20

我的问题有点像这个问题,但我有更多的限制:

  • 我知道文件是合理的
  • 他们非常有规律(他们都来自同一个来源
  • 我想要大约 99% 的可见文本
  • 大约 99% 的可行内容是文本(它们或多或少是 RTF 转换为 HTML)
  • 我不关心格式甚至分节符。

是否有任何工具可以做到这一点,或者我最好只是打破 RegexBuddy 和 C#?

我对命令行或批处理工具以及 C/C#/D 库持开放态度。

4

9 回答 9

22

我今天用HTML Agility Pack破解的这段代码将提取无格式的修剪文本。

public static string ExtractText(string html)
{
    if (html == null)
    {
        throw new ArgumentNullException("html");
    }

    HtmlDocument doc = new HtmlDocument();
    doc.LoadHtml(html);

    var chunks = new List<string>(); 

    foreach (var item in doc.DocumentNode.DescendantNodesAndSelf())
    {
        if (item.NodeType == HtmlNodeType.Text)
        {
            if (item.InnerText.Trim() != "")
            {
                chunks.Add(item.InnerText.Trim());
            }
        }
    }
    return String.Join(" ", chunks);
}

如果您想保持某种程度的格式,您可以在源代码提供的示例上进行构建。

public string Convert(string path)
{
    HtmlDocument doc = new HtmlDocument();
    doc.Load(path);

    StringWriter sw = new StringWriter();
    ConvertTo(doc.DocumentNode, sw);
    sw.Flush();
    return sw.ToString();
}

public string ConvertHtml(string html)
{
    HtmlDocument doc = new HtmlDocument();
    doc.LoadHtml(html);

    StringWriter sw = new StringWriter();
    ConvertTo(doc.DocumentNode, sw);
    sw.Flush();
    return sw.ToString();
}

public void ConvertTo(HtmlNode node, TextWriter outText)
{
    string html;
    switch (node.NodeType)
    {
        case HtmlNodeType.Comment:
            // don't output comments
            break;

        case HtmlNodeType.Document:
            ConvertContentTo(node, outText);
            break;

        case HtmlNodeType.Text:
            // script and style must not be output
            string parentName = node.ParentNode.Name;
            if ((parentName == "script") || (parentName == "style"))
                break;

            // get text
            html = ((HtmlTextNode) node).Text;

            // is it in fact a special closing node output as text?
            if (HtmlNode.IsOverlappedClosingElement(html))
                break;

            // check the text is meaningful and not a bunch of whitespaces
            if (html.Trim().Length > 0)
            {
                outText.Write(HtmlEntity.DeEntitize(html));
            }
            break;

        case HtmlNodeType.Element:
            switch (node.Name)
            {
                case "p":
                    // treat paragraphs as crlf
                    outText.Write("\r\n");
                    break;
            }

            if (node.HasChildNodes)
            {
                ConvertContentTo(node, outText);
            }
            break;
    }
}


private void ConvertContentTo(HtmlNode node, TextWriter outText)
{
    foreach (HtmlNode subnode in node.ChildNodes)
    {
        ConvertTo(subnode, outText);
    }
}
于 2012-05-10T03:09:35.197 回答
10

您可以使用支持从 HTML 中提取文本的NUglify

var result = Uglify.HtmlToText("<div>  <p>This is <em>   a text    </em></p>   </div>");
Console.WriteLine(result.Code);   // prints: This is a text

由于它使用的是 HTML5 自定义解析器,因此它应该非常健壮(特别是如果文档不包含任何错误)并且速度非常快(不涉及正则表达式,而是纯递归下降解析器)

于 2016-05-17T02:12:31.310 回答
8

您需要使用HTML Agility Pack

您可能希望使用 LINQ antDescendants调用找到一个元素,然后获取它的InnerText.

于 2010-01-21T23:08:00.807 回答
5

这是我正在使用的代码:

using System.Web;
public static string ExtractText(string html)
{
    Regex reg = new Regex("<[^>]+>", RegexOptions.IgnoreCase);
    string s =reg.Replace(html, " ");
    s = HttpUtility.HtmlDecode(s);
    return s;
}
于 2015-04-08T19:13:39.723 回答
3

这是最好的方法:

  public static string StripHTML(string HTMLText)
    {
        Regex reg = new Regex("<[^>]+>", RegexOptions.IgnoreCase);
        return reg.Replace(HTMLText, "");
    }
于 2012-11-19T17:07:38.870 回答
2

如果将 HTML 加载到 C# 中,然后在 C#/WinForms 中使用 mshtml.dll 或 WebBrowser 控件,则相对简单,然后可以将整个 HTML 文档视为一棵树,遍历树捕获 InnerText 对象。

或者,您也可以使用 document.all,它获取树并将其展平,然后您可以遍历树,再次捕获 InnerText。

这是一个例子:

        WebBrowser webBrowser = new WebBrowser();
        webBrowser.Url = new Uri("url_of_file"); //can be remote or local
        webBrowser.DocumentCompleted += delegate
        {
            HtmlElementCollection collection = webBrowser.Document.All;
            List<string> contents = new List<string>();

            /*
             * Adds all inner-text of a tag, including inner-text of sub-tags
             * ie. <html><body><a>test</a><b>test 2</b></body></html> would do:
             * "test test 2" when collection[i] == <html>
             * "test test 2" when collection[i] == <body>
             * "test" when collection[i] == <a>
             * "test 2" when collection[i] == <b>
             */
            for (int i = 0; i < collection.Count; i++)
            {
                if (!string.IsNullOrEmpty(collection[i].InnerText))
                {
                    contents.Add(collection[i].InnerText);
                }
            }

            /*
             * <html><body><a>test</a><b>test 2</b></body></html>
             * outputs: test test 2|test test 2|test|test 2
             */
            string contentString = string.Join("|", contents.ToArray());
            MessageBox.Show(contentString);
        };

希望有帮助!

于 2010-01-21T23:12:49.587 回答
2

这是我开发的一个类来完成同样的事情。所有可用的 HTML 解析库都太慢了,正则表达式也太慢了。代码注释中解释了功能。根据我的基准测试,在亚马逊的登录页面(包括在下面)上测试时,此代码比 HTML Agility Pack 的等效代码快 10 倍多一点。

/// <summary>
/// The fast HTML text extractor class is designed to, as quickly and as ignorantly as possible,
/// extract text data from a given HTML character array. The class searches for and deletes
/// script and style tags in a first and second pass, with an optional third pass to do the same
/// to HTML comments, and then copies remaining non-whitespace character data to an ouput array.
/// All whitespace encountered is replaced with a single whitespace in to avoid multiple
/// whitespace in the output.
///
/// Note that the returned text content still may have named character and numbered character
/// references within that, when decoded, may produce multiple whitespace.
/// </summary>
public class FastHtmlTextExtractor
{

    private readonly char[] SCRIPT_OPEN_TAG = new char[7] { '<', 's', 'c', 'r', 'i', 'p', 't' };
    private readonly char[] SCRIPT_CLOSE_TAG = new char[9] { '<', '/', 's', 'c', 'r', 'i', 'p', 't', '>' };

    private readonly char[] STYLE_OPEN_TAG = new char[6] { '<', 's', 't', 'y', 'l', 'e' };
    private readonly char[] STYLE_CLOSE_TAG = new char[8] { '<', '/', 's', 't', 'y', 'l', 'e', '>' };

    private readonly char[] COMMENT_OPEN_TAG = new char[3] { '<', '!', '-' };
    private readonly char[] COMMENT_CLOSE_TAG = new char[3] { '-', '-', '>' };

    private int[] m_deletionDictionary;

    public string Extract(char[] input, bool stripComments = false)
    {
        var len = input.Length;
        int next = 0;

        m_deletionDictionary = new int[len];

        // Whipe out all text content between style and script tags.
        FindAndWipe(SCRIPT_OPEN_TAG, SCRIPT_CLOSE_TAG, input);
        FindAndWipe(STYLE_OPEN_TAG, STYLE_CLOSE_TAG, input);

        if(stripComments)
        {
            // Whipe out everything between HTML comments.
            FindAndWipe(COMMENT_OPEN_TAG, COMMENT_CLOSE_TAG, input);
        }

        // Whipe text between all other tags now.
        while(next < len)
        {
            next = SkipUntil(next, '<', input);

            if(next < len)
            {
                var closeNext = SkipUntil(next, '>', input);

                if(closeNext < len)
                {
                    m_deletionDictionary[next] = (closeNext + 1) - next;
                    WipeRange(next, closeNext + 1, input);
                }

                next = closeNext + 1;
            }
        }

        // Collect all non-whitespace and non-null chars into a new
        // char array. All whitespace characters are skipped and replaced
        // with a single space char. Multiple whitespace is ignored.
        var lastSpace = true;
        var extractedPos = 0;
        var extracted = new char[len];

        for(next = 0; next < len; ++next)
        {
            if(m_deletionDictionary[next] > 0)
            {
                next += m_deletionDictionary[next];
                continue;
            }

            if(char.IsWhiteSpace(input[next]) || input[next] == '\0')
            {
                if(lastSpace)
                {
                    continue;
                }

                extracted[extractedPos++] = ' ';
                lastSpace = true;
            }
            else
            {
                lastSpace = false;
                extracted[extractedPos++] = input[next];
            }
        }

        return new string(extracted, 0, extractedPos);
    }

    /// <summary>
    /// Does a search in the input array for the characters in the supplied open and closing tag
    /// char arrays. Each match where both tag open and tag close are discovered causes the text
    /// in between the matches to be overwritten by Array.Clear().
    /// </summary>
    /// <param name="openingTag">
    /// The opening tag to search for.
    /// </param>
    /// <param name="closingTag">
    /// The closing tag to search for.
    /// </param>
    /// <param name="input">
    /// The input to search in.
    /// </param>
    private void FindAndWipe(char[] openingTag, char[] closingTag, char[] input)
    {
        int len = input.Length;
        int pos = 0;

        do
        {
            pos = FindNext(pos, openingTag, input);

            if(pos < len)
            {
                var closenext = FindNext(pos, closingTag, input);

                if(closenext < len)
                {
                    m_deletionDictionary[pos - openingTag.Length] = closenext - (pos - openingTag.Length);
                    WipeRange(pos - openingTag.Length, closenext, input);
                }

                if(closenext > pos)
                {
                    pos = closenext;
                }
                else
                {
                    ++pos;
                }
            }
        }
        while(pos < len);
    }

    /// <summary>
    /// Skips as many characters as possible within the input array until the given char is
    /// found. The position of the first instance of the char is returned, or if not found, a
    /// position beyond the end of the input array is returned.
    /// </summary>
    /// <param name="pos">
    /// The starting position to search from within the input array.
    /// </param>
    /// <param name="c">
    /// The character to find.
    /// </param>
    /// <param name="input">
    /// The input to search within.
    /// </param>
    /// <returns>
    /// The position of the found character, or an index beyond the end of the input array.
    /// </returns>
    private int SkipUntil(int pos, char c, char[] input)
    {
        if(pos >= input.Length)
        {
            return pos;
        }

        do
        {
            if(input[pos] == c)
            {
                return pos;
            }

            ++pos;
        }
        while(pos < input.Length);

        return pos;
    }

    /// <summary>
    /// Clears a given range in the input array.
    /// </summary>
    /// <param name="start">
    /// The start position from which the array will begin to be cleared.
    /// </param>
    /// <param name="end">
    /// The end position in the array, the position to clear up-until.
    /// </param>
    /// <param name="input">
    /// The source array wherin the supplied range will be cleared.
    /// </param>
    /// <remarks>
    /// Note that the second parameter is called end, not lenghth. This parameter is meant to be
    /// a position in the array, not the amount of entries in the array to clear.
    /// </remarks>
    private void WipeRange(int start, int end, char[] input)
    {
        Array.Clear(input, start, end - start);
    }

    /// <summary>
    /// Finds the next occurance of the supplied char array within the input array. This search
    /// ignores whitespace.
    /// </summary>
    /// <param name="pos">
    /// The position to start searching from.
    /// </param>
    /// <param name="what">
    /// The sequence of characters to find.
    /// </param>
    /// <param name="input">
    /// The input array to perform the search on.
    /// </param>
    /// <returns>
    /// The position of the end of the first matching occurance. That is, the returned position
    /// points to the very end of the search criteria within the input array, not the start. If
    /// no match could be found, a position beyond the end of the input array will be returned.
    /// </returns>
    public int FindNext(int pos, char[] what, char[] input)
    {
        do
        {
            if(Next(ref pos, what, input))
            {
                return pos;
            }
            ++pos;
        }
        while(pos < input.Length);

        return pos;
    }

    /// <summary>
    /// Probes the input array at the given position to determine if the next N characters
    /// matches the supplied character sequence. This check ignores whitespace.
    /// </summary>
    /// <param name="pos">
    /// The position at which to check within the input array for a match to the supplied
    /// character sequence.
    /// </param>
    /// <param name="what">
    /// The character sequence to attempt to match. Note that whitespace between characters
    /// within the input array is accebtale.
    /// </param>
    /// <param name="input">
    /// The input array to check within.
    /// </param>
    /// <returns>
    /// True if the next N characters within the input array matches the supplied search
    /// character sequence. Returns false otherwise.
    /// </returns>
    public bool Next(ref int pos, char[] what, char[] input)
    {
        int z = 0;

        do
        {
            if(char.IsWhiteSpace(input[pos]) || input[pos] == '\0')
            {
                ++pos;
                continue;
            }

            if(input[pos] == what[z])
            {
                ++z;
                ++pos;
                continue;
            }

            return false;
        }
        while(pos < input.Length && z < what.Length);

        return z == what.Length;
    }
}

HtmlAgilityPack 中的等价物:

// Where m_whitespaceRegex is a Regex with [\s].
// Where sampleHtmlText is a raw HTML string.

var extractedSampleText = new StringBuilder();
HtmlDocument doc = new HtmlDocument();
doc.LoadHtml(sampleHtmlText);

if(doc != null && doc.DocumentNode != null)
{
    foreach(var script in doc.DocumentNode.Descendants("script").ToArray())
    {
        script.Remove();
    }

    foreach(var style in doc.DocumentNode.Descendants("style").ToArray())
    {
        style.Remove();
    }

    var allTextNodes = doc.DocumentNode.SelectNodes("//text()");
    if(allTextNodes != null && allTextNodes.Count > 0)
    {
        foreach(HtmlNode node in allTextNodes)
        {
            extractedSampleText.Append(node.InnerText);
        }
    }

    var finalText = m_whitespaceRegex.Replace(extractedSampleText.ToString(), " ");
}
于 2017-01-13T15:23:20.373 回答
1

在这里,您可以下载一个工具及其在 HTML 和 XAML 之间来回转换的源代码:XAML/HTML 转换器

它包含一个 HTML 解析器(这样的东西显然必须比您的标准 XML 解析器更宽容),并且您可以遍历与 XML 非常相似的 HTML。

于 2010-01-22T16:03:15.073 回答
1

从命令行,您可以像这样使用Lynx文本浏览器:

如果您想下载格式化输出的网页(即,没有 HTML 标记,而是显示在 Lynx 中),请输入:

lynx -dump URL > filename

如果页面上有任何链接,这些链接的 URL 将包含在下载页面的末尾。

您可以使用禁用链接列表-nolist。例如:

lynx -dump -nolist http://stackoverflow.com/a/10469619/724176 > filename
于 2012-05-06T09:52:10.957 回答