// parse a `,` delimiter sequence of expr
|
||||
// need to consume )
|
||||
if (!tryConsume(cursor, char('('))) {
|
||||
throw {
|
||||
tag: "ExpectedFunctionCallStart",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
const args = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_CLOSE_PAREN, expr);
|
||||
return Expr.call(token.name, args);
|
||||
case "keyword":
|
||||
switch (token.kw) {
|
||||
case "let":
|
||||
// let { p0 = e0, p1 = e2 . body }
|
||||
if (!tryConsume(cursor, char('{'))) {
|
||||
throw {
|
||||
tag: "ExpectedLetBlockOpen",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
const bindings = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_DOT, productPatternBinding);
|
||||
const body = expr(cursor);
|
||||
|
||||
if (!tryConsume(cursor, TERMINATOR_CLOSE_BRACE)) {
|
||||
throw {
|
||||
tag: "ExpectedLetBlockClose",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
return Expr.let_(bindings, body);
|
||||
case "fn": {
|
||||
// fn { p0, p1, p2 . body }
|
||||
if (!tryConsume(cursor, char('{'))) {
|
||||
throw {
|
||||
tag: "ExpectedLambdaBlockOpen",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
|
||||
const parameters = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_DOT, productPattern);
|
||||
const body = expr(cursor);
|
||||
if (!tryConsume(cursor, TERMINATOR_CLOSE_BRACE)) {
|
||||
throw {
|
||||
tag: "ExpectedLambdaBlockClose",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
return Expr.lambda(parameters, body);
|
||||
}
|
||||
case "apply":
|
||||
// apply(e ! e0, e1, e2)
|
||||
if (!tryConsume(cursor, char('('))) {
|
||||
throw {
|
||||
tag: "ExpectedApplyStart",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
const callee = expr(cursor);
|
||||
if (!tryConsume(cursor, char('!'))) {
|
||||
throw {
|
||||
tag: "ExpectedApplySeparator",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
|
||||
const args = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_CLOSE_PAREN, expr);
|
||||
return Expr.apply(callee, args);
|
||||
case "match":
|
||||
// match e { branch0 | branch1 | branch2 }
|
||||
const arg = expr(cursor);
|
||||
if (!tryConsume(cursor, char('{'))) {
|
||||
throw {
|
||||
tag: "ExpectedMatchBlockOpen",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
|
||||
const branches = delimitedTerminalSequence(cursor, DELIMITER_PIPE, TERMINATOR_CLOSE_BRACE, matchBranch);
|
||||
return Expr.match(arg, branches)
|
||||
case "=":
|
||||
case "|":
|
||||
case "!":
|
||||
// These keywords CANNOT start an expression.
|
||||
throw {
|
||||
tag: "ExpectedExpression",
|
||||
span: token.span
|
||||
} as ParseError;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
function matchBranch(cursor: Cursor): MatchBranch {
|
||||
// p . body
|
||||
const p = pattern(cursor);
|
||||
|
||||
if (!tryConsume(cursor, char("."))) {
|
||||
throw {
|
||||
tag: "ExpectedPatternBindingSymbol",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
const e = expr(cursor);
|
||||
return Expr.matchBranch(p, e);
|
||||
}
|
||||
|
||||
function productPatternBinding(cursor: Cursor): ExprBinding {
|
||||
// TODO: There's a potential here to do a lot of work on nice errors.
|
||||
// `p = e`
|
||||
// here there could be problems like the pattern being just a variable that uses `=` as its part
|
||||
// `x= = 123` is valid. Maybe in case of erroneous things like
|
||||
// `x=123` - which just parses as an identifier - we should analyze the identifier and product a suggestion? idk...
|
||||
// or even...
|
||||
// `x= 123` - which just parses as an identifier - we should analyze the identifier and product a suggestion? idk...
|
||||
// or
|
||||
// `x =123` - this one technically is a sequence of two identifiers.
|
||||
const pattern = productPattern(cursor);
|
||||
|
||||
if (!tryConsume(cursor, char('='))) {
|
||||
throw {
|
||||
tag: "ExpectedPatternBindingSymbol",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
const e = expr(cursor);
|
||||
return Expr.exprBinding(pattern, e);
|
||||
}
|
||||
|
||||
function fieldAssignment(cursor: Cursor): FieldAssignment {
|
||||
// `f = e`
|
||||
const { name, span } = identifier(cursor, 'identifier');
|
||||
|
||||
if (!tryConsume(cursor, char('='))) {
|
||||
throw {
|
||||
tag: "ExpectedFieldAssignmentSymbol", // Specific error for this context
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError;
|
||||
}
|
||||
|
||||
const value = expr(cursor);
|
||||
return Expr.fieldAssignment(name, value);
|
||||
}
|
||||
|
||||
function pattern(cursor: Cursor): Pattern {
|
||||
// x
|
||||
// (x, y, z)
|
||||
// ((x, y), z)
|
||||
// { a = x, b = y }
|
||||
// { a, b = y }
|
||||
// #foo
|
||||
// #foo x
|
||||
// #foo (x, y)
|
||||
const token = patternStartToken(cursor);
|
||||
|
||||
// === Tag Pattern logic ===
|
||||
if (token.tag === "tag") {
|
||||
// Check Lookahead: Is this a Tagged Value? e.g. #foo x or #foo (a,b)
|
||||
if (isNextTokenProductPatternStart(cursor)) {
|
||||
// Parse the payload (must be a product pattern)
|
||||
const payload = productPattern(cursor);
|
||||
return Pattern.tagged(token.name, payload);
|
||||
} else {
|
||||
// Standalone Tag: #foo
|
||||
return Pattern.tag(token.name);
|
||||
}
|
||||
}
|
||||
|
||||
return finishProductPattern(cursor, token);
|
||||
}
|
||||
|
||||
function productPattern(cursor: Cursor): ProductPattern {
|
||||
// x
|
||||
// (x, y, z)
|
||||
// ((x, y), z)
|
||||
// { a = x, b = y }
|
||||
// { a, b = y }
|
||||
//
|
||||
// tag patterns are syntax errors i.e. unexpected-tag-pattern (renderer then could explain that we don't allow tag patterns to be deeply nested)
|
||||
|
||||
const token = patternStartToken(cursor);
|
||||
return finishProductPattern(cursor, token);
|
||||
}
|
||||
|
||||
function finishProductPattern(cursor: Cursor, token: PatternStartToken): ProductPattern {
|
||||
switch (token.tag) {
|
||||
case "pattern_binding":
|
||||
// foo
|
||||
return ProductPattern.any(token.name);
|
||||
|
||||
case "tuple_start": {
|
||||
// ( p1, p2 )
|
||||
const items = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_CLOSE_PAREN, productPattern);
|
||||
return ProductPattern.tuple(items);
|
||||
}
|
||||
|
||||
case "record_start": {
|
||||
// { a = p, b }
|
||||
const fields = delimitedTerminalSequence(cursor, DELIMITER_COMMA, TERMINATOR_CLOSE_BRACE, recordPatternField);
|
||||
return ProductPattern.record(fields);
|
||||
}
|
||||
|
||||
case "tag":
|
||||
throw { tag: "UnexpectedTagPattern", span: token.span } as ParseError;
|
||||
|
||||
case "EOF":
|
||||
throw { tag: "ExpectedPattern", span: token.span } as ParseError;
|
||||
}
|
||||
}
|
||||
|
||||
function recordPatternField(cursor: Cursor): FieldPattern {
|
||||
const { name, span } = identifier(cursor, 'identifier'); // Reuse existing identifier scanner
|
||||
if (tryConsume(cursor, char('='))) {
|
||||
const p = productPattern(cursor);
|
||||
return ProductPattern.fieldPattern(name, p);
|
||||
} else {
|
||||
// Punning: { a } -> { a = a }
|
||||
return ProductPattern.fieldPattern(name, ProductPattern.any(name));
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
export function parse(input: string): Result<Expr, ParseError> {
|
||||
const source = new SourceText(input);
|
||||
const cursor = new Cursor(source);
|
||||
|
||||
try {
|
||||
skipWhitespaceAndComments(cursor);
|
||||
const expression = expr(cursor);
|
||||
|
||||
if (!cursor.eof()) {
|
||||
return Result.error({
|
||||
tag: "UnexpectedToken",
|
||||
expected: "End of File",
|
||||
span: cursor.makeSpan(cursor.currentLocation())
|
||||
} as ParseError);
|
||||
}
|
||||
|
||||
return Result.ok(expression);
|
||||
} catch (e) {
|
||||
// TODO: This is a bit sketchy. We maybe forced to have "checked" Exceptions for `ParseError` by wrapping it in something that has a proper tag.
|
||||
return Result.error(e as ParseError);
|
||||
}
|
||||
}
|
||||
|
|
@ -6,7 +6,7 @@ import type { Cursor, CursorState, GenericScanError, NumberError, StringError }
|
|||
import { Result } from '../result';
|
||||
import { Expr } from 'src/value';
|
||||
|
||||
function skipWhitespaceAndComments(cursor: Cursor): number {
|
||||
export function skipWhitespaceAndComments(cursor: Cursor): number {
|
||||
let totalConsumed = 0;
|
||||
|
||||
while (true) {
|
||||
|
|
@ -60,7 +60,13 @@ export type ExprScanError =
|
|||
| { tag: "InvalidIdentifier", text: string, kind: IdentifierKind, reason: IdentifierErrorReason, span: Span }
|
||||
|
||||
// What kind of identifier were we trying to parse?
|
||||
export type IdentifierKind = "identifier" | "variable_use" | "tag_construction" | "function_call";
|
||||
export type IdentifierKind =
|
||||
| "identifier"
|
||||
| "variable_use"
|
||||
| "tag_construction"
|
||||
| "function_call"
|
||||
| "pattern_binding";
|
||||
|
||||
export type IdentifierErrorReason =
|
||||
| { tag: "StartsWithDigit" }
|
||||
| { tag: "IsKeyword", kw: Keyword }
|
||||
|
|
@ -73,12 +79,20 @@ export type ExprStartToken =
|
|||
| { tag: "function_name", name: string, span: Span }
|
||||
| { tag: "variable_use", name: string, span: Span }
|
||||
| { tag: "tag", name: string, span: Span }
|
||||
| { tag: "tagged", name: string, span: Span } // TODO: This may be a bit weird. Actually we can lookahead and see if the next char after the identifier is NOT-AN-EXPRESSION start.
|
||||
| { tag: "tuple_start", span: Span }
|
||||
| { tag: "record_start", span: Span }
|
||||
| { tag: "keyword", kw: Keyword, span: Span }
|
||||
// TODO: ger rid of EOF
|
||||
| { tag: "EOF", span: Span }
|
||||
|
||||
export type PatternStartToken =
|
||||
| { tag: "pattern_binding", name: string, span: Span }
|
||||
| { tag: "tag", name: string, span: Span }
|
||||
| { tag: "tuple_start", span: Span }
|
||||
| { tag: "record_start", span: Span }
|
||||
// TODO: ger rid of EOF
|
||||
| { tag: "EOF", span: Span };
|
||||
|
||||
// === Identifier Scanners ===
|
||||
|
||||
// Returns the raw string.
|
||||
|
|
@ -146,12 +160,14 @@ export function identifier(cursor: Cursor, kind: IdentifierKind): { name: string
|
|||
|
||||
// === Literal Scanners ===
|
||||
// throws ExprScanError
|
||||
// TODO: handle trailing whitespace
|
||||
function number(cursor: Cursor): { value: number, span: Span } {
|
||||
const res = scanNumber(cursor);
|
||||
if (res.tag === "ok") { return res.value; } else { throw (res.error as ExprScanError); }
|
||||
}
|
||||
|
||||
// throws ExprScanError
|
||||
// TODO: handle trailing whitespace
|
||||
function string(cursor: Cursor): { value: string, span: Span } {
|
||||
const res = scanString(cursor);
|
||||
if (res.tag === "ok") { return res.value; } else { throw (res.error as ExprScanError); }
|
||||
|
|
@ -159,10 +175,10 @@ function string(cursor: Cursor): { value: string, span: Span } {
|
|||
|
||||
// === complex scanners ===
|
||||
|
||||
// TODO: in exprStart/patternStart make sure whitespace is consumed after they succesfuly produce token. Should we build it in the functions? Or should that be in `parser.ts`?
|
||||
|
||||
// throws ExprScanError
|
||||
export function exprStart(cursor: Cursor): ExprStartToken {
|
||||
skipWhitespaceAndComments(cursor);
|
||||
|
||||
const start = cursor.currentLocation();
|
||||
if (cursor.eof()) {
|
||||
return { tag: "EOF", span: cursor.makeSpan(start) };
|
||||
|
|
@ -220,53 +236,105 @@ export function exprStart(cursor: Cursor): ExprStartToken {
|
|||
}
|
||||
}
|
||||
|
||||
export function patternStart(cursor: Cursor): PatternStartToken {
|
||||
const start = cursor.currentLocation();
|
||||
|
||||
// TODO: Need a Token to TokenKind function
|
||||
// TODO: Need is_start_of_expression(token): boolean
|
||||
// identifier -> true
|
||||
// symbol # -> true
|
||||
// symbol $ -> true
|
||||
// symbol @ -> true
|
||||
// symbol ( -> true
|
||||
// symbol { -> true // this is actually context dependent. Sometimes its a start of a binding context { params . body } or { let-params . body }, and sometimes it is a record. But this function is gonna be used only in the first context
|
||||
// symbol _ -> false
|
||||
// number -> true
|
||||
// string -> true
|
||||
// keyword let -> true
|
||||
// keyword fn -> true
|
||||
// keyword apply -> true
|
||||
// keyword = -> false
|
||||
// keyword | -> false
|
||||
// EOF -> false
|
||||
//
|
||||
// TODO: function that matches a token with a token_type (returns bool)
|
||||
if (cursor.eof()) {
|
||||
return { tag: "EOF", span: cursor.makeSpan(start) };
|
||||
}
|
||||
|
||||
// TODO: forbidden characters are
|
||||
// '('
|
||||
// ')'
|
||||
// '{'
|
||||
// '}'
|
||||
// '.'
|
||||
// ','
|
||||
// '|'
|
||||
// '$'
|
||||
// '#'
|
||||
// '@'
|
||||
// '"'
|
||||
// ' '
|
||||
// '\r'
|
||||
// '\t'
|
||||
// '\n'
|
||||
// TODO: need function is_forbidden_char
|
||||
const c = cursor.peek()!;
|
||||
// === tuple ===
|
||||
if (c === char('(')) {
|
||||
cursor.next();
|
||||
return { tag: "tuple_start", span: cursor.makeSpan(start) };
|
||||
}
|
||||
|
||||
// === record ===
|
||||
if (c === char('{')) {
|
||||
cursor.next();
|
||||
return { tag: "record_start", span: cursor.makeSpan(start) };
|
||||
}
|
||||
|
||||
// === tag ===
|
||||
if (c === char('#')) {
|
||||
cursor.next();
|
||||
const { name } = identifier(cursor, 'tag_construction');
|
||||
return { tag: "tag", name, span: cursor.makeSpan(start) };
|
||||
}
|
||||
|
||||
// === scanner functions ===
|
||||
// TODO: whitespace - consumes whitespace
|
||||
// TODO: comment - consumes token
|
||||
// TODO: raw_identifier - consumes raw identifier - then we can decide whether that was a keyword or an identifier
|
||||
// TODO: string - consumes string like "foo bar\njfjdsajfksd"
|
||||
// TODO: number - consumes number like 123123 or 000123 or 23919233.123
|
||||
//
|
||||
// TODO: token - gives next token
|
||||
// TODO: This is more subtle... -foo is a valid pattern name... I think I should restrict: can't have identifiers start with `-`? But then `-` itself can't be an identifier, which is a bit sad.
|
||||
// TODO: This is gonna be different once we allow number/string literals as patterns.
|
||||
if (isDigit(c) || c === char('"') || c === char('-')) {
|
||||
throw {
|
||||
tag: "UnexpectedCharacter",
|
||||
char: c,
|
||||
span: cursor.makeSpan(start)
|
||||
} as ExprScanError;
|
||||
}
|
||||
|
||||
// === pattern binding ===
|
||||
const { name } = identifier(cursor, 'pattern_binding');
|
||||
return { tag: "pattern_binding", name, span: cursor.makeSpan(start) };
|
||||
}
|
||||
|
||||
export function isNextTokenExprStart(cursor: Cursor): boolean {
|
||||
const state = cursor.save();
|
||||
try {
|
||||
const token = exprStart(cursor);
|
||||
|
||||
switch (token.tag) {
|
||||
case "number":
|
||||
case "string":
|
||||
case "variable_use":
|
||||
case "tag":
|
||||
case "tuple_start":
|
||||
case "record_start":
|
||||
case "function_name": // e.g. my_func(x)
|
||||
return true;
|
||||
|
||||
case "keyword":
|
||||
switch (token.kw) {
|
||||
case "let":
|
||||
case "fn":
|
||||
case "match":
|
||||
case "apply":
|
||||
return true;
|
||||
case "=":
|
||||
case "|":
|
||||
case "!":
|
||||
return false;
|
||||
}
|
||||
|
||||
case "EOF":
|
||||
return false;
|
||||
|
||||
default:
|
||||
return false;
|
||||
}
|
||||
|
||||
} catch (e) {
|
||||
return false;
|
||||
} finally {
|
||||
cursor.restore(state);
|
||||
}
|
||||
}
|
||||
|
||||
export function isNextTokenProductPatternStart(cursor: Cursor): boolean {
|
||||
const state = cursor.save();
|
||||
try {
|
||||
const token = patternStart(cursor);
|
||||
switch (token.tag) {
|
||||
case "pattern_binding":
|
||||
case "tuple_start":
|
||||
case "record_start":
|
||||
return true;
|
||||
default:
|
||||
return false;
|
||||
}
|
||||
} catch (e) {
|
||||
return false;
|
||||
} finally {
|
||||
cursor.restore(state);
|
||||
}
|
||||
}
|
||||
|
|
|
|||
44
src/value.ts
44
src/value.ts
|
|
@ -145,8 +145,8 @@ export type Expr =
|
|||
| { tag: "tag", tag_name: Tag }
|
||||
| { tag: "tagged", tag_name: Tag, expr: Expr }
|
||||
| { tag: "tuple", exprs: Expr[] }
|
||||
| { tag: "record", fields: { name: FieldName, expr: Expr }[] }
|
||||
| { tag: "match", arg: Expr, branches: { pattern: Pattern, body: Expr }[] }
|
||||
| { tag: "record", fields: FieldAssignment[] }
|
||||
| { tag: "match", arg: Expr, branches: MatchBranch[] }
|
||||
| { tag: "lambda", parameters: ProductPattern[], body: Expr }
|
||||
| { tag: "apply", callee: Expr, args: Expr[] }
|
||||
|
||||
|
|
@ -155,14 +155,23 @@ export type Literal =
|
|||
| { tag: "string", value: string }
|
||||
|
||||
export type ExprBinding = {
|
||||
var: ProductPattern,
|
||||
pattern: ProductPattern,
|
||||
expr: Expr,
|
||||
}
|
||||
|
||||
export type MatchBranch = {
|
||||
pattern: Pattern,
|
||||
body: Expr,
|
||||
}
|
||||
|
||||
export type FieldAssignment = { name: FieldName, expr: Expr };
|
||||
|
||||
export type ProductPattern =
|
||||
| { tag: "any", name: VariableName }
|
||||
| { tag: "tuple", patterns: ProductPattern[] }
|
||||
| { tag: "record", fields: { field_name: FieldName, pattern: ProductPattern }[] }
|
||||
| { tag: "record", fields: FieldPattern[] }
|
||||
|
||||
export type FieldPattern = { fieldName: FieldName, pattern: ProductPattern };
|
||||
|
||||
export type Pattern =
|
||||
| ProductPattern
|
||||
|
|
@ -212,12 +221,16 @@ export namespace Expr {
|
|||
export const tag = (tag_name: Tag): Expr => ({ tag: "tag", tag_name, });
|
||||
export const tagged = (tag_name: Tag, expr: Expr): Expr => ({ tag: "tagged", tag_name, expr, });
|
||||
export const tuple = (exprs: Expr[]): Expr => ({ tag: "tuple", exprs });
|
||||
export const record = (fields: { name: FieldName, expr: Expr }[]): Expr => ({ tag: "record", fields });
|
||||
export const match = (arg: Expr, branches: { pattern: Pattern; body: Expr }[]): Expr => ({ tag: "match", arg, branches, });
|
||||
export const record = (fields: FieldAssignment[]): Expr => ({ tag: "record", fields });
|
||||
export const match = (arg: Expr, branches: MatchBranch[]): Expr => ({ tag: "match", arg, branches, });
|
||||
export const var_use = (name: VariableName): Expr => ({ tag: "var_use", name, });
|
||||
export const let_ = (bindings: ExprBinding[], body: Expr): Expr => ({ tag: "let", bindings, body, });
|
||||
export const apply = (callee: Expr, args: Expr[]): Expr => ({ tag: "apply", callee, args, });
|
||||
export const lambda = (parameters: ProductPattern[], body: Expr): Expr => ({ tag: "lambda", parameters, body, });
|
||||
|
||||
export const matchBranch = (pattern: Pattern, expr: Expr): MatchBranch => ({ pattern, body: expr });
|
||||
export const exprBinding = (pattern: ProductPattern, expr: Expr): ExprBinding => ({ pattern, expr });
|
||||
export const fieldAssignment = (name: FieldName, expr: Expr): FieldAssignment => ({ name, expr });
|
||||
}
|
||||
|
||||
export namespace Value {
|
||||
|
|
@ -230,6 +243,19 @@ export namespace Value {
|
|||
export const closure = (closure: Closure): Value => ({ tag: "closure", closure });
|
||||
}
|
||||
|
||||
export namespace ProductPattern {
|
||||
export const any = (name: VariableName): ProductPattern => ({ tag: "any", name });
|
||||
export const tuple = (patterns: ProductPattern[]): ProductPattern => ({ tag: "tuple", patterns });
|
||||
export const record = (fields: FieldPattern[]): ProductPattern => ({ tag: "record", fields });
|
||||
|
||||
export const fieldPattern = (fieldName: FieldName, pattern: ProductPattern): FieldPattern => ({ fieldName, pattern });
|
||||
}
|
||||
|
||||
export namespace Pattern {
|
||||
export const tag = (tag_name: Tag): Pattern => ({ tag: "tag", tag_name });
|
||||
export const tagged = (tag_name: Tag, pattern: Pattern): Pattern => ({ tag: "tagged", tag_name, pattern });
|
||||
}
|
||||
|
||||
// ===Errors===
|
||||
type RuntimeError =
|
||||
| { tag: "FunctionLookupFailure", name: FunctionName }
|
||||
|
|
@ -369,7 +395,7 @@ function eval_expr(program: Program, env: Env, e: Expr): Value {
|
|||
function eval_bindings(program: Program, env: Env, bindings: ExprBinding[]): Env {
|
||||
// note that `let { x = 123, y = x + 1 ... } is allowed. Ofcourse later bindings can't be referenced by earlier bindings (i.e. no recursion).
|
||||
let cur_env = env;
|
||||
for (const { var: var_name, expr } of bindings) {
|
||||
for (const { pattern: var_name, expr } of bindings) {
|
||||
const value = eval_expr(program, cur_env, expr);
|
||||
const res = match_product_pattern(var_name, value);
|
||||
if (res.tag === "failure") {
|
||||
|
|
@ -514,8 +540,8 @@ function match_product_pattern_mut(frame: EnvFrame, pattern: ProductPattern, val
|
|||
case "record":
|
||||
if (value.tag !== "record") return { tag: "failure", pattern, value };
|
||||
|
||||
for (const { field_name, pattern: p } of pattern.fields) {
|
||||
const field_value = value.fields.get(field_name);
|
||||
for (const { fieldName, pattern: p } of pattern.fields) {
|
||||
const field_value = value.fields.get(fieldName);
|
||||
if (field_value === undefined) {
|
||||
return { tag: "failure", pattern, value };
|
||||
} else {
|
||||
|
|
|
|||
|
|
@ -15,3 +15,5 @@ npm install -D sass-embedded
|
|||
npx ts-node src/parser/cursor.test.ts
|
||||
|
||||
|
||||
npx ts-node src/debug/repl.ts
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue