NewsParser/news_parser.js

// ==UserScript==
// @name         News parser
// @namespace    http://zakonvremeni.ru
// @version      0.2
// @description  Parse news
// @author       AlexeiBv+mirocod@narod.ru
// @match        https://tass.ru/*
// @match        https://ria.ru/*
// @match        https://rg.ru/*
// @match        https://russian.rt.com/*
// @match        https://zakonvremeni.ru/*
// @icon         https://icons.duckduckgo.com/ip2/zakonvremeni.ru.ico
// @grant        none
// ==/UserScript==

// Общественное достояние, 2023, Алексей Безбородов (Alexei Bezborodov) <AlexeiBv+mirocod_news_parser@narod.ru>

(function() {
    'use strict';

    // Поиск элементов по регулярному выражению

    function GetElementClassName(a_Element) {
        return a_Element.className;
    }

    function GetNodeName(a_Element) {
        return a_Element.nodeName;
    }

    function CheckRegExp(a_GetElementNameFunc, a_RegExpPattern, a_Element) {
        let re = new RegExp("(?:^|\\s)" + a_RegExpPattern + "(?!\\S)");
        return re.test(a_GetElementNameFunc(a_Element));
    }

    function FindElementsByRegExp(a_GetElementNameFunc, a_RegExpPattern, a_ElementParent) {
        a_ElementParent || (a_ElementParent=document);
        let descendants = a_ElementParent.getElementsByTagName('*'), i=-1, e, result=[];
        while (e = descendants[++i]) {
            if (CheckRegExp(a_GetElementNameFunc, a_RegExpPattern, e)){
                result.push(e);
            }
        }
        return result;
    }

    // Работа со строками

    function TrimString(s) {
        return ( s || '' ).replace( /^\s+|\s+$/g, '' );
    }

    function RemoveBeforeSplitter(a_String, a_Splitter) {
        let index = a_String.indexOf(a_Splitter)
        if (index != -1) {
            return a_String.substring(index + a_Splitter.length);
        }
        return a_String;
    }

    function RemoveAfterSplitter(a_String, a_Splitter, a_SaveSplitter) {
        let index = a_String.indexOf(a_Splitter)
        if (index != -1) {
            let spl_len = a_Splitter.length
            if (!a_SaveSplitter) {
                spl_len = 0;
            }
            return a_String.substring(0, index + spl_len);
        }
        return a_String;
    }

    function ClearUrl(a_Url) {
        const separator = '?';
        return RemoveAfterSplitter(a_Url, separator, false);
    }

    function ClearTextFuncTemplate(a_RemoveBeforeList, a_OnlyFirstIndex) {
        function ClearTextFunc(a_Content, a_ElementIndex) {
            if (a_OnlyFirstIndex && a_ElementIndex && a_ElementIndex != 1) {
                return a_Content;
            }
            let content = a_Content;
            for (let i = 0; i < a_RemoveBeforeList.length; i++) {
                let r = a_RemoveBeforeList[i];
                content = RemoveBeforeSplitter(content, r);
            }
            return content;
        }
        return ClearTextFunc
    }

    // Работа с контейнерами

    function GetImageInContainers(a_Elements, a_TextAlign) {
        let i;
        let img_src = '';
        let re = new RegExp("(https?:\/\/.*\.(?:png|jpg))");
        for (i in a_Elements) {
            let e = a_Elements[i];
            let 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) {
        let result = '';
        let element_index = 1;
        for (var i in a_Elements) {
            let e = a_Elements[i];

            result += a_FinishWorkFunc(e, element_index);
            element_index += 1;
        }
        return result;
    }

    // Фильтрация элементов

    function FIlterElements(a_Elements, a_ElementChecker) {
        let result = [];
        for (let i = 0; i < a_Elements.length; i++) {
            let e = a_Elements[i];
            if (a_ElementChecker(e)) {
                result.push(e);
            }
        }

        return result;
    }

    function ElementCheckerTrue(a_Element) {
        return true;
    }

    function ElementCheckerRia(a_Element) {
        if (a_Element.dataset.type == 'article' || a_Element.dataset.type == 'banner' || a_Element.dataset.type == 'media') {
            return false;
        }

        return true;
    }

    function ElementCheckerZV(a_Element) {
        if (a_Element.itemprop == 'articleBody') {
            return true;
        }
        return false;
    }

    // Обработка элементов

    function GrubTextFuncTemplate() {
        function GrubTextFunc(a_Element) {
            var content = '';
            if (a_Element.innerText) {
                content = TrimString(a_Element.textContent);
            }
            return content;
        }
        return GrubTextFunc
    }

    function RemoveAllAttributes(a_Element) {
        let new_el = document.createElement(a_Element.nodeName);
        new_el.innerHTML = a_Element.innerHTML;
        a_Element.outherHTML = new_el.outherHTML;
    }

    function RemoveCurrentElementSaveChild(a_Element, a_Parent) {
        var parent = a_Element.parentNode || a_Parent;
        while(a_Element.firstChild) parent.insertBefore(a_Element.firstChild, a_Element);
        parent.removeChild(a_Element);
    }

    function GetClearHtml(a_Element, a_OutTag, a_TextAlign) {
        let clear_element = a_Element.cloneNode(true);

        let elements = clear_element.querySelectorAll('*');
        elements.forEach(function (element) {
            let for_rt_com = 'read-more|article__cover';
            let for_rg_com = 'portal|rg-incut|article-img|Section';
            let delete_in_rg = element.nodeName == 'RG-VIDEO' || element.nodeName == 'RG-INCUT';
            if (element && (element.parentNode && CheckRegExp(GetElementClassName, '.*(' + for_rt_com + '|' + for_rg_com + ').*', element) || delete_in_rg)) {
                element.parentNode.removeChild(element);
            }
            else {
                RemoveAllAttributes(element);
                element.removeAttribute('class');
                element.removeAttribute('id');
            }
        });

        let tags_to_delete = ['div', 'span', 'em', 'svg', 'path'];

        tags_to_delete.forEach(function (del_tag_name) {
            let elements = clear_element.querySelectorAll(del_tag_name);
            elements.forEach(function (element) {
                RemoveCurrentElementSaveChild(element, clear_element);
            });
        });

        elements = clear_element.querySelectorAll('a');
        elements.forEach(function (element) {
            if (element.host == location.hostname) {
                RemoveCurrentElementSaveChild(element, clear_element);
            } 
        });

        let tags_to_align = ['p', 'h2'];

        tags_to_align.forEach(function (align_tag_name) {
            let elements = clear_element.querySelectorAll(align_tag_name);
            elements.forEach(function (element) {
                element.style.textAlign = a_TextAlign;
                if (element.innerHTML == '&nbsp;' || element.innerHTML == '') {
                    element.parentNode.removeChild(element);
                }
            });
        });

        let result = clear_element.innerHTML;
        if (a_OutTag && a_TextAlign) {
            result = '<' + a_OutTag + ' style = "text-align:' + a_TextAlign + ';">' + result + '</' + a_OutTag + '>';
        }
        return result;
    }


    function FinishWorkFuncTemplate(a_OutTag, a_TextAlign, a_ClearTextFunc) {
        function FinishWorkFunc(a_Element, a_ElementIndex) {
            let out_tag = a_OutTag;
            if (a_Element && CheckRegExp(GetElementClassName, 'Title_title.*', a_Element)) {
                out_tag = 'h2';
            }
            let a_Content = GetClearHtml(a_Element, out_tag, a_TextAlign)
            if (a_ClearTextFunc) {
                a_Content = a_ClearTextFunc(a_Content, a_ElementIndex);
            }

            if (a_Element && a_Element.dataset && a_Element.dataset.type == 'quote') {
                a_Content = '<blockquote>' + a_Content + '</blockquote>';
            }

            return a_Content;
        }
        return FinishWorkFunc
    }

    // Создание контента для стандартных новостей

    function MakeContentByNews(a_BaseElementTitle, a_BaseElementImage, a_BaseElementText, a_TitleRegExpElementPattern, a_ImageRegExpElementPattern, a_TextRegExpElementPattern, a_ElementChecker, a_ClearTextPatterns) {
        const title_tag = 'h2';
        const p_tag = 'p';
        const title_finish_text_func = FinishWorkFuncTemplate(title_tag, 'center')
        const grub_func = GrubTextFuncTemplate();

        let content = '';
        const paragraph_finish_text_func = FinishWorkFuncTemplate(p_tag, 'justify', ClearTextFuncTemplate(a_ClearTextPatterns, true));
        content += GetContentInContainers(FindElementsByRegExp(GetElementClassName, a_TitleRegExpElementPattern, a_BaseElementTitle), grub_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_func, paragraph_finish_text_func);
        return content;
    }

    // Создание контента для сайта

    function MakeContent() {
        let content = '';
        let source_add = true;
        const zero_tag_func = FinishWorkFuncTemplate()
        const grub_text_func = GrubTextFuncTemplate()

        if (location.hostname == 'tass.ru') {
            // test: https://tass.ru/proisshestviya/19117971
            const base_element = document.getElementById('content_box');
            content = MakeContentByNews(
                base_element,
                base_element,
                base_element,
                '(ArticleHeader_titles|tass_pkg_title--variant_h1_default).*',
                'Image_wrapper_.*',
                '(Paragraph_paragraph|Title_title).*',
                ElementCheckerTrue,
                ['. ']
                );
        }
        else if (location.hostname == 'ria.ru') {
            // test: https://ria.ru/20231020/ssha-1904210900.html
            const base_element = document.getElementsByClassName('article__header')[0];
            const base_element_text = document.getElementsByClassName('article__body')[0];
            const tire = ['-', '–', '—', '‒', '―', '⸺', '⸻'];
            let clear_text = [];
            for (let i in tire) {
                let t = tire[i];
                clear_text.push(t + ' РИА Новости. ');
            }

            content = MakeContentByNews(
                base_element,
                base_element,
                base_element_text,
                'article__title',
                'photoview__open',
                'article__block',
                ElementCheckerRia,
                ['</strong>']
                );
        }
        else if (location.hostname == 'rg.ru') {
            // test: https://rg.ru/2023/10/28/volontery-iz-evrosoiuza-privezli-dlia-zhitelej-donbassa-20-tonn-gumanitarnogo-gruza.html
            const base_element = document.getElementsByClassName('article__header')[0];
            const base_element_text = document.getElementsByClassName('article__body')[0];

            content = MakeContentByNews(
                document,
                document,
                base_element_text,
                'PageArticleContent_title.*',
                'PageArticleContent_image.*',
                '(PageContentCommonStyling_text|PageArticleContent_lead).*',
                ElementCheckerRia,
                []
                );
        }
        else if (location.hostname == 'russian.rt.com') {
            // test: https://russian.rt.com/business/article/1222163-centrobank-stavka-oktyabr-2023
            const base_element = document.getElementsByClassName('article article_article-page')[0];
            const base_element_text = document.getElementsByClassName('article__body')[0];

            content = MakeContentByNews(
                base_element,
                base_element,
                base_element_text,
                'article__heading',
                'article__cover article__cover_article-page',
                'article__text',
                ElementCheckerRia,
                []
                );
        }
        else if (location.hostname == 'zakonvremeni.ru') {
            const base_element = document.getElementsByClassName('item-page')[0];
            const title = GetContentInContainers(FindElementsByRegExp(GetElementClassName, 'page-header', base_element), grub_text_func);
            const parent_category = GetContentInContainers(FindElementsByRegExp(GetElementClassName, 'parent-category-name', base_element), grub_text_func);
            const category = GetContentInContainers(FindElementsByRegExp(GetElementClassName, 'category-name', base_element), grub_text_func);
            const 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;
        }

        let result = '';
        if (content.length > 0) {
            result = '<textarea id = "news_content" rows="10" cols="100">' + content;
            if (source_add) {
                result += '<p style="text-align: justify;">Источник: <a href = "' + ClearUrl(document.URL) + '">' + location.hostname + '</a></p>';
            }
            result += '</textarea>';
        }
        return result;
    }

    let content = MakeContent();
    let logo = document.createElement("div");
    logo.innerHTML = '<div style="margin: 0pt auto; width: 800px; text-align: center;">' + content + '</div>';

    document.body.insertBefore(logo, document.body.firstChild);

})();