Refactored Mike Adriano scraper. Changed logo and favicon. Added style methods to qu.
This commit is contained in:
DebaucheryLibrarian
@@ -68,7 +68,7 @@ async function fetchActorReleases({ query }, url, remainingPages, actorName, acc
|
||||
return accReleases.concat(releases);
|
||||
}
|
||||
|
||||
async function scrapeProfile({ query, window }, actorName, url, include) {
|
||||
async function scrapeProfile({ query }, actorName, url, include) {
|
||||
const profile = {};
|
||||
|
||||
profile.avatar = {
|
||||
@@ -85,7 +85,7 @@ async function scrapeProfile({ query, window }, actorName, url, include) {
|
||||
|
||||
if (include.releases) {
|
||||
const availablePages = query.all('.pageboxdropdown option', 'value');
|
||||
profile.releases = await fetchActorReleases(qu.init(query.q('#episodes > table'), window), url, availablePages.slice(1), actorName);
|
||||
profile.releases = await fetchActorReleases(qu.init(query.q('#episodes > table')), url, availablePages.slice(1), actorName);
|
||||
}
|
||||
|
||||
return profile;
|
||||
@@ -105,7 +105,9 @@ async function fetchScene(url, channel) {
|
||||
|
||||
async function fetchProfile(actorName, entity, include) {
|
||||
const url = `http://killergram.com/episodes.asp?page=episodes&model=${encodeURI(actorName)}&ct=model`;
|
||||
const res = await qu.get(url, '#content');
|
||||
const res = await qu.get(url, '#content', null, {
|
||||
followRedirects: false,
|
||||
});
|
||||
|
||||
return res.ok ? scrapeProfile(res.item, actorName, url, include) : res.status;
|
||||
}
|
||||
|
||||
@@ -1,237 +1,88 @@
|
||||
'use strict';
|
||||
|
||||
/* eslint-disable newline-per-chained-call */
|
||||
const { JSDOM } = require('jsdom');
|
||||
const cheerio = require('cheerio');
|
||||
const moment = require('moment');
|
||||
// const bhttp = require('bhttp');
|
||||
const qu = require('../utils/qu');
|
||||
|
||||
const { get } = require('../utils/http');
|
||||
|
||||
const descriptionTags = {
|
||||
'anal cream pie': 'anal creampie',
|
||||
'ass to mouth': 'ass to mouth',
|
||||
'cream pie in her ass': 'anal creampie',
|
||||
'eats ass': 'ass eating',
|
||||
facial: 'facial',
|
||||
gaped: 'gaping',
|
||||
gapes: 'gaping',
|
||||
gape: 'gaping',
|
||||
'rectal cream pie': 'anal creampie',
|
||||
rimming: 'ass eating',
|
||||
};
|
||||
|
||||
function deriveTagsFromDescription(description) {
|
||||
const matches = (description || '').toLowerCase().match(new RegExp(Object.keys(descriptionTags).join('|'), 'g'));
|
||||
|
||||
return matches
|
||||
? matches.map(match => descriptionTags[match])
|
||||
: [];
|
||||
}
|
||||
|
||||
async function scrapeLatestA(html, site) {
|
||||
const { document } = new JSDOM(html).window;
|
||||
const sceneElements = document.querySelectorAll('.content-item-large, .content-item');
|
||||
|
||||
return Promise.all(Array.from(sceneElements, async (element) => {
|
||||
const $ = cheerio.load(element.innerHTML, { normalizeWhitespace: true });
|
||||
|
||||
const titleElement = element.querySelector('h3.title a');
|
||||
const title = titleElement.textContent;
|
||||
const url = titleElement.href;
|
||||
const entryId = url.split('/').slice(-2)[0];
|
||||
|
||||
const descriptionElement = element.querySelector('.desc');
|
||||
const description = descriptionElement && descriptionElement.textContent.trim();
|
||||
const date = moment(element.querySelector('.date, time').textContent, 'Do MMM YYYY').toDate();
|
||||
|
||||
const actors = Array.from(element.querySelectorAll('h4.models a'), actorElement => actorElement.textContent);
|
||||
|
||||
const durationString = element.querySelector('.total-time').textContent.trim();
|
||||
// timestamp is sometimes 00:00, sometimes 0:00:00
|
||||
const duration = durationString.split(':').length === 3
|
||||
? moment.duration(durationString).asSeconds()
|
||||
: moment.duration(`00:${durationString}`).asSeconds();
|
||||
|
||||
const ratingElement = element.querySelector('.rating');
|
||||
const stars = ratingElement && ratingElement.dataset.rating;
|
||||
|
||||
const [poster, ...primaryPhotos] = Array.from(element.querySelectorAll('img'), imageElement => imageElement.src);
|
||||
const secondaryPhotos = $('.thumb-top, .thumb-bottom')
|
||||
.map((photoIndex, photoElement) => $(photoElement).css()['background-image'])
|
||||
.toArray()
|
||||
.map(photoUrl => photoUrl.slice(photoUrl.indexOf('http'), photoUrl.indexOf('.jpg') + 4));
|
||||
|
||||
const photos = [...primaryPhotos, ...secondaryPhotos];
|
||||
const tags = deriveTagsFromDescription(description);
|
||||
|
||||
const scene = {
|
||||
url,
|
||||
entryId,
|
||||
title,
|
||||
description,
|
||||
actors,
|
||||
director: 'Mike Adriano',
|
||||
date,
|
||||
duration,
|
||||
tags,
|
||||
poster,
|
||||
photos,
|
||||
rating: {
|
||||
stars,
|
||||
},
|
||||
site,
|
||||
};
|
||||
|
||||
return scene;
|
||||
}));
|
||||
}
|
||||
|
||||
async function scrapeLatestB(html) {
|
||||
const { document } = new JSDOM(html).window;
|
||||
const sceneElements = document.querySelectorAll('.content-border');
|
||||
|
||||
return Promise.all(Array.from(sceneElements, async (element) => {
|
||||
const $ = cheerio.load(element.innerHTML, { normalizeWhitespace: true });
|
||||
async function scrapeAll(scenes) {
|
||||
return scenes.map(({ query }) => {
|
||||
const release = {
|
||||
director: 'Mike Adriano',
|
||||
};
|
||||
|
||||
const titleElement = element.querySelector('.content-title-wrap a');
|
||||
release.title = titleElement.title || titleElement.textContent.trim();
|
||||
release.url = titleElement.href;
|
||||
release.entryId = release.url.split('/').slice(-2)[0];
|
||||
release.title = query.cnt('h3.title a, .content-title-wrap a');
|
||||
release.url = query.url('h3.title a, .content-title-wrap a');
|
||||
release.entryId = new URL(release.url).pathname.match(/\/view\/(\d+)/)[1];
|
||||
|
||||
release.description = element.querySelector('.content-description').textContent.trim();
|
||||
release.date = (moment(element.querySelector('.mobile-date').textContent, 'MM/DD/YYYY')
|
||||
|| moment(element.querySelector('.date').textContent, 'Do MMM YYYY')).toDate();
|
||||
release.actors = Array.from(element.querySelectorAll('.content-models a'), actorElement => actorElement.textContent);
|
||||
release.description = query.cnt('.desc, .content-description');
|
||||
release.date = query.date('.date, time, .hide', 'Do MMM YYYY');
|
||||
|
||||
const durationString = element.querySelector('.total-time').textContent.trim();
|
||||
// timestamp is somethines 00:00, sometimes 0:00:00
|
||||
release.duration = durationString.split(':').length === 3
|
||||
? moment.duration(durationString).asSeconds()
|
||||
: moment.duration(`00:${durationString}`).asSeconds();
|
||||
release.actors = query.cnts('h4.models a, .content-models a');
|
||||
release.duration = query.dur('.total-time');
|
||||
|
||||
const [poster, ...primaryPhotos] = Array.from(element.querySelectorAll('a img'), imageElement => imageElement.src);
|
||||
const secondaryPhotos = $('.thumb-mouseover')
|
||||
.map((photoIndex, photoElement) => $(photoElement).css()['background-image'])
|
||||
.toArray()
|
||||
.map(photoUrl => photoUrl.slice(photoUrl.indexOf('http'), photoUrl.indexOf('.jpg') + 4));
|
||||
const [poster, ...primaryPhotos] = query.imgs('a img');
|
||||
const secondaryPhotos = query.styles('.thumb-top, .thumb-bottom, .thumb-mouseover', 'background-image').map(style => style.match(/url\((.*)\)/)[1]);
|
||||
|
||||
release.poster = poster;
|
||||
release.photos = [...primaryPhotos, ...secondaryPhotos];
|
||||
release.photos = primaryPhotos.concat(secondaryPhotos);
|
||||
|
||||
release.tags = deriveTagsFromDescription(release.description);
|
||||
return release;
|
||||
}));
|
||||
});
|
||||
}
|
||||
|
||||
async function scrapeSceneA(html, url) {
|
||||
const { document } = new JSDOM(html).window;
|
||||
const element = document.querySelector('.content-page-info');
|
||||
const release = {
|
||||
url,
|
||||
director: 'Mike Adriano',
|
||||
};
|
||||
function scrapeScene({ query }, url) {
|
||||
const release = { director: 'Mike Adriano' };
|
||||
|
||||
release.entryId = url.split('/').slice(-2)[0];
|
||||
release.title = element.querySelector('.title').textContent.trim();
|
||||
release.description = element.querySelector('.desc').textContent.trim();
|
||||
release.date = moment(element.querySelector('.post-date').textContent.trim(), 'Do MMM YYYY').toDate();
|
||||
if (query.exists('a[href*="stackpath.com"]')) {
|
||||
throw new Error('URL blocked by StackPath');
|
||||
}
|
||||
|
||||
release.actors = Array.from(element.querySelectorAll('.models a'), actorElement => actorElement.textContent);
|
||||
release.entryId = new URL(url).pathname.match(/\/view\/(\d+)/)[1];
|
||||
|
||||
const durationString = element.querySelector('.total-time').textContent.trim();
|
||||
// timestamp is sometimes 00:00, sometimes 0:00:00
|
||||
release.duration = durationString.split(':').length === 3
|
||||
? moment.duration(durationString).asSeconds()
|
||||
: moment.duration(`00:${durationString}`).asSeconds();
|
||||
release.title = query.cnt('.content-page-info .title');
|
||||
release.description = query.cnt('.content-page-info .desc');
|
||||
release.date = query.date('.content-page-info .date, .content-page-info .hide', 'Do MMM YYYY');
|
||||
|
||||
const { poster } = document.querySelector('.content-page-header video');
|
||||
const { src, type } = document.querySelector('.content-page-header source');
|
||||
release.actors = query.cnts('.content-page-info .models a');
|
||||
release.duration = query.dur('.content-page-info .total-time:last-child');
|
||||
|
||||
release.poster = poster;
|
||||
release.trailer = { src, type };
|
||||
release.poster = query.poster('.content-page-header video, .content-page-header-inner video');
|
||||
|
||||
release.tags = deriveTagsFromDescription(release.description);
|
||||
const trailerEl = query.q('.content-page-header source, .content-page-header-inner source');
|
||||
|
||||
if (trailerEl) {
|
||||
release.trailer = {
|
||||
src: trailerEl.src,
|
||||
type: trailerEl.type,
|
||||
};
|
||||
}
|
||||
|
||||
return release;
|
||||
}
|
||||
|
||||
async function scrapeSceneB(html, url, site) {
|
||||
const { document } = new JSDOM(html).window;
|
||||
const element = document.querySelector('.content-page-info');
|
||||
|
||||
const entryId = url.split('/').slice(-2)[0];
|
||||
const title = element.querySelector('.title').textContent.trim();
|
||||
const description = element.querySelector('.desc').textContent.trim();
|
||||
const date = moment(element.querySelector('.date').textContent.trim(), 'Do MMM YYYY').toDate();
|
||||
|
||||
const actors = Array.from(element.querySelectorAll('.models a'), actorElement => actorElement.textContent);
|
||||
|
||||
const durationString = element.querySelector('.total-time').textContent.trim();
|
||||
// timestamp is somethines 00:00, sometimes 0:00:00
|
||||
const duration = durationString.split(':').length === 3
|
||||
? moment.duration(durationString).asSeconds()
|
||||
: moment.duration(`00:${durationString}`).asSeconds();
|
||||
|
||||
const { poster } = document.querySelector('.content-page-header-inner video');
|
||||
const { src, type } = document.querySelector('.content-page-header-inner source');
|
||||
|
||||
const tags = deriveTagsFromDescription(description);
|
||||
|
||||
const scene = {
|
||||
url,
|
||||
entryId,
|
||||
title,
|
||||
description,
|
||||
actors,
|
||||
director: 'Mike Adriano',
|
||||
date,
|
||||
duration,
|
||||
tags,
|
||||
poster,
|
||||
trailer: {
|
||||
src,
|
||||
type,
|
||||
},
|
||||
site,
|
||||
};
|
||||
|
||||
return scene;
|
||||
}
|
||||
|
||||
async function fetchLatest(site, page = 1) {
|
||||
const { host } = new URL(site.url);
|
||||
async function fetchLatest(channel, page = 1) {
|
||||
const { host } = new URL(channel.url);
|
||||
const url = `https://tour.${host}/videos?page=${page}`;
|
||||
|
||||
const res = await get(url);
|
||||
const res = await qu.get(url);
|
||||
|
||||
if (res.code === 200) {
|
||||
if (host === 'trueanal.com' || host === 'swallowed.com') {
|
||||
return scrapeLatestA(res.html, site);
|
||||
if (res.ok) {
|
||||
if (res.item.query.exists('a[href*="stackpath.com"]')) {
|
||||
throw new Error('URL blocked by StackPath');
|
||||
}
|
||||
|
||||
return scrapeLatestB(res.html, site);
|
||||
return scrapeAll(qu.initAll(res.item.el, '.content-item-large, .content-item, .content-border'), channel);
|
||||
}
|
||||
|
||||
return res.code;
|
||||
return res.status;
|
||||
}
|
||||
|
||||
async function fetchScene(url, site) {
|
||||
const { host } = new URL(site.url);
|
||||
const res = await get(url);
|
||||
async function fetchScene(url, channel) {
|
||||
const res = await qu.get(url);
|
||||
|
||||
if (res.code === 200) {
|
||||
if (host === 'trueanal.com' || host === 'swallowed.com') {
|
||||
return scrapeSceneA(res.body.toString(), url, site);
|
||||
}
|
||||
|
||||
return scrapeSceneB(res.body.toString(), url, site);
|
||||
if (res.ok) {
|
||||
return scrapeScene(res.item, url, channel);
|
||||
}
|
||||
|
||||
return res.code;
|
||||
return res.status;
|
||||
}
|
||||
|
||||
/* API protected
|
||||
|
||||
@@ -78,7 +78,7 @@ function scrapeAll(scenes, site, origin) {
|
||||
release.url = `${site?.url || origin}${scene.targetUrl}`;
|
||||
|
||||
release.date = moment.utc(scene.releaseDate).toDate();
|
||||
release.shootDate = moment.utc(scene.shootDate).toDate();
|
||||
release.datePrecision = 'minute';
|
||||
|
||||
release.actors = scene.models;
|
||||
release.stars = Number(scene.textRating) / 2;
|
||||
@@ -104,7 +104,7 @@ function scrapeUpcoming(scene, site) {
|
||||
release.url = `${site.url}${scene.targetUrl}`;
|
||||
|
||||
release.date = moment.utc(scene.releaseDate).toDate();
|
||||
release.shootDate = moment.utc(scene.shootDate).toDate();
|
||||
release.datePrecision = 'minute';
|
||||
|
||||
release.actors = scene.models;
|
||||
|
||||
@@ -133,7 +133,8 @@ async function scrapeScene(data, url, site, baseRelease) {
|
||||
release.entryId = scene.newId;
|
||||
|
||||
release.date = moment.utc(scene.releaseDate).toDate();
|
||||
release.shootDate = moment.utc(scene.shootDate).toDate();
|
||||
release.productionDate = moment.utc(scene.shootDate).toDate();
|
||||
release.datePrecision = 'minute';
|
||||
|
||||
release.actors = baseRelease?.actors || scene.models;
|
||||
|
||||
|
||||
@@ -31,6 +31,7 @@ function curateReleaseEntry(release, batchId, existingRelease) {
|
||||
shoot_id: release.shootId || null,
|
||||
url: release.url,
|
||||
date: Number(release.date) ? release.date : null,
|
||||
production_date: Number(release.productionDate) ? release.productionDate : null,
|
||||
date_precision: release.datePrecision,
|
||||
slug,
|
||||
description: release.description,
|
||||
|
||||
17
src/utils/jsdom-url.js
Normal file
17
src/utils/jsdom-url.js
Normal file
@@ -0,0 +1,17 @@
|
||||
'use strict';
|
||||
|
||||
const { JSDOM } = require('jsdom');
|
||||
|
||||
const el = new JSDOM(`
|
||||
<span id="url" style="background-image: url(https://w.wallhaven.cc/full/dg/wallhaven-dg7y23.jpg);">
|
||||
<span id="urlQuotes" style="background-image: url('https://w.wallhaven.cc/full/dg/wallhaven-dg7y23.jpg');">
|
||||
<span id="color" style="color: rgb(255, 0, 0);">
|
||||
<span id="urlSpaces" style="background-image: url( https://w.wallhaven.cc/full/md/wallhaven-mdvmrm.jpg );">
|
||||
<span id="colorSpaces" style="color: rgb( 255, 0, 0 );">
|
||||
`).window.document;
|
||||
|
||||
console.log(el.querySelector('#url').style.backgroundImage);
|
||||
console.log(el.querySelector('#urlQuotes').style.backgroundImage);
|
||||
console.log(el.querySelector('#color').style.color);
|
||||
console.log(el.querySelector('#urlSpaces').style.backgroundImage);
|
||||
console.log(el.querySelector('#colorSpaces').style.color);
|
||||
@@ -4,9 +4,11 @@ const { JSDOM } = require('jsdom');
|
||||
const moment = require('moment');
|
||||
const http = require('./http');
|
||||
|
||||
const { window: globalWindow } = new JSDOM('');
|
||||
|
||||
function trim(str) {
|
||||
if (typeof str !== 'string') {
|
||||
return null;
|
||||
return str;
|
||||
}
|
||||
|
||||
return str.trim().replace(/\s+/g, ' ');
|
||||
@@ -55,9 +57,9 @@ function q(context, selector, attrArg, applyTrim = true) {
|
||||
if (attr) {
|
||||
const value = selector
|
||||
? context.querySelector(selector)?.[attr] || context.querySelector(selector)?.attributes[attr]?.value
|
||||
: context[attr] || context.attributes[attr]?.value;
|
||||
: context[attr] || context.getAttribute(attr);
|
||||
|
||||
return applyTrim && value ? trim(value) : value;
|
||||
return applyTrim && typeof value === 'string' ? trim(value) : value;
|
||||
}
|
||||
|
||||
return selector ? context.querySelector(selector) : context;
|
||||
@@ -77,6 +79,14 @@ function exists(context, selector) {
|
||||
return !!q(context, selector);
|
||||
}
|
||||
|
||||
function content(context, selector, applyTrim = true) {
|
||||
return q(context, selector, 'textContent', applyTrim);
|
||||
}
|
||||
|
||||
function contents(context, selector, applyTrim) {
|
||||
return all(context, selector, 'textContent', applyTrim);
|
||||
}
|
||||
|
||||
function html(context, selector) {
|
||||
const el = q(context, selector, null, true);
|
||||
|
||||
@@ -103,6 +113,33 @@ function text(context, selector, applyTrim = true) {
|
||||
return applyTrim ? trim(textValue) : textValue;
|
||||
}
|
||||
|
||||
function removeStyleFunctionSpaces(el) {
|
||||
// jsdom appears to have a bug where it ignores inline CSS attributes set to a function() containing spaces, e.g. url( image.png )
|
||||
el.setAttribute('style', el.getAttribute('style').replace(/\(\s+(.*)\s+\)/g, (match, cssArgs) => `(${cssArgs})`));
|
||||
}
|
||||
|
||||
function style(context, selector, styleAttr) {
|
||||
const el = q(context, selector);
|
||||
|
||||
if (el) {
|
||||
removeStyleFunctionSpaces(el);
|
||||
|
||||
return styleAttr ? el.style[styleAttr] : el.style;
|
||||
}
|
||||
|
||||
return null;
|
||||
}
|
||||
|
||||
function styles(context, selector, styleAttr) {
|
||||
const elStyles = Array.from(context.querySelectorAll(selector), (el) => {
|
||||
removeStyleFunctionSpaces(el);
|
||||
|
||||
return styleAttr ? el.style[styleAttr] : el.style;
|
||||
});
|
||||
|
||||
return elStyles;
|
||||
}
|
||||
|
||||
function number(context, selector, attr = true) {
|
||||
const value = q(context, selector, attr);
|
||||
|
||||
@@ -236,6 +273,10 @@ const legacyFuncs = {
|
||||
const quFuncs = {
|
||||
all,
|
||||
html,
|
||||
content,
|
||||
contents,
|
||||
cnt: content,
|
||||
cnts: contents,
|
||||
date,
|
||||
dur: duration,
|
||||
duration,
|
||||
@@ -250,6 +291,8 @@ const quFuncs = {
|
||||
num: number,
|
||||
poster,
|
||||
q,
|
||||
style,
|
||||
styles,
|
||||
text,
|
||||
texts,
|
||||
trailer: video,
|
||||
@@ -265,7 +308,7 @@ function init(element, window) {
|
||||
const legacyContextFuncs = Object.entries(legacyFuncs) // dynamically attach methods with context
|
||||
.reduce((acc, [key, func]) => ({
|
||||
...acc,
|
||||
[key]: (...args) => (window && args[0] instanceof window.HTMLElement // allow for different context
|
||||
[key]: (...args) => (args[0] instanceof globalWindow.HTMLElement // allow for different context
|
||||
? func(...args)
|
||||
: func(element, ...args)),
|
||||
}), {});
|
||||
@@ -273,7 +316,7 @@ function init(element, window) {
|
||||
const quContextFuncs = Object.entries(quFuncs) // dynamically attach methods with context
|
||||
.reduce((acc, [key, func]) => ({
|
||||
...acc,
|
||||
[key]: (...args) => (window && args[0] instanceof window.HTMLElement // allow for different context
|
||||
[key]: (...args) => (args[0] instanceof globalWindow.HTMLElement // allow for different context
|
||||
? func(...args)
|
||||
: func(element, ...args)),
|
||||
}), {});
|
||||
@@ -319,7 +362,7 @@ function extractAll(htmlValue, selector) {
|
||||
}
|
||||
|
||||
async function get(urlValue, selector, headers, options, queryAll = false) {
|
||||
const res = await http.get(urlValue, headers);
|
||||
const res = await http.get(urlValue, headers, options);
|
||||
|
||||
if (res.statusCode === 200) {
|
||||
const item = queryAll
|
||||
|
||||
Reference in New Issue
Block a user