Screen-scraping was a popular method for slowly converting mainframe applications into applications that would run on PCs. The application would connect to the mainframe, read data from the screen, and re-display it in a Windows-based application. Data entered into the Windows application would then be transmitted back to the mainframe.
If you have a Web-based application that doesn’t support Web services, you can do a Web-based screen scraping using the HttpWebRequest and HttpWebResponse covered in a previous tip. The example in this tip posts a query to the Weather Channel and extracts the current temperature from the data sent back to the Web page. Here’s the code you can put into a Web page for testing:
public partial class PostingData : System.Web.UI.Page
protected void Page_Load(object sender, EventArgs e)
string outputBuffer = “where=46038”;
HttpWebRequest req =
req.Method = “POST”;
req.ContentLength = outputBuffer.Length;
req.ContentType = “application/x-www-form-urlencoded”;
StreamWriter swOut = new StreamWriter(req.GetRequestStream());
HttpWebResponse resp = (HttpWebResponse)req.GetResponse();
StreamReader sr = new StreamReader(resp.GetResponseStream());
string buffer = sr.ReadToEnd();
int start = 0, end = 0;
string startTag = “<B CLASS=obsTempTextA>”;
string endTag = “</B>”;
start = buffer.IndexOf(startTag, StringComparison.CurrentCultureIgnoreCase);
end = buffer.IndexOf(endTag, start, StringComparison.CurrentCultureIgnoreCase);
Response.Write(“Current temperature in ZIP 46038: ”
+ buffer.Substring(start + startTag.Length, end – start – startTag.Length));
It starts by creating a HttpWebRequest to weather.com’s search URL, which I found by looking at its home page search form. Part of the “fun” of webscraping is trying to figure out what all has to be sent on a post in order to get back valid results. In this case, you have to send only a value of where with the ZIP code you want to query. That information is stored in the outputBuffer variable in POST format, which means each name/value pair is separated by ampersands, similar to what you would see in a query string.
Next, the example populates the request with the post information and then requests the response, which has the effect of sending the data to the remote server. It retrieves the information into a string buffer and closes up the response stream.
This, unfortunately, is the tedious part of webscraping. You have to find the information you want in the response buffer. For this page, the resulting HTML (which can be dumped out to the page using the commented line at the end of the code) is 224KB of HTML to search through. However, the data you want is stashed between a reasonably easy tag to find. Using some simple string manipulation, you can extract the value and show it on the screen.
As you might guess, this is fairly “fragile” code. If the Weather Channel decides to change its page or the tag you’re looking for, the code will fail to find the information it needs. That’s one of the major reasons why Web services have become popular. The Weather Channel’s page is designed for humans to read, not computers. The Web services that handle weather, on the other hand, send back only the relevant content and not all the formatting found in the page. However, if you don’t have another option, webscraping can be a handy tool.
About the Author
Eric Smith is the owner of Northstar Computer Systems, a Web-hosting company based in Indianapolis, Indiana. He is also a MCT and MCSD who has been developing with .NET since 2001. In addition, he has written or contributed to 12 books covering .NET, ASP, and Visual Basic. Send him your questions and feedback via e-mail at [email protected].