Use the native lexer instead of a TS reimplementation

master
Araozu 2024-07-05 19:12:30 -05:00
parent 49faed4fcb
commit a542071af9
12 changed files with 148 additions and 469 deletions

1
.env.example Normal file
View File

@ -0,0 +1 @@
THP_BINARY=/path/to/rust/thp/binary

View File

@ -1,10 +1,10 @@
--- ---
import { leftTrimDedent } from "./utils"; import { native_highlighter } from "../lexer/highlighter";
import { thp_highlighter } from "../lexer/highlighter";
const { thpcode } = Astro.props; const { thpcode } = Astro.props;
const html_code = thp_highlighter(leftTrimDedent(thpcode).join("\n"));
const native_html = await native_highlighter(thpcode);
--- ---
<pre <pre
class="language-thp"><code class="language-thp" set:html={html_code} /><span class="absolute top-2 right-2 inline-block text-sm select-none opacity-75">thp</span></pre> class="language-thp"><code class="language-thp" set:html={native_html} /><span class="absolute top-2 right-2 inline-block text-sm select-none opacity-75">thp</span></pre>

View File

@ -1,13 +1,148 @@
import { lex } from "./lexer"; import { spawn } from "node:child_process";
import { leftTrimDedent } from "../components/utils";
export function thp_highlighter(code: string) { export interface LexResult {
let tokens = lex(code); Ok?: Token[]
Err?: Err
let highlighted_code = "";
for (let token of tokens) {
highlighted_code += `<span class="token ${token.token_type}">${token.v}</span>`;
} }
return highlighted_code; export interface Token {
token_type: TokenType
value: string
position: number
} }
type TokenType =
"Identifier" |
"Datatype" |
"Int" |
"Float" |
"String" |
"Operator" |
"LeftParen" |
"RightParen" |
"LeftBracket" |
"RightBracket" |
"LeftBrace" |
"RightBrace" |
"NewLine" |
"Comment" |
"Comma" |
"INDENT" |
"DEDENT" |
"VAL" |
"VAR" |
"EOF" |
"FUN";
export interface Err {
Lex: LexError
}
export interface LexError {
position: number
reason: string
}
export async function native_highlighter(code: string): Promise<string> {
let formatted_code = leftTrimDedent(code).join("\n");
const result = await native_lex(formatted_code);
if (result.Err) {
throw new Error(JSON.stringify(result.Err.Lex) + "\n" + code);
}
const tokens = result.Ok!;
const input_chars = formatted_code.split("");
let output = "";
let current_pos = 0;
for (let i = 0; i < tokens.length; i += 1) {
const t = tokens[i]!;
const token_start = t.position;
const token_end = t.position + t.value.length;
// There are some tokens that are empty, ignore them
if (t.value == "") {
continue;
}
// Append all characters before the token
output += input_chars.slice(current_pos, token_start).join("");
// Append the token
const token_value = t.value.replaceAll(/</g, "&lt;").replaceAll(/>/g, "&gt;");
const token_type = translate_token_type(t.token_type, token_value);
output += `<span class="token ${token_type}">${token_value}</span>`;
current_pos = token_end;
}
return output;
}
function translate_token_type(tt: TokenType, value: string): string {
const keywords = ["throws", "extends", "constructor", "case", "static", "const", "enum", "union", "loop", "use", "break", "catch", "continue", "as", "do", "else", "finally", "for", "fun", "if", "in", "fn", "nil", "return", "throw", "try", "while", "type", "match", "with", "of", "abstract", "class", "interface", "private", "pub", "override", "open", "init", "val", "var", "mut", "clone"];
switch (tt) {
case "Datatype":
return "class-name";
case "Identifier": {
if (keywords.includes(value)) {
return "keyword";
}
return "identifier";
}
case "Int":
return "number";
case "Float":
return "number";
case "String":
return "string";
case "Comment":
return "comment";
// keywords:
case "VAL":
case "VAR":
case "FUN":
return "keyword";
default:
return tt;
}
}
const native_lex = (code: string) => new Promise<LexResult>((resolve, reject) => {
// Get binary path from .env
const binary = import.meta.env.THP_BINARY;
if (!binary) {
throw new Error("THP_BINARY not set in .env");
}
const subprocess = spawn(binary, ["tokenize"]);
let response = "";
let error = "";
subprocess.stdin.write(code);
subprocess.stdin.end();
subprocess.stdout.on("data", (data) => {
response += data.toString();
});
subprocess.stderr.on("data", (data) => {
error += data.toString();
});
subprocess.on("close", (code) => {
if (code === 0) {
resolve(JSON.parse(response));
} else {
reject(error);
}
});
})

View File

@ -1,55 +0,0 @@
import { expect, test, describe } from "vitest";
import { scan_identifier } from "./identifier_lexer";
describe("Identifier Lexer", () => {
test("should return an identifier token", () => {
const code = "a";
const token = scan_identifier(code, 0);
expect(token).toEqual([{ v: "a", token_type: "identifier" }, 1]);
});
test("should scan an underscore", () => {
const code = "_";
const token = scan_identifier(code, 0);
expect(token).toEqual([{ v: "_", token_type: "identifier" }, 1]);
});
test("should scan an identifier with an underscore", () => {
const code = "a_";
const token = scan_identifier(code, 0);
expect(token).toEqual([{ v: "a_", token_type: "identifier" }, 2]);
});
test("should scan an identifier that starts with an underscore", () => {
const code = "_a";
const token = scan_identifier(code, 0);
expect(token).toEqual([{ v: "_a", token_type: "identifier" }, 2]);
});
test("should scan an identifier with numbers and uppercase letters", () => {
const code = "aA1";
const token = scan_identifier(code, 0);
expect(token).toEqual([{ v: "aA1", token_type: "identifier" }, 3]);
});
test("should scan a keyword", () => {
const code = "val";
const token = scan_identifier(code, 0);
expect(token).toEqual([{ v: "val", token_type: "keyword" }, 3]);
});
test("should scan a datatype", () => {
const code = "Int";
const token = scan_identifier(code, 0, true);
expect(token).toEqual([{ v: "Int", token_type: "class-name" }, 3]);
});
});

View File

@ -1,44 +0,0 @@
import type { Token } from "./lexer";
import { is_identifier_char } from "./utils";
/**
* Scans an identifier, at the given position in the input string.
* This function assumes that the character at the given position is a letter.
*
* @param input the input string
* @param starting_position the position to start scanning from
* @param is_datatype whether the identifier is a datatype
*/
export function scan_identifier(input: string, starting_position: number, is_datatype = false): [Token, number] {
let value = input[starting_position]!;
let pos = starting_position + 1;
while (pos < input.length) {
const c = input[pos]!;
if (is_identifier_char(c)) {
pos += 1;
value += c;
}
else {
break;
}
}
if (is_datatype) {
return [{ v: value, token_type: "class-name" }, pos];
}
else {
return [{ v: value, token_type: check_keyword(value) }, pos];
}
}
function check_keyword(value: string): string {
const keywords = ["throws", "extends", "constructor", "case", "static", "const", "enum", "union", "loop", "use", "break", "catch", "continue", "as", "do", "else", "finally", "for", "fun", "if", "in", "fn", "nil", "return", "throw", "try", "while", "type", "match", "with", "of", "abstract", "class", "interface", "private", "pub", "override", "open", "init", "val", "var", "mut", "clone"];
if (keywords.includes(value)) {
return "keyword";
}
return "identifier";
}

View File

@ -1,45 +0,0 @@
import { expect, test, describe } from "vitest";
import { lex } from "./lexer";
describe("Lexer", () => {
test("empty program should return no tokens", () => {
const code = "";
const tokens = lex(code);
expect(tokens).toEqual([]);
});
test("program with whitespace should return a single token", () => {
const code = " ";
const tokens = lex(code);
expect(tokens).toEqual([{v: " ", token_type: ""}]);
})
test("program with newlines should return a single token", () => {
const code = "\n";
const tokens = lex(code);
expect(tokens).toEqual([{v: "\n", token_type: ""}]);
});
test("program with random unicode should return the same unicode", () => {
const code = "🍕";
const tokens = lex(code);
expect(tokens).toEqual([{v: "🍕", token_type: ""}]);
});
test("should scan integers", () => {
const code = "12345";
const tokens = lex(code);
expect(tokens).toEqual([{v: "12345", token_type: "number"}]);
});
test("should scan integers and whitespace around", () => {
const code = " 12345 \n ";
const tokens = lex(code);
expect(tokens).toEqual([
{v: " ", token_type: ""},
{v: "12345", token_type: "number"},
{v: " \n ", token_type: ""},
]);
});
});

View File

@ -1,166 +0,0 @@
import { scan_identifier } from "./identifier_lexer";
import { scan_number } from "./number_lexer";
import { scan_string } from "./string_lexer";
import { is_digit, is_lowercase, is_uppercase } from "./utils";
export type Token = {
v: string,
token_type: string,
};
/**
* Lexes a string of THP code, and returns an array of tokens. Unlike a regular
* lexer, whitespace and other characters are not ignored, and are instead treated
* as a default token.
*
* This lexer implements a subset of the grammar defined in the THP language specification,
* only recognizing the following tokens:
* - Identifier
* - Datatype
* - String
* - Number
* - Single line comment
* - Multi line comment
* - Keywords
*
* @param code Code to lex
* @returns An array of all the tokens found
*/
export function lex(code: string, start = 0): Array<Token> {
const code_len = code.length;
const tokens: Array<Token> = [];
let current_pos = start;
let current_default_token = "";
while (current_pos < code_len) {
const c = code[current_pos]!;
// try to scan a number
if (is_digit(c)) {
// if the current default token is not empty, push it to the tokens array
if (current_default_token !== "") {
tokens.push({ v: current_default_token, token_type: "" });
current_default_token = "";
}
// lex a number
const [token, next] = scan_number(code, current_pos);
current_pos = next;
tokens.push(token);
continue;
}
// try to scan an identifier/keyword
else if (is_lowercase(c) || c === "_") {
// if the current default token is not empty, push it to the tokens array
if (current_default_token !== "") {
tokens.push({ v: current_default_token, token_type: "" });
current_default_token = "";
}
const [token, next] = scan_identifier(code, current_pos);
current_pos = next;
tokens.push(token);
continue;
}
// try to scan a datatype
else if (is_uppercase(c)) {
// if the current default token is not empty, push it to the tokens array
if (current_default_token !== "") {
tokens.push({ v: current_default_token, token_type: "" });
current_default_token = "";
}
const [token, next] = scan_identifier(code, current_pos, true);
current_pos = next;
tokens.push(token);
continue;
}
// try to scan a string
else if (c === "\"") {
// if the current default token is not empty, push it to the tokens array
if (current_default_token !== "") {
tokens.push({ v: current_default_token, token_type: "" });
current_default_token = "";
}
const [token, next] = scan_string(code, current_pos);
current_pos = next;
tokens.push(token);
continue;
}
// try to scan a comment
else if (c === "/" && code[current_pos + 1] === "/") {
// if the current default token is not empty, push it to the tokens array
if (current_default_token !== "") {
tokens.push({ v: current_default_token, token_type: "" });
current_default_token = "";
}
let comment = "";
let pos = current_pos;
while (pos < code_len) {
const char = code[pos];
if (char === "\n") {
break;
}
comment += char;
pos++;
}
tokens.push({ v: comment, token_type: "comment" });
current_pos = pos;
continue;
}
// try to scan a multiline comment
else if (c === "/" && code[current_pos + 1] === "*") {
// if the current default token is not empty, push it to the tokens array
if (current_default_token !== "") {
tokens.push({ v: current_default_token, token_type: "" });
current_default_token = "";
}
let comment = "";
let pos = current_pos;
while (pos < code_len) {
const char = code[pos];
if (char === "*" && code[pos + 1] === "/") {
pos += 2;
comment += "*/";
break;
}
comment += char;
pos++;
}
tokens.push({ v: comment, token_type: "comment" });
current_pos = pos;
continue;
}
// replace < with &lt;
else if (c === "<") {
current_default_token += "&lt;";
current_pos++;
continue;
}
current_default_token += c;
current_pos++;
}
// if there was a default token, push it to the tokens array
if (current_default_token !== "") {
tokens.push({ v: current_default_token, token_type: "" });
current_default_token = "";
}
return tokens;
}

View File

@ -1,19 +0,0 @@
import { expect, test, describe } from "vitest";
import { scan_number } from "./number_lexer";
describe("Number Lexer", () => {
test("should return a whole number token", () => {
const code = "1";
const token = scan_number(code, 0);
expect(token).toEqual([{ v: "1", token_type: "number" }, 1]);
});
test("should return a whole number token pt 2", () => {
const code = "12345";
const token = scan_number(code, 0);
expect(token).toEqual([{ v: "12345", token_type: "number" }, 5]);
});
});

View File

@ -1,47 +0,0 @@
import type { Token } from "./lexer";
import { is_digit } from "./utils";
/**
* Scans a number, at the given position in the input string.
* This function assumes that the character at the given position is a digit.
* It follows this grammar:
*
* @param input the input string
* @param pos the position to start scanning from
* @returns
*/
export function scan_number(input: string, pos: number): [Token, number] {
const [token_value, next] = scan_decimal(input, pos);
return [{ v: token_value, token_type: "number" }, next];
}
function scan_decimal(input: string, starting_position: number): [string, number] {
let current_value = "";
let pos = starting_position;
while (pos < input.length) {
const c = input[pos]!;
if (c === ".") {
// todo
return [current_value, pos];
}
else if (c == "e" || c == "E") {
// todo
return [current_value, pos];
}
else if (is_digit(c)) {
current_value += c;
pos += 1;
}
else {
break;
}
}
return [current_value, pos];
}

View File

@ -1,32 +0,0 @@
import { expect, test, describe } from "vitest";
import { scan_string } from "./string_lexer";
describe("String Lexer", () => {
test("should scan an empty string", () => {
const code = "\"\"";
const token = scan_string(code, 0);
expect(token).toEqual([{ v: "\"\"", token_type: "string" }, 2]);
});
test("should scan a string with a single character", () => {
const code = "\"a\"";
const token = scan_string(code, 0);
expect(token).toEqual([{ v: "\"a\"", token_type: "string" }, 3]);
});
test("should scan a string with multiple characters", () => {
const code = "\"hello\"";
const token = scan_string(code, 0);
expect(token).toEqual([{ v: "\"hello\"", token_type: "string" }, 7]);
});
test("should scan a string with an escape character", () => {
const code = "\"\\n\"";
const token = scan_string(code, 0);
expect(token).toEqual([{ v: "\"\\n\"", token_type: "string" }, 4]);
});
});

View File

@ -1,49 +0,0 @@
import type { Token } from "./lexer";
export function scan_string(input: string, starting_position: number): [Token, number] {
let value = "\"";
let pos = starting_position + 1;
while (pos < input.length) {
const c = input[pos];
if (c === "\"") {
value += c;
pos += 1;
break;
}
if (c === "\n") {
// todo: error handling, return an error indicator and the caller should render a red wavy underline
break;
}
if (c === "\\") {
const next_char = input[pos + 1];
value += handle_escape_char(next_char);
pos += 2;
continue;
}
value += c;
pos += 1;
}
return [{ v: value, token_type: "string" }, pos];
}
function handle_escape_char(next_char: string): string {
switch (next_char) {
case "n":
return "\\n"
case "t":
return "\\t"
case "r":
return "\\r"
case "\"":
return "\\\""
case "\\":
return "\\\\"
default:
return "\\" + next_char
}
}

View File

@ -196,7 +196,7 @@ $cat->meow();
<Code thpcode={` <Code thpcode={`
// THP // THP
val cat = Cat("Michifu", 7) val cat = Cat("Michifu", 7)
cat.meow(); cat.meow()
`} /> `} />
- Instantiate classes without `new` - Instantiate classes without `new`