From 92c1b3abe15b82486427ef2e9e2455524e0c6c84 Mon Sep 17 00:00:00 2001
From: lyg <1543117173@qq.com>
Date: 星期五, 14 六月 2024 23:01:48 +0800
Subject: [PATCH] 回写图书信息到Excel
---
src/main.mjs | 230 ++++++++++++++++++++++++++++++++++++++++++++++++++-------
1 files changed, 201 insertions(+), 29 deletions(-)
diff --git a/src/main.mjs b/src/main.mjs
index 0e0a42f..bbf02b6 100644
--- a/src/main.mjs
+++ b/src/main.mjs
@@ -8,7 +8,10 @@
import { Worker, isMainThread, parentPort, workerData, threadId } from 'worker_threads';
import { HttpsProxyAgent } from "https-proxy-agent";
import { resolve } from "path";
-
+import { execFileSync } from "child_process";
+import wordsjs from 'wordlist-js';
+import usPlaceList from "./us-place-list.mjs";
+import usPeronNameList from "./us-pseron-name-list.mjs";
/*-------------璇诲彇閰嶇疆---------------*/
let config = JSON.parse(fs.readFileSync('./config.json'));
@@ -19,7 +22,7 @@
if (!fs.existsSync('./logs')) {
fs.mkdirSync('./logs', { recursive: true });
}
- logFile = fs.createWriteStream(`./logs/logs-${config.startRow}-${config.endRow}-thread${threadId}.log`, { flags: 'a', encoding: 'utf8' });
+ logFile = fs.createWriteStream(`./logs/logs-thread${threadId}.log`, { flags: 'a', encoding: 'utf8' });
console.log = function (...text) {
text = `${new Date().toLocaleString()} ${text.join(' ') ?? ''}`;
_log(text);
@@ -33,6 +36,135 @@
proxy: false,
httpsAgent,
});
+
+function allWords() {
+ const words = {};
+ wordsjs.usPlaces = usPlaceList;
+ wordsjs.usPeronNameList = usPeronNameList;
+ for (const key in wordsjs.default) {
+ if (Object.hasOwnProperty.call(wordsjs.default, key)) {
+ for (const word of wordsjs.default[key]) {
+ words[word] = true;
+ }
+ }
+ }
+ return words;
+}
+
+const wordsMap = allWords();
+
+/**
+ * 缁熻鍗曡瘝鏁伴噺
+ * @param {string} str 瀛楃涓�
+ * @returns 鍗曡瘝鏁伴噺
+ */
+function countWordSize(str) {
+ let count = 0;
+ str = str.replace(/[ ]{2,}/g, ' ');
+ for (let i = 0; i < str.length; i++) {
+ if (str[i] === ' ') {
+ count++;
+ }
+ }
+ return count;
+}
+
+/**
+ * 鑾峰彇閿欒鍗曡瘝姣斾緥
+ * @param {string} text 鏂囨湰
+ * @returns 閿欒鍗曡瘝姣斾緥
+ */
+function incorrectWordRatio(text) {
+ text = text.replace(/[ ]+/g, ' ').replace(/([a-zA-Z])[\.!?,;"')]/g, "$1");
+ const words = text.split(' ');
+ const incorrectWordCnt = words.filter(word => !wordsMap[word.toLocaleLowerCase()] && !/\d+/g.test(word)).length;
+ return incorrectWordCnt / words.length;
+}
+
+/**
+ * 绗﹀彿鍗犳瘮 0 ~ 1
+ * @param {string} text 鏂囨湰
+ */
+function symbolRatio(text) {
+ // 闈炲瓧姣嶆暟瀛楀瓧绗﹀崰姣�
+ return (text.match(/[^a-zA-Z0-9 ]/g) || []).length / text.length;
+}
+
+/**
+ * 娓呯悊鏂囨湰
+ * @param {string} text 瑕佹竻鐞嗙殑鏂囨湰
+ */
+function cleanText(text) {
+ text = text.replace(/(\r)/g, '');
+ const googlePage = text.substring(0, 10000);
+ if (googlePage.includes('google')) {
+ text = googlePage.replace(/^(.|\n)*books[ ]*\.[ ]*google[ ]*\.[ ]*com/ig, '') + text.substring(10000);
+ }
+ // if (!/.{170,}/g.test(text) || text.includes('google')) {
+ text = text.replace(/[ ]{2,}/g, ' ')
+ if (!/.{170,}/g.test(text)) {
+ // 姣忚涓嶈秴杩�170涓瓧绗�
+ text = text.replace(/(.{170,})\n/g, '$1');
+ }
+ text = text.replace(/\n+/g, '\n');
+ text = text.replace(/-\n/g, '-');
+ const lines = text.split('\n');
+ const result = [];
+ for (const line of lines) {
+ // 绗﹀彿姣斿お楂樼殑涓嶈
+ const incorrectRatio = incorrectWordRatio(line);
+ if (symbolRatio(line) > 0.2) {
+ if (incorrectRatio > 0.65) {
+ continue;
+ }
+ }
+ // 鍘婚櫎绌烘牸鍚� 杩炵画閲嶅鍗曚釜瀛楃3娆″強浠ヤ笂涓嶈
+ const wordSize = countWordSize(line);
+ if (/([\D])\1{2,}/.test(line.replace(/[ ]+/g, ''))) {
+ if (wordSize < 5 || incorrectRatio > 0.65) {
+ continue;
+ }
+ }
+ // 杩炵画涓変釜鏍囩偣绗﹀彿鍙婁互涓�,閿欒鐜囧ぇ浜�0.65涓嶈
+ if (incorrectRatio > 0.65 && /([\.,'";:|!@#$%^&*\(\)<>?`~鈥�*卢禄芦]){3,}/.test(line)) {
+ continue;
+ }
+ // 鍗曡瘝鏁伴噺澶皯鐨勪笉瑕�
+ if (wordSize > 5 && incorrectRatio > 0.65) {
+ continue;
+ }
+ // 鏈塯oogle鐨勪笉瑕�
+ if (/.*(google).*/ig.test(line)) {
+ continue;
+ }
+ // 鍙湁涓�涓瓧绗︿笉瑕�
+ const ret = line.trim().replace(/[鈻犫��*卢禄芦^-]/g, '');
+ if (ret.length <= 1) {
+ continue;
+ }
+ if (ret == 'Digitized by') {
+ continue;
+ }
+ result.push(ret);
+ }
+ text = result.join('\n');
+ // }
+ return text;
+}
+
+/**
+ * 瑙ｅ帇鏂囨湰鏂囦欢
+ * @param {string} zipFile 鍘嬬缉鏂囦欢璺緞
+ * @param {string} txtFile 鏂囨湰鏂囦欢璺緞
+ */
+function unzip(zipFile, txtFile) {
+ const tmpdir = `./tmpdir/${threadId}`;
+ execFileSync('./7za.exe', ['x', '-aoa', zipFile, `-o${tmpdir}`])
+ const file = fs.readdirSync(tmpdir).map(file => ({ size: fs.statSync(`${tmpdir}/${file}`), name: file }))
+ .sort((a, b) => a.size.size - b.size.size).pop();
+ fs.cpSync(`${tmpdir}/${file.name}`, txtFile, { overwrite: true });
+ fs.rmSync(`${tmpdir}`, { recursive: true });
+}
/**
* 鑾峰彇瑕佷笅杞界啛鍥句功淇℃伅
@@ -99,14 +231,12 @@
* @returns 澶勭悊鍚庣殑鍏抽敭瀛�
*/
function formatKw(text, titleWithNumbers) {
- // 鍙繚鐣欑┖鏍笺�佷腑鏂囥�佽嫳鏂囥�佹硶鏂囥�佸痉鏂囥�佸笇鑵婃枃
- const regex = /[^\u4e00-\u9fa5\w\s\d]/g;
if (titleWithNumbers) {
- text = text.replace(/[^\u4e00-\u9fa5a-zA-Z\u00c0-\u024f \d]/g, "");
+ text = text;
} else {
- text = text.replace(/[^\u4e00-\u9fa5a-zA-Z\u00c0-\u024f ]/g, "");
+ text = text.replace(/[\d]/g, "");
}
- text = text.split(' ').slice(0, 10).join("+");
+ text = text.split(' ').slice(0, 6).join("+");
return text;
}
@@ -135,10 +265,10 @@
* @param {*} book
*/
async function openSearchPage(book, titleWithNumbers) {
- console.log(`鎵撳紑鎼滅储: https://archive.org/search?query=${formatKw(book.title, titleWithNumbers)}`);
+ console.log(`鎵撳紑鎼滅储: https://archive.org/search?query=${formatKw(book.title, titleWithNumbers)}&sin=TXT`);
return await retry(async () => {
// 鑾峰彇椤甸潰
- const searchUrl = `https://archive.org/search?query=${formatKw(book.title, titleWithNumbers)}`;
+ const searchUrl = `https://archive.org/search?query=${formatKw(book.title, titleWithNumbers)}&sin=TXT`;
await driver.get(searchUrl);
}).then(() => true)
.catch(() => false);
@@ -236,19 +366,34 @@
} else if (textUrl) {
return textUrl;
} else {
- book.state = "娌℃湁pdf鎴杢ext鏂囦欢";
+ book.state = "娌℃湁text鏂囦欢";
return ''
}
})
.catch(() => {
- book.state = "娌℃湁pdf鎴杢ext鏂囦欢";
+ book.state = "娌℃湁text鏂囦欢";
return '';
});
}
+/**
+ * 浠嶩TML鎻愬彇鏂囨湰
+ * @param {string} text html鏂囨湰
+ * @returns 鏂囨湰
+ */
+function getTextFromHtml(text) {
+ if (text.includes("<!DOCTYPE html>")) {
+ const s = text.indexOf('<pre>') + 6;
+ const e = text.indexOf('</pre>');
+ text = text.substring(s, e);
+ // text = /(.|\n)*<pre>((.|\n)*)<\/pre>(.|\n)*/g.exec(text)[2];
+ }
+ return text;
+}
+
async function downloadFile(book, url) {
console.log(`涓嬭浇鏂囦欢: ${url}`);
- const ext = url.split(".").pop();
+ const ext = url.split(".").pop().toLowerCase();
const filepath = `./downloads/${book.id} ${book.isbn}.${ext}`;
if (fs.existsSync(filepath)) {
book.state = `涓嬭浇瀹屾垚`;
@@ -259,8 +404,13 @@
return;
}
await retry(() => {
+ const timeoutTime = 10 * 60 * 1000;
+ const source = axios.CancelToken.source();
+ const timeout = setTimeout(() => {
+ source.cancel("timeout");
+ }, timeoutTime);
return new Promise((resolve, reject) => myAxios
- .get(url, { responseType: "stream" })
+ .get(url, { responseType: "stream", timeout: timeoutTime, cancelToken: source.token })
.then((response) => {
const len = response.headers['content-length'];
if (ext !== "pdf" && ext !== "txt" && len > 200 * 1024 * 1024) {
@@ -272,17 +422,19 @@
return;
}
const stream = response.data;
- const out = fs.createWriteStream(filepath);
+ const _filepath = `./downloads/${book.id} ${book.isbn}.${ext}`;
+ const out = fs.createWriteStream(_filepath);
stream.pipe(out);
stream.on("end", () => {
+ clearTimeout(timeout);
book.state = `涓嬭浇瀹屾垚`;
book.format = ext;
book.file = filepath;
book.url = url;
- console.log(`涓嬭浇瀹屾垚锛�${filepath}`);
resolve(true);
});
stream.on("error", (err) => {
+ clearTimeout(timeout);
console.error(err);
book.state = "涓嬭浇澶辫触";
book.url = url;
@@ -297,6 +449,7 @@
});
})
.catch((e) => {
+ clearTimeout(timeout);
console.error(e);
book.state = "涓嬭浇澶辫触";
book.url = url;
@@ -327,6 +480,15 @@
});
}
+function getBookInfo(book) {
+ return retry(async () => {
+ book.publisher = await driver.executeScript(`return document.querySelector("span[itemprop=publisher]").textContent`).catch(e => 0);
+ book.pubDate = await driver.executeScript(`return document.querySelector("span[itemprop=datePublished]").textContent`).catch(e => 0);
+ let pages = await driver.executeScript(`return document.querySelector("span[data-id=resultsCount]").textContent`).catch(e => 0);
+ if (pages) { book.pages = pages.split(' / ')[1]; }
+ });
+}
+
async function downloadBooks(books) {
driver = await createDriver();
@@ -341,15 +503,15 @@
break;
}
bookCount++;
- if (isAlreadyDownloaded(book)) {
+ /*if (isAlreadyDownloaded(book)) {
skipCount++;
continue;
}
- if (book.state && (book.state === "娌℃湁鎼滅储缁撴灉" || book.state === "娌℃湁pdf鎴杢ext鏂囦欢" || book.state === "涓嬭浇瀹屾垚")) {
+ if (book.state && (book.state === "娌℃湁鎼滅储缁撴灉" || book.state === "娌℃湁pdf鎴杢ext鏂囦欢" || book.state === "涓嬭浇瀹屾垚")) {
// 璺宠繃娌℃湁鎼滅储缁撴灉鎴栨病鏈塸df鎴杢ext鏂囦欢鐨勪功绫�
skipCount++;
continue;
- }
+ } */
console.log(`寮�濮嬩笅杞�: ${book.id} ${book.title}`);
// 鎵撳紑鎼滅储椤甸潰骞舵悳绱�
if (!await openSearchPage(book, true)) {
@@ -373,22 +535,24 @@
continue;
}
// 绛変竴娈垫椂闂村啀鎵撳紑璇︽儏椤�
- sleep(getRandomNumber(1000, 30000));
+ sleep(getRandomNumber(500, 10000));
// 鎵撳紑璇︽儏椤�
await openBookDetailPage(book, detailPageUrl);
+ await getBookInfo(book);
// 鑾峰彇涓嬭浇閾炬帴
const url = await getDownloadUrl(book);
if (!url) { continue; }
// 绛夊緟涓�娈垫椂闂村啀涓嬭浇
- await sleep(getRandomNumber(1000, 30000));
+ await sleep(getRandomNumber(500, 10000));
// 涓嬭浇鏂囦欢
try {
await downloadFile(book, url);
console.log(`涓嬭浇瀹屾垚: ${book.id} ${book.title}`);
+ console.log('finish: ' + JSON.stringify(book));
} catch (e) { }
successCount++;
// 绛変竴娈垫椂闂村啀涓嬩竴涓�
- sleep(getRandomNumber(1000, 30000));
+ sleep(getRandomNumber(500, 10000));
}
}
@@ -400,6 +564,9 @@
for (const book of books) {
const index = data.findIndex((row) => row[0] === book.id);
if (index > -1) {
+ data[index][5] = book.publisher;
+ data[index][6] = book.pubDate;
+ data[index][11] = book.pages;
data[index][12] = book.state;
data[index][13] = book.format;
data[index][14] = book.file;
@@ -461,7 +628,7 @@
const books = [];
downloadBooks(books)
.then(() => {
- console.log(`鍏ㄩ儴瀹屾垚锛屽叡涓嬭浇${bookCount}鏈紝鎴愬姛涓嬭浇${successCount}鏈紝璺宠繃${skipCount}鏈紝澶辫触${bookCount - skipCount - successCount}鏈紝鑰楁椂锛� ${msFormat(Date.now() - startTime)}銆俙);
+ console.log(`绾跨▼锛�${threadId}鍏ㄩ儴瀹屾垚锛屽叡涓嬭浇${bookCount}鏈紝鎴愬姛涓嬭浇${successCount}鏈紝璺宠繃${skipCount}鏈紝澶辫触${bookCount - skipCount - successCount}鏈紝鑰楁椂锛� ${msFormat(Date.now() - startTime)}銆俙);
})
.catch(e => {
console.error(e);
@@ -477,6 +644,13 @@
});
}
+if (!fs.existsSync('tmpdir')) {
+ fs.mkdirSync('tmpdir', { recursive: true });
+}
+if (!fs.existsSync('downloads')) {
+ fs.mkdirSync('downloads', { recursive: true });
+}
+
// 澶氳繘绋嬫墽琛�
if (isMainThread) {
initLogger();
@@ -485,15 +659,10 @@
console.log(`绾跨▼鏁帮細${threadSize}, 寮�濮嬭锛�${startRow}, 缁撴潫琛岋細${endRow}`);
let finishCnt = 0;
const finishBooks = [];
- const thBookSize = (endRow - startRow) / threadSize;
const books = getBooksFromExcel(startRow, endRow);
- for (let sr = startRow; sr < endRow; sr += thBookSize) {
- let er = sr + thBookSize;
- if (er > endRow) {
- er = endRow;
- }
- const worker = new Worker("./src/main.mjs", { workerData: { startRow: sr, endRow: er, alreadyDownloadedBooks } });
+ for (let i = 0; i < threadSize; i++) {
+ const worker = new Worker("./src/main.mjs", { workerData: { alreadyDownloadedBooks } });
worker.on("message", (message) => {
if (message.type === 'books') {
finishBooks.push(...message.data);
@@ -511,3 +680,6 @@
main();
}
+// const filepath = "D:\\projects\\book-crawler\\downloads\\10482686 978-1-333-27648-5.txt";
+// let text = fs.readFileSync(filepath, 'utf8');
+// fs.writeFileSync(filepath + '.result.txt', cleanText(text), 'utf-8');
\ No newline at end of file
--
Gitblit v1.9.1