readme: formato de repo

.gitignore

@@ -1,3 +1,5 @@
 node_modules/
 dataJsons/
 log
+prueba
+datos.gob.ar*

download_json.js

@@ -1,28 +1,17 @@
 // @ts-check
-import { mkdir, open } from "node:fs/promises";
-import { Agent, fetch } from "undici";
+import { mkdir, open, writeFile } from "node:fs/promises";
+import { Agent, fetch, request, setGlobalDispatcher } from "undici";
 import { join, normalize } from "node:path";
 import { pipeline } from "node:stream/promises";
 
-// lista de dominios que permitimos usar http: porque tienen HTTPS roto..
-const brokenHttps = [
-  "datos.mindef.gov.ar", // cert para otro dominio
-  "datos.energia.gob.ar", // cert para otro dominio
-  "datos.minem.gob.ar", // vencido 2022-17-06
-  "datos.agroindustria.gob.ar", // vencido 2022-03-10
-  "andino.siu.edu.ar", // self signed, igual parece que todo tira 404 en este..
-  "datos.salud.gob.ar", // timeout en HTTPS
-  "datos.jus.gob.ar", // HTTPS redirige incorrectamente a URLs inexistentes
-  "www.hidro.gob.ar", // no HTTPS
-];
-
 // 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
 
-const dispatcher = new Agent({
-  pipelining: 10,
-  maxRedirections: 20,
-});
+setGlobalDispatcher(
+  new Agent({
+    pipelining: 0,
+  })
+);
 
 class StatusCodeError extends Error {
   /**
@@ -33,107 +22,142 @@ class StatusCodeError extends Error {
     this.code = code;
   }
 }
+class TooManyRedirectsError extends Error {}
 
-const outputPath = process.argv[2];
-if (!outputPath) {
-  console.error("Especificamente el output porfa");
+let jsonUrlString = process.argv[2];
+if (!jsonUrlString) {
+  console.error("Especificamente el url al json porfa");
   process.exit(1);
 }
+const jsonUrl = new URL(jsonUrlString);
+const outputPath = jsonUrl.host;
 await mkdir(outputPath, { recursive: true });
+const errorFile = await open(join(outputPath, "errors.jsonl"), "w");
 
-// Leer JSON de stdin
-const json = await process.stdin.toArray();
-const jsonString = json.join("");
-/** @type {{ dataset: Dataset[] }} */
-const parsed = JSON.parse(jsonString);
+const jsonRes = await fetch(jsonUrl);
+// prettier-ignore
+const parsed = /** @type {{ dataset: Dataset[] }} */(await jsonRes.json())
+await writeFile(join(outputPath, "data.json"), JSON.stringify(parsed));
 
 const jobs = parsed.dataset.flatMap((dataset) =>
-  dataset.distribution.map((dist) => ({ dataset, dist })),
+  dataset.distribution.map((dist) => ({
+    dataset,
+    dist,
+    url: patchUrl(new URL(dist.downloadURL)),
+  }))
 );
-// forma barrani de distribuir carga entre servidores
-shuffleArray(jobs);
 const totalJobs = jobs.length;
 let nFinished = 0;
+let nErrors = 0;
 
-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",
-  );
+// por las dudas verificar que no hayan archivos duplicados
+chequearIdsDuplicados();
+
+/** @type {Map< string, DownloadJob[] >} */
+let jobsPerHost = new Map();
+for (const job of jobs) {
+  jobsPerHost.set(job.url.host, [
+    ...(jobsPerHost.get(job.url.host) || []),
+    job,
+  ]);
 }
 
-const greens = Array(128)
+const greens = [...jobsPerHost.entries()].flatMap(([host, jobs]) => {
+  const nThreads = 8;
+  return Array(nThreads)
     .fill(0)
     .map(() =>
       (async () => {
         let job;
         while ((job = jobs.pop())) {
-        const { dataset, dist } = job;
-        request: do {
           try {
-            await downloadDist(dataset, dist);
+            await downloadDistWithRetries(job);
           } catch (error) {
-            if (error instanceof StatusCodeError) {
-              // algunos servidores usan 403 como coso para decir "calmate"
-              if (error.code === 403) {
-                console.debug(
-                  `debug: reintentando ${dist.downloadURL} porque tiró 403`,
+            await errorFile.write(
+              JSON.stringify({
+                url: job.url.toString(),
+                ...encodeError(error),
+              }) + "\n"
             );
-                await wait(15000);
-                continue request;
-              }
-              error = error.toString();
-            }
-            console.error(
-              `error: Failed to download URL ${dist.downloadURL} (${dataset.identifier}/${dist.identifier}):`,
-              error,
-            );
-            if (!(error instanceof StatusCodeError)) continue request;
+            nErrors++;
           } finally {
             nFinished++;
           }
-        } while (0);
         }
-    })(),
+      })()
     );
+});
+process.stderr.write(`greens: ${greens.length}\n`);
 
 const interval = setInterval(() => {
-  console.info(`info: ${nFinished}/${totalJobs} done`);
-}, 15000);
+  process.stderr.write(`info: ${nFinished}/${totalJobs} done\n`);
+}, 30000);
 await Promise.all(greens);
 clearInterval(interval);
+if (nErrors > 0) console.error(`Finished with ${nErrors} errors`);
 
 /**
- * @argument {Dataset} dataset
- * @argument {Distribution} dist
+ * @argument {DownloadJob} job
+ * @argument {number} tries
  */
-async function downloadDist(dataset, dist) {
-  const url = new URL(dist.downloadURL);
+async function downloadDistWithRetries(job, tries = 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" &&
+      tries < 15
+    ) {
+      await wait(15000);
+      return await downloadDistWithRetries(job, tries + 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) &&
+      tries < 5
+    ) {
+      await wait(5000);
+      return await downloadDistWithRetries(job, tries + 1);
+    } else throw error;
+  }
+}
 
-  // Siempre usar HTTPS excepto cuando está roto
-  if (brokenHttps.includes(url.host)) {
-    url.protocol = "http:";
-    // console.debug(url);
-  } else url.protocol = "https:";
+/**
+ * @argument {DownloadJob} job
+ */
+async function downloadDist({ dist, dataset, url }) {
+  // sharepoint no le gusta compartir a bots lol
+  const spoofUserAgent = url.host.endsWith("sharepoint.com");
 
-  const res = await fetch(url.toString(), {
-    dispatcher,
+  const res = await request(url.toString(), {
+    maxRedirections: 20,
+    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.status >= 400) {
-    throw new StatusCodeError(res.status);
+  if (res.statusCode >= 300 && res.statusCode <= 399)
+    throw new TooManyRedirectsError();
+  if (res.statusCode < 200 || res.statusCode > 299) {
+    throw new StatusCodeError(res.statusCode);
   }
 
   const fileDirPath = join(
     outputPath,
     sanitizeSuffix(dataset.identifier),
-    sanitizeSuffix(dist.identifier),
+    sanitizeSuffix(dist.identifier)
   );
   await mkdir(fileDirPath, { recursive: true });
   const filePath = join(
     fileDirPath,
-    sanitizeSuffix(dist.fileName || dist.identifier),
+    sanitizeSuffix(dist.fileName || dist.identifier)
   );
   const outputFile = await open(filePath, "w");
 
@@ -141,11 +165,16 @@ async function downloadDist(dataset, dist) {
   await pipeline(res.body, outputFile.createWriteStream());
 }
 
-/** @typedef {object} Dataset
+/** @typedef DownloadJob
+ * @prop {Dataset} dataset
+ * @prop {Distribution} dist
+ * @prop {URL} url
+ */
+/** @typedef Dataset
  * @prop {string} identifier
  * @prop {Distribution[]} distribution
  */
-/** @typedef {object} Distribution
+/** @typedef Distribution
  * @prop {string} identifier
  * @prop {string} fileName
  * @prop {string} downloadURL
@@ -159,31 +188,46 @@ function sanitizeSuffix(path) {
   return normalize(path).replace(/^(\.\.(\/|\\|$))+/, "");
 }
 
+function chequearIdsDuplicados() {
+  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"
+    );
+  }
+}
 // https://stackoverflow.com/a/7376645
-/**
- * @argument {any[]} array
- */
+/** @argument {any[]} array */
 function hasDuplicates(array) {
   return new Set(array).size !== array.length;
 }
 
-// https://stackoverflow.com/a/12646864
-/**
- * @argument {any[]} array
- */
-function shuffleArray(array) {
-  for (var i = array.length - 1; i > 0; i--) {
-    var j = Math.floor(Math.random() * (i + 1));
-    var temp = array[i];
-    array[i] = array[j];
-    array[j] = temp;
-  }
-}
-
-/**
- * @argument {number} ms
- */
+/** @argument {number} ms */
 function wait(ms) {
   if (ms < 0) return Promise.resolve();
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
+
+function encodeError(error) {
+  if (error instanceof StatusCodeError)
+    return { kind: "http_error", status_code: error.code };
+  else if (error instanceof TooManyRedirectsError)
+    return { kind: "infinite_redirect" };
+  else {
+    return { kind: "generic_error", error: error.code || error.message };
+  }
+}
+
+/**
+ * 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;
+}

package.json

@@ -13,5 +13,8 @@
   "dependencies": {
     "node_extra_ca_certs_mozilla_bundle": "^1.0.5",
     "undici": "^5.28.0"
+  },
+  "devDependencies": {
+    "@types/node": "^20.10.0"
   }
 }

pnpm-lock.yaml

@@ -12,6 +12,11 @@ dependencies:
     specifier: ^5.28.0
     version: 5.28.0
 
+devDependencies:
+  '@types/node':
+    specifier: ^20.10.0
+    version: 20.10.0
+
 packages:
 
   /@fastify/busboy@2.1.0:
@@ -19,6 +24,12 @@ packages:
     engines: {node: '>=14'}
     dev: false
 
+  /@types/node@20.10.0:
+    resolution: {integrity: sha512-D0WfRmU9TQ8I9PFx9Yc+EBHw+vSpIub4IDvQivcp26PtPrdMGAq5SDcpXEo/epqa/DXotVpekHiLNTg3iaKXBQ==}
+    dependencies:
+      undici-types: 5.26.5
+    dev: true
+
   /asynckit@0.4.0:
     resolution: {integrity: sha512-Oei9OH4tRh0YqU3GxhX79dM/mwVgvbZJaSNaRk+bshkj0S5cfHcgYakreBjrHwatXKbz+IoIdYLxrKim2MjW0Q==}
     dev: false
@@ -153,6 +164,10 @@ packages:
       is-utf8: 0.2.1
     dev: false
 
+  /undici-types@5.26.5:
+    resolution: {integrity: sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA==}
+    dev: true
+
   /undici@5.28.0:
     resolution: {integrity: sha512-gM12DkXhlAc5+/TPe60iy9P6ETgVfqTuRJ6aQ4w8RYu0MqKuXhaq3/b86GfzDQnNA3NUO6aUNdvevrKH59D0Nw==}
     engines: {node: '>=14.0'}

readme.md

@@ -9,5 +9,15 @@ pnpm install
 ## correr
 
 ```
-pnpm run run download_json.js carpeta_output < dataJsons/datos.gob.ar.data.json
+pnpm run run download_json.js https://datos.gob.ar/data.json
+# guarda en ./datos.gob.ar
 ```
+
+## formato de repo guardado
+
+- `{dominio de repo}`
+  - `data.json`
+  - `errors.jsonl`
+  - `{identifier de dataset}`
+    - `{identifier de distribution}`
+      - `{fileName (o, si no existe, identifier de distribution)}`