/// <summary>
/// Extracts the complete html table the given path is pointing to. If the path points
/// to a cell of a table the complete table is extracted still.
/// <remarks>
/// Returns null if table not found by path. Currently we cannot handle thead
/// and tfoot. The number of the column is defined by the html table row with the most
/// html columns
/// </remarks>
/// </summary>
/// <param name="doc">the HTML document</param>
/// <param name="path">the path to the table</param>
/// <param name="textOnly">set this to true to get only the text of the cell, otherwise the
/// cell itself as HtmlElement is returned</param>
public static FallibleActionResult <DataTable> ExtractTable(this IHtmlDocument doc, HtmlPath path, bool textOnly)
{
doc.Require(x => doc != null);
path.Require(x => path != null);
HtmlTable htmlTable = doc.GetTableByPath(path);
if (htmlTable == null)
{
return(FallibleActionResult <DataTable> .CreateFailureResult("Could not get table by path"));
}
DataTable table = new DataTable();
// TODO: should we get the culture from the HTML page somehow?
table.Locale = CultureInfo.InvariantCulture;
Func <IHtmlElement, object> GetContent = element => (textOnly ? (object)element.InnerText : element);
foreach (var tr in htmlTable.Rows)
{
var htmlRow = new List <IHtmlElement>();
foreach (var td in tr.Children)
{
if (td.TagName == "TD" || td.TagName == "TH")
{
htmlRow.Add(td);
}
}
// add columns if necessary
if (htmlRow.Count > table.Columns.Count)
{
(htmlRow.Count - table.Columns.Count).Times(x => table.Columns.Add(string.Empty, typeof(object)));
}
// add new row to table
DataRow row = table.NewRow();
table.Rows.Add(row);
table.AcceptChanges();
// add data
htmlRow.ForeachIndex((element, idx) => row[idx] = GetContent(element));
}
if (table.Rows.Count == 0)
{
table.Dispose();
return(FallibleActionResult <DataTable> .CreateFailureResult("Table was empty"));
}
return(FallibleActionResult <DataTable> .CreateSuccessResult(table));
}
/// <summary>
/// Returns the element specified by the given <see cref="HtmlPath"/>.
/// </summary>
public static IHtmlElement GetElementByPath(this IHtmlDocument doc, HtmlPath path)
{
doc.Require(d => doc != null);
path.Require(p => path != null);
var root = doc.Body.GetRoot();
if (root == null)
{
return(null);
}
foreach (var element in path.Elements)
{
root = root.GetChildAt(element.TagName, element.Position);
if (root == null)
{
return(null);
}
}
return(root);
}
