transicion-desordenada-diablo/download_json.js

252 lines
7 KiB
JavaScript
Raw Normal View History

2023-11-27 20:01:56 +00:00
// @ts-check
import { mkdir, open, writeFile } from "node:fs/promises";
import { Agent, fetch, request, setGlobalDispatcher } from "undici";
2023-11-27 20:01:56 +00:00
import { join, normalize } from "node:path";
2023-11-28 22:19:35 +00:00
import pLimit from "p-limit";
2023-11-27 20:01:56 +00:00
// FYI: al menos los siguientes dominios no tienen la cadena completa de certificados en HTTPS. tenemos que usar un hack (node_extra_ca_certs_mozilla_bundle) para conectarnos a estos sitios. (se puede ver con ssllabs.com) ojalá lxs administradorxs de estos servidores lo arreglen.
// www.enargas.gov.ar, transparencia.enargas.gov.ar, www.energia.gob.ar, www.economia.gob.ar, datos.yvera.gob.ar
setGlobalDispatcher(
new Agent({
pipelining: 0,
})
);
2023-11-27 20:01:56 +00:00
2023-11-28 22:19:35 +00:00
/** key es host
* @type {Map<string, import("p-limit").LimitFunction>} */
const limiters = new Map();
const nThreads = process.env.N_THREADS ? parseInt(process.env.N_THREADS) : 16;
2023-11-27 20:01:56 +00:00
class StatusCodeError extends Error {
/**
* @param {number} code
*/
constructor(code) {
super(`Status code: ${code}`);
this.code = code;
}
}
class TooManyRedirectsError extends Error {}
2023-11-27 20:01:56 +00:00
let jsonUrlString = process.argv[2];
if (!jsonUrlString) {
console.error("Especificamente el url al json porfa");
2023-11-27 20:01:56 +00:00
process.exit(1);
}
downloadFromData(jsonUrlString);
/**
* @param {string} jsonUrlString
*/
async function downloadFromData(jsonUrlString) {
const jsonUrl = new URL(jsonUrlString);
const outputPath = jsonUrl.host;
await mkdir(outputPath, { recursive: true });
const errorFile = await open(join(outputPath, "errors.jsonl"), "w");
const jsonRes = await fetch(jsonUrl);
// prettier-ignore
const parsed = /** @type {{ dataset: Dataset[] }} */(await jsonRes.json())
await writeFile(join(outputPath, "data.json"), JSON.stringify(parsed));
2023-11-27 20:01:56 +00:00
/** @type {DownloadJob[]} */
const jobs = parsed.dataset.flatMap((dataset) =>
dataset.distribution.map((dist) => ({
dataset,
dist,
url: patchUrl(new URL(dist.downloadURL)),
outputPath,
attempts: 0,
}))
);
const totalJobs = jobs.length;
let nFinished = 0;
let nErrors = 0;
// por las dudas verificar que no hayan archivos duplicados
chequearIdsDuplicados(jobs);
shuffleArray(jobs);
2023-11-28 22:19:35 +00:00
const promises = jobs.map((job) => {
let limit = limiters.get(job.url.host);
if (!limit) {
limit = pLimit(nThreads);
limiters.set(job.url.host, limit);
}
return limit(async () => {
try {
await downloadDistWithRetries(job);
} catch (error) {
await errorFile.write(
JSON.stringify({
url: job.url.toString(),
...encodeError(error),
}) + "\n"
);
nErrors++;
} finally {
nFinished++;
}
});
});
2023-11-27 20:01:56 +00:00
2023-11-28 22:19:35 +00:00
process.stderr.write(`info: 0/${totalJobs} done\n`);
const interval = setInterval(() => {
process.stderr.write(`info: ${nFinished}/${totalJobs} done\n`);
}, 30000);
2023-11-28 22:19:35 +00:00
await Promise.all(promises);
clearInterval(interval);
if (nErrors > 0) console.error(`Finished with ${nErrors} errors`);
2023-11-28 01:43:58 +00:00
}
2023-11-28 22:19:35 +00:00
/**
* @argument {DownloadJob} job
* @argument {number} attempts
*/
async function downloadDistWithRetries(job, attempts = 0) {
const { url } = job;
try {
await downloadDist(job);
} catch (error) {
// algunos servidores usan 403 como coso para decir "calmate"
// intentar hasta 15 veces con 15 segundos de por medio
if (
error instanceof StatusCodeError &&
error.code === 403 &&
url.host === "minsegar-my.sharepoint.com" &&
attempts < 15
) {
await wait(15000);
return await downloadDistWithRetries(job, attempts + 1);
}
// si no fue un error de http, reintentar hasta 5 veces con 5 segundos de por medio
else if (
!(error instanceof StatusCodeError) &&
!(error instanceof TooManyRedirectsError) &&
attempts < 5
) {
await wait(5000);
return await downloadDistWithRetries(job, attempts + 1);
} else throw error;
}
}
2023-11-28 01:43:58 +00:00
/**
* @argument {DownloadJob} job
*/
async function downloadDist({ dist, dataset, url, outputPath }) {
2023-11-28 03:32:53 +00:00
// sharepoint no le gusta compartir a bots lol
const spoofUserAgent = url.host.endsWith("sharepoint.com");
const res = await request(url.toString(), {
maxRedirections: 20,
2023-11-28 03:32:53 +00:00
headers: {
"User-Agent": spoofUserAgent
? "Mozilla/5.0 (X11; Linux x86_64; rv:120.0) Gecko/20100101 Firefox/120.0"
: "transicion-desordenada (https://nulo.ar)",
},
});
if (res.statusCode >= 300 && res.statusCode <= 399)
throw new TooManyRedirectsError();
if (res.statusCode < 200 || res.statusCode > 299) {
throw new StatusCodeError(res.statusCode);
2023-11-27 20:01:56 +00:00
}
const fileDirPath = join(
outputPath,
sanitizeSuffix(dataset.identifier),
2023-11-27 23:10:57 +00:00
sanitizeSuffix(dist.identifier)
2023-11-27 20:01:56 +00:00
);
await mkdir(fileDirPath, { recursive: true });
const filePath = join(
fileDirPath,
2023-11-27 23:10:57 +00:00
sanitizeSuffix(dist.fileName || dist.identifier)
2023-11-27 20:01:56 +00:00
);
if (!res.body) throw new Error("no body");
await writeFile(filePath, res.body);
2023-11-27 20:01:56 +00:00
}
2023-11-28 01:43:58 +00:00
/** @typedef DownloadJob
* @prop {Dataset} dataset
* @prop {Distribution} dist
* @prop {URL} url
* @prop {string} outputPath
* @prop {number} attempts
* @prop {Date=} waitUntil
2023-11-28 01:43:58 +00:00
*/
/** @typedef Dataset
2023-11-27 20:01:56 +00:00
* @prop {string} identifier
* @prop {Distribution[]} distribution
*/
2023-11-28 01:43:58 +00:00
/** @typedef Distribution
2023-11-27 20:01:56 +00:00
* @prop {string} identifier
* @prop {string} fileName
* @prop {string} downloadURL
*/
// https://security.stackexchange.com/a/123723
/**
* @argument {string} path
*/
function sanitizeSuffix(path) {
return normalize(path).replace(/^(\.\.(\/|\\|$))+/, "");
}
/**
* @param {DownloadJob[]} jobs
*/
function chequearIdsDuplicados(jobs) {
2023-11-28 01:43:58 +00:00
const duplicated = hasDuplicates(
jobs.map((j) => `${j.dataset.identifier}/${j.dist.identifier}`)
);
if (duplicated) {
console.error(
"ADVERTENCIA: ¡encontré duplicados! es posible que se pisen archivos entre si"
);
}
}
2023-11-27 20:01:56 +00:00
// https://stackoverflow.com/a/7376645
2023-11-28 01:43:58 +00:00
/** @argument {any[]} array */
2023-11-27 20:01:56 +00:00
function hasDuplicates(array) {
return new Set(array).size !== array.length;
}
2023-11-28 01:43:58 +00:00
/** @argument {number} ms */
2023-11-27 20:01:56 +00:00
function wait(ms) {
return new Promise((resolve) => setTimeout(resolve, ms));
}
2023-11-28 01:43:58 +00:00
function encodeError(error) {
if (error instanceof StatusCodeError)
return { kind: "http_error", status_code: error.code };
else if (error instanceof TooManyRedirectsError)
return { kind: "infinite_redirect" };
2023-11-28 01:43:58 +00:00
else {
return { kind: "generic_error", error: error.code || error.message };
2023-11-28 01:43:58 +00:00
}
}
2023-11-28 03:41:25 +00:00
/**
* parchea URLs que se rompen solas
* @param {URL} url
*/
function patchUrl(url) {
if (url.host === "www.ign.gob.ar") {
// por defecto, 'http://www.ign.gob.ar' redirige a 'https://ign.gob.ar' pero su certificado solo aplica para '*.ign.gob.ar'. se sirve todo el contenido correctamente en 'https://www.ign.gob.ar', así que vamos para ahí.
url.protocol = "https:";
}
return url;
}
// https://stackoverflow.com/a/12646864
/** @param {any[]} array */
function shuffleArray(array) {
for (let i = array.length - 1; i > 0; i--) {
const j = Math.floor(Math.random() * (i + 1));
[array[i], array[j]] = [array[j], array[i]];
}
}