Sample ColdFusion script to parse a webpage and extract table data using jsoup.

JsoupTableParse.cfm

<cfhttp url="http://target_website_with-table.com/" username="#CGI.Http_User_Agent#"></cfhttp>


<cfscript>
jsoup = CreateObject("java", "org.jsoup.Jsoup");
HTMLDocument = jsoup.parse(CFHTTP.fileContent);

/* Identify a specific table containing the data to scrape */
TheTable = HTMLDocument.select("##tableByID");
/*  Alternate DOM select methods if table doesn't have a unique ID
TheTable = HTMLDocument.select("table.tableClass").first();
TheTable = HTMLDocument.select("table").first();   */

/* Auto sanitize HTML. Removes most attributes  http://jsoup.org/apidocs/org/jsoup/safety/Whitelist.html
Whitelist = CreateObject("java", "org.jsoup.safety.Whitelist");
TheTable = jsoup.clean(TheTable.toString(), Whitelist.relaxed());
*/

/* Manually remove all class and other misc attributes from table (optional) */
TheTable.removeAttr("class").addClass("cellpadding3").removeAttr("width").removeAttr("cellspacing");
/* remove misc attributes from TH/TD (optional) */
TheTable.select("td,th").removeAttr("colspan").removeAttr("rowspan").removeAttr("align").removeAttr("class");

/* Add CSS class to table */
TheTable.addClass("hTable");

/* for tables with only TD, convert first column to TH using tagName() (optional)
   http://jsoup.org/apidocs/org/jsoup/nodes/Element.html */
rows = TheTable.select("tr");
for ( row in rows ){
	TheTable.select("tr td:eq(0),tr th:eq(0)").tagName("th");
}

/* Add thead, header row and cells (optional) */
TheTable.prepend("<thead><tr><th>Col1</th><th>Col2</th></tr></thead>")

/* remove last row (optional) */
TheTable.select("tr").last().remove();

writeOutput(TheTable.toString());
</cfscript>