Добавлены загаловки тасс, исправлена очистка ХТМЛ, остаются теги таблиц и списков и много чего ещё. #5 #1

news_parser.js

@@ -113,26 +113,7 @@
         for (var i in a_Elements) {
             let e = a_Elements[i];
 
-            let content = '';
-            if (e.querySelectorAll) {
-                var children = e.querySelectorAll("*");
-                if (children.length == 0 || e.innerText) {
-                    content += a_GrubTextFunc(e);
-                }
-                else {
-                    for (let i = 0; i < children.length; i++) {
-                        let c = children[i];
-                        content += a_GrubTextFunc(c);
-                    }
-                }
-            }
-
-            if (a_FinishWorkFunc) {
-                result += a_FinishWorkFunc(content, e);
-            }
-            else {
-                result += content;
-            }
+            result += a_FinishWorkFunc(e);
         }
         return result;
     }
@@ -183,8 +164,57 @@
         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 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);
+            });
+        });
+
+        let elements = clear_element.querySelectorAll('*');
+        elements.forEach(function (element) {
+            RemoveAllAttributes(element);
+        });
+
+        elements = clear_element.querySelectorAll('a');
+        let re = new RegExp("(https?:\/\/.*)");
+        elements.forEach(function (element) {
+            if (!re.test(element.pathname)) {
+                RemoveCurrentElementSaveChild(element, clear_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_Content, a_Element) {
+        function FinishWorkFunc(a_Element) {
+            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);
             }
@@ -204,12 +234,9 @@
                     let c = childrens[i];
                     content += FinishWorkFunc(GrubTextFuncTemplate()(c), c);
                 }
-                return content
+                return content;
             }
 
-            if (a_OutTag && a_TextAlign) {
-                a_Content = '<' + a_OutTag + ' style = "text-align:' + a_TextAlign + ';">' + a_Content + '</' + a_OutTag + '>';
-            }
             if (a_Element && a_Element.dataset && a_Element.dataset.type == 'quote') {
                 a_Content = '<blockquote>' + a_Content + '</blockquote>';
             }
@@ -244,14 +271,15 @@
         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,
-                'tass_pkg_title--variant_h1_default.*',
+                '(ArticleHeader_titles|tass_pkg_title--variant_h1_default).*',
                 'Image_wrapper_.*',
-                'Paragraph_paragraph.*',
+                '(Paragraph_paragraph|Title_title).*',
                 ElementCheckerTrue,
                 ['/ТАСС/. ']
                 );