4
0
Fork 0
Wiki Issues Projects Code Pull Requests Releases Activity
Browse Source

Версия 0.2 полностью обновлена. Функционал тот же.

master
Alexei 1 year ago
parent
b3bb8fdc87
commit
b8dc6c2792
1 changed files with 126 additions and 81 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 207
      news_parser.js

207
news_parser.js
Unescape View File

@ -1,7 +1,7 @@
// ==UserScript==
// @name News parser
// @namespace http://zakonvremeni.ru
// @version 0.1
// @version 0.2
// @description Parse news
// @author AlexeiBv+mirocod@narod.ru
// @match https://tass.ru/*
@ -11,49 +11,32 @@
// @grant none
// ==/UserScript==
// Общественное достояние, 2023, Алексей Безбородов (Alexei Bezborodov) <AlexeiBv+mirocod_platform_bot@narod.ru>
// Общественное достояние, 2023, Алексей Безбородов (Alexei Bezborodov) <AlexeiBv+mirocod_news_parser@narod.ru>
(function() {
'use strict';
function getByClass (className, parent) {
parent || (parent=document);
var descendants=parent.getElementsByTagName('*'), i=-1, e, result=[];
var re = new RegExp("(?:^|\\s)" + className + "(?!\\S)");
// Поиск элементов по регулярному выражению
function GetElementClassName(a_Element) {
return a_Element.className;
}
function FindElementsByRegExp(a_GenElementNameFunc, a_RegExpPattern, a_ElementParent) {
a_ElementParent || (a_ElementParent=document);
var descendants = a_ElementParent.getElementsByTagName('*'), i=-1, e, result=[];
var re = new RegExp("(?:^|\\s)" + a_RegExpPattern + "(?!\\S)");
while (e=descendants[++i]) {
if (re.test(e.className)){
if (re.test(a_GenElementNameFunc(e))){
result.push(e);
}
}
return result;
}
function GetImageInContainers(baseClass, parent, textAlign) {
var elems = getByClass(baseClass, parent);
if (!elems) {
return '';
}
// Работа со строками
var i;
var img_src = '';
var re = new RegExp("(https?:\/\/.*\.(?:png|jpg))");
for (i in elems) {
var e = elems[i];
var children = e.querySelectorAll("*");
for(let i = 0; i < children.length; i++){
var c = children[i];
if (c.nodeName == 'IMG' && re.test(c.src)) {
img_src = c.src;
}
}
}
if (img_src.length > 0) {
return '<p style = "text-align:' + textAlign + ';"><img src = "'+ img_src + '" width = "600px"/></p>';
}
return '';
}
function Trim(s) {
function TrimString(s) {
return ( s || '' ).replace( /^\s+|\s+$/g, '' );
}
@ -77,8 +60,11 @@
return a_String;
}
function FinishWorkFuncTemplate(a_OutTag, a_TextAlign) {
function FinishWorkFuncTemplate(a_OutTag, a_TextAlign, a_ClearTextFunc) {
function FinishWorkFunc(a_Content, a_Element) {
if (a_ClearTextFunc) {
a_Content = a_ClearTextFunc(a_Content);
}
if (a_OutTag && a_TextAlign) {
return '<' + a_OutTag + ' style = "text-align:' + a_TextAlign + ';">' + a_Content + '</' + a_OutTag + '>';
}
@ -89,18 +75,49 @@
return FinishWorkFunc
}
function GetContentInContainers(a_FinishWorkFunc, a_GrubTextFunc, baseClass, parent, a_ElementFilterFunc, a_ClearTextFunc) {
var elems = getByClass(baseClass, parent);
if (!elems) {
return 'Не удалось найти ' + baseClass;
function ClearUrl(a_Url) {
var separator = '?';
return RemoveAfterSplitter(a_Url, separator, false);
}
var result = '';
for (var i in elems) {
var e = elems[i];
if (a_ElementFilterFunc && !a_ElementFilterFunc(e)) {
continue;
function ClearTextFuncTemplate(a_RemoveBeforeList) {
function ClearTextFunc(a_Content) {
var content = a_Content;
for (let i = 0; i < a_RemoveBeforeList.length; i++) {
var r = a_RemoveBeforeList[i];
content = RemoveBeforeSplitter(content, r);
}
return content;
}
return ClearTextFunc
}
// Работа с контейнерами
function GetImageInContainers(a_Elements, a_TextAlign) {
var i;
var img_src = '';
var re = new RegExp("(https?:\/\/.*\.(?:png|jpg))");
for (i in a_Elements) {
var e = a_Elements[i];
var children = e.querySelectorAll("*");
for(let i = 0; i < children.length; i++){
var c = children[i];
if (c.nodeName == 'IMG' && re.test(c.src)) {
img_src = c.src;
}
}
}
if (img_src.length > 0) {
return '<p style = "text-align:' + a_TextAlign + ';"><img src = "'+ img_src + '" width = "600px"/></p>';
}
return '';
}
function GetContentInContainers(a_Elements, a_GrubTextFunc, a_FinishWorkFunc) {
var result = '';
for (var i in a_Elements) {
var e = a_Elements[i];
var content = '';
if (e.querySelectorAll) {
@ -115,9 +132,6 @@
}
}
}
if (a_ClearTextFunc) {
content = a_ClearTextFunc(content);
}
if (a_FinishWorkFunc) {
result += a_FinishWorkFunc(content, e);
@ -129,76 +143,107 @@
return result;
}
function ClearUrl(a_Url) {
var separator = '?';
return RemoveAfterSplitter(a_Url, separator, false);
// Фильтрация элементов
function FIlterElements(a_Elements, a_ElementChecker) {
var result = [];
for (let i = 0; i < a_Elements.length; i++) {
var e = a_Elements[i];
if (a_ElementChecker(e)) {
result.push(e);
}
}
function FIlterTrue(element) {
return true;
return result;
}
function FIlterRia(element) {
if (element.dataset.type == 'text' || element.dataset.type == 'quote' || element.dataset.type == 'list') {
function ElementCheckerTrue(a_Element) {
return true;
}
return false;
}
function FIlterZV(element) {
if (element.itemprop == 'articleBody') {
function ElementCheckerRia(a_Element) {
if (a_Element.dataset.type == 'text' || a_Element.dataset.type == 'quote' || a_Element.dataset.type == 'list') {
return true;
}
return false;
}
function ClearTextFuncTemplate(a_RemoveBeforeList) {
function ClearTextFunc(a_Content) {
var content = a_Content;
for (let i = 0; i < a_RemoveBeforeList.length; i++) {
var r = a_RemoveBeforeList[i];
content = RemoveBeforeSplitter(content, r);
}
return content;
function ElementCheckerZV(a_Element) {
if (a_Element.itemprop == 'articleBody') {
return true;
}
return ClearTextFunc
return false;
}
function GrubTextFuncTemplate() {
function GrubTextFunc(a_Element) {
var content = '';
if (a_Element.innerText) {
content = Trim(a_Element.textContent);
content = TrimString(a_Element.textContent);
}
return content;
}
return GrubTextFunc
}
function MakeContent() {
var content = '';
// Создание контента для стандартных новостей
function MakeContentByNews(a_BaseElementTitle, a_BaseElementImage, a_BaseElementText, a_TitleRegExpElementPattern, a_ImageRegExpElementPattern, a_TextRegExpElementPattern, a_ElementChecker, a_ClearTextPatterns) {
var title_tag = 'h2';
var p_tag = 'p';
var zero_tag = '';
var title_finish_text_func = FinishWorkFuncTemplate(title_tag, 'center')
var grub_text_func = GrubTextFuncTemplate()
var content = '';
var paragraph_finish_text_func = FinishWorkFuncTemplate(p_tag, 'justify', ClearTextFuncTemplate(a_ClearTextPatterns));
content += GetContentInContainers(FindElementsByRegExp(GetElementClassName, a_TitleRegExpElementPattern, a_BaseElementTitle), grub_text_func, title_finish_text_func);
content += GetImageInContainers(FindElementsByRegExp(GetElementClassName, a_ImageRegExpElementPattern, a_BaseElementImage), 'center');
content += GetContentInContainers(FIlterElements(FindElementsByRegExp(GetElementClassName, a_TextRegExpElementPattern, a_BaseElementText), a_ElementChecker), grub_text_func, paragraph_finish_text_func);
return content;
}
// Создание контента для сайта
function MakeContent() {
var content = '';
var source_add = true;
var title_func = FinishWorkFuncTemplate(title_tag, 'center')
var paragraph_func = FinishWorkFuncTemplate(p_tag, 'justify')
var zero_tag_func = FinishWorkFuncTemplate()
var grub_text_func = GrubTextFuncTemplate()
if (location.hostname == 'tass.ru') {
content += GetContentInContainers(title_func, GrubTextFuncTemplate(), 'tass_pkg_title--variant_h1_default.*', document.getElementById('content_box'));
content += GetImageInContainers('Image_wrapper_.*', document.getElementById('content_box'), 'center');
content += GetContentInContainers(paragraph_func, GrubTextFuncTemplate(), 'Paragraph_paragraph.*', document.getElementById('content_box'), FIlterTrue, ClearTextFuncTemplate(['/ТАСС/. ']));
let base_element = document.getElementById('content_box');
content = MakeContentByNews(
base_element,
base_element,
base_element,
'tass_pkg_title--variant_h1_default.*',
'Image_wrapper_.*',
'Paragraph_paragraph.*',
ElementCheckerTrue,
['/ТАСС/. ']
);
}
else if (location.hostname == 'ria.ru') {
content += GetContentInContainers(title_func, GrubTextFuncTemplate(), title_tag, document.getElementsByClassName('article__header')[0]);
content += GetImageInContainers('photoview__open', document.getElementsByClassName('article__header')[0], 'center');
content += GetContentInContainers(paragraph_func, GrubTextFuncTemplate(), p_tag, 'article__block', document.getElementsByClassName('article__body')[0], 'justify', FIlterRia, ClearTextFuncTemplate(['– РИА Новости. ', '— РИА Новости. ']));
let base_element = document.getElementsByClassName('article__header')[0];
var base_element_text = document.getElementsByClassName('article__body')[0];
content = MakeContentByNews(
base_element,
base_element,
base_element_text,
'article__title',
'photoview__open',
'article__block',
ElementCheckerRia,
['– РИА Новости. ', '— РИА Новости. ']
);
}
else if (location.hostname == 'zakonvremeni.ru') {
var title = GetContentInContainers(zero_tag_func, GrubTextFuncTemplate(), 'page-header', document.getElementsByClassName('item-page')[0]);
var parent_category = GetContentInContainers(zero_tag_func, GrubTextFuncTemplate(), 'parent-category-name', document.getElementsByClassName('item-page')[0]);
var category = GetContentInContainers(zero_tag_func, GrubTextFuncTemplate(), 'category-name', document.getElementsByClassName('item-page')[0]);
var page = RemoveAfterSplitter(Trim(document.getElementsByClassName('item-page')[0].querySelector('[itemprop=articleBody]').textContent), '.', true);
let base_element = document.getElementsByClassName('item-page')[0];
var title = GetContentInContainers(FindElementsByRegExp(GetElementClassName, 'page-header', base_element), grub_text_func);
var parent_category = GetContentInContainers(FindElementsByRegExp(GetElementClassName, 'parent-category-name', base_element), grub_text_func);
var category = GetContentInContainers(FindElementsByRegExp(GetElementClassName, 'category-name', base_element), grub_text_func);
var page = RemoveAfterSplitter(TrimString(document.getElementsByClassName('item-page')[0].querySelector('[itemprop=articleBody]').textContent), '.', true);
content = title + '\n' + parent_category + ' ' + category + '\n\n' + page + '\n' + document.URL;
source_add = false;
}

Write Preview
Loading…
Cancel
Save