When parsing more than one file "start" and "end" accumulate values

[lostsequence]

const files = ['file1', 'file2']  
files.forEach(f => {  
    const text = Deno.readTextFileSync(f);  
    const ast = parse(text, {  
		target: "es2020",  
		syntax: "typescript",  
		comments: false,  
		tsx: true,  
	});  
})  

In that case for 'file1' "span.start" for Module will be "0"
But for 'file2' "span.start" will be "span.end" of another file.

Is it possible to have "0" as "span.start" for each file ?

[eibens]

Any update on this?

---

I patched this on my end by running the parser in a worker. It works, but starting up the worker takes about 900 ms on my system.

import { parse } from "swc";

self.postMessage("ready");

self.onmessage = (evt) => {
  try {
    const program = parse(evt.data, {
      target: "es2019",
      syntax: "typescript",
      comments: true,
      tsx: true,
    });
    const ast = JSON.stringify(program);
    self.postMessage(ast);
  } catch (e) {
    self.postMessage({
      error: {
        message: e.message,
      },
    });
  }
};

export function parse(source: string): Promise<string> {
  return new Promise((resolve, reject) => {
    const url = new URL("./parse.worker.js", import.meta.url);
    const worker = new Worker(url.href, { type: "module" });

    worker.addEventListener("message", (event) => {
      if (event.data !== "ready") {
        reject("Worker failed to initialize");
        return;
      }

      worker.addEventListener("message", (event) => {
        resolve(event.data);
      }, { once: true });

      worker.postMessage(source);
    }, { once: true });

    worker.addEventListener("error", (event) => {
      reject(event.error);
    });

    worker.addEventListener("messageerror", (event) => {
      reject(event.data);
    });
  });
}

[eibens]

The problem comes from the main package. Here is the issue.

Unfortunately, most of the solutions there do not currently work. You have to either instantiate the module for each file, or subtract program.span.start from all spans in the AST. I am now using the latter approach, since downloading the WASM file for every parse call caused unacceptable delays and instability on bad networks.

[rizrmd]

What happen when span.start exceeding Number.MAX_SAFE_INTEGER ? Does swc crash ?

[eibens]

@rizrmd You have to parse 9,007,199,254,740,991 bytes (~9 Petabytes) of source code to get to that point. What are you planning to do?

[rizrmd]

Haha, sorry, just curious… I never thought MAX_SAFE_INTEGER is 9 petabytes