import java.net.*;
import java.io.*;

/**
 * <H2>NetRand Project</H2>
 * <H4>Software Engineering - CS536</H4>
 * <H4>University of Wisconsin - Milwaukee</H4>
 * <H4>Authors:</H4>
 * <UL>
 * <LI>Spring 1998 - Francis William Kasper
 * </UL>
 * <BR>File: BinaryLinkExtractor.java
 * <HR>
 * <P>
 * BinaryLinkExtractor extracts links to binary files from HTML files.
 */
  
public class BinaryLinkExtractor {

		// --------- Class Variables ----------- //

    /** the url of the HTML to parse */
    private URL url;

    /** handles the URL connection */
    private URLConnection urlConn;

    /** the name of the directory to store the cache file */
    private String tmp_dir = "./";

    /** the cache file */
    private File cacheFile;

		// ---------- Class Methods ----------- //

    /** 
     * constructor for the BinaryLinkExtractor
     */
    public BinaryLinkExtractor()
    {
        url = null;
	urlConn = null;
	cacheFile = null;
    }

		// ------------------------------------- //

    /**
     * calls the appropriate functions to extract the binary file links from
     * given URL.
     * @param urlString - the target URL
     * @return a LinkedList containing the URLs to the found binary files
     */
    public LinkedList extractBinaryLinks( String urlString )
    {   
        String baseURL = null;
	String binaryFile = null;
	String tag = null;
 	FileInputStream theHTML = null;
	LinkedList linkList = new LinkedList();

        try {
	    baseURL = cacheHTMLDocument( urlString );
	    baseURL = baseURL.substring(0, baseURL.lastIndexOf('/')+1);
	    theHTML = new FileInputStream( cacheFile );
	   
	    while ( (tag = readTag(theHTML)) != null ) {
	        if ( isBaseTag(tag) ) {
		    baseURL = extract(tag, "href");
                }
		    
	        if ( isHypertextLink(tag) ) {
                    binaryFile = extract(tag, "href"); 
		    if ( binaryFile.indexOf("http") < 0 )
		        binaryFile = baseURL.concat(binaryFile);
	            if ( isBinaryFile(binaryFile) )
		        linkList.append( new String(binaryFile) );
		}

                if ( isImageSourceTag(tag) ) {
		    binaryFile = extract(tag, "src");
                    if ( binaryFile.indexOf("http") < 0 )
                        binaryFile = baseURL.concat(binaryFile);
                    if ( isBinaryFile(binaryFile) )  
                        linkList.append( new String(binaryFile) );
                }
	    }

	    theHTML.close();
	    deleteCacheDocument();
	}
	catch ( Exception e ) {
	    System.err.println( e );
	    return null;
	}

        if ( linkList.size() > 0 )
	    return linkList;
	else
            return null;
    }

		// ------------------------------------ //

    /**
     * gets the HTMl file from the remote serve and writes it to a local cache
     * file
     * @param urlString - the URL of the target HTML file
     * @exception IOException if there was an error reading the HTML file
     */
    private String cacheHTMLDocument( String urlString )
    throws IOException
    {
        String thisLine = null;
	String urlName = null;

	if ( cacheFile != null ) deleteCacheDocument();

	try {
	    url = new URL( urlString );
	    urlConn = url.openConnection();
	    cacheFile = new File( tmp_dir + "html.cache" );

	    BufferedReader in = new BufferedReader( new InputStreamReader(urlConn.getInputStream()) );
	    PrintWriter out = new PrintWriter( new FileOutputStream(cacheFile), true );
	    
	    urlName = new String(urlConn.toString());
	    urlName = urlName.substring(urlName.lastIndexOf("http://"));
	    while ( (thisLine = in.readLine()) != null )
	        out.println(thisLine);
	    
	    in.close();
	    out.close();
	}
        catch ( MalformedURLException me ) {
	    System.err.println( me );
            deleteCacheDocument();
            throw new IOException( urlString + ": is not a URL I understand" );
        }
	catch ( Exception e ) {
            System.err.println( e );
            deleteCacheDocument();
	    throw new IOException( urlName + ":cannot cache HTML file" );
	}

	return urlName;
    }

    		// --------------------------------------- //

    /**
     * deletes the cache file
     */
    private void deleteCacheDocument()
    {
        if ( cacheFile != null ) {
	    cacheFile.delete();
	    cacheFile = null;
	}
    }
    
		// --------------------------------------- //
 
    /**
     * parses for a HTML tag in a HTML file
     * @param theHTML - the HTML file's input stream
     * @return a String containing the HTML tag, or null if no tag was found
     */
    private String readTag( FileInputStream theHTML )
    {
        int c;
	StringBuffer tagBuf = new StringBuffer();

	try {
	    while ( ((c = theHTML.read()) != -1) && ((char)c != '<') );

	    if ( (char)c == '<' ) {
	        while( ((c = theHTML.read()) != -1) && ((char)c != '>') )
		    tagBuf.append((char)c);
            }
 	}
	catch ( IOException e ) {
	    System.err.println( e );
	}

	if ( tagBuf.length() > 0 )
	    return ( new String(tagBuf.toString()) );
	else
	    return null;
    }

		// --------------------------------------- //

    /**
     * extracts the text within quotes from a line immediately after a given key
     * @param line - the line to extract the test from
     * @param key - the key to search for
     * @return a String containing the extracted text 
     */
    private String extract( String line, String key )
    {
	try {
	    key = key.toLowerCase();
	    String lower_case = line.toLowerCase();
	    int i = lower_case.indexOf(key);
	    if ( i < 0 ) return null;
 	    i += key.length();
	    if ( line.charAt(i) != '=') return null;
	    i++;
	    int i2;
            if ( line.charAt(i) == '"' ) 
            {
		i++;
		i2 = line.indexOf('"', i);
		if ( i2 < 0 )
		    return line.substring(i);
                else
		    return line.substring(i, i2);
            }
	    else
            {
		int targ = line.length();
		for ( i2 = i; i2 < targ; i2++ )
		    if ( Character.isSpaceChar(line.charAt(i2)) ) break;
                return line.substring(i, i2);
            }
	}
        catch ( StringIndexOutOfBoundsException e ) {}

	return null;
    }

		// ------------------------------------- //

    /**
     * @param tag - the HTML tag to test
     * @return true iff the tag represents the HTML BASE tag, otherwise false
     */
    private boolean isBaseTag( String tag )
    {
        if ( (tag.indexOf("BASE") != -1 || tag.indexOf("base") != -1) &&
	     (tag.indexOf("HREF") != -1 || tag.indexOf("href") != -1) )
	     return true;
	else
	     return false;
    }

		// ------------------------------------- //

    /**
     * @param tag - the HTML tag to test
     * @return true iff the HTML tag contains a hypertext link, otherwise false
     */
    private boolean isHypertextLink( String tag )
    {
   	if ( tag.indexOf("HREF") != -1 || tag.indexOf("href") != -1 )
            return true;
	else
	    return false;
    }

		// -------------------------------------- //

    /**
     * @param tag - the HTML tag to test
     * @return true iff the HTML tag is an IMG SRC tag, otherwise false
     */
    private boolean isImageSourceTag( String tag )
    {
        if ( (tag.indexOf("IMG") != -1 || tag.indexOf("img") != -1) &&
             (tag.indexOf("SRC") != -1 || tag.indexOf("src") != -1) )
            return true;
        else
	    return false;
    }

		// -------------------------------------- //

    /**
     * @param link - the hypertext link to test
     * @return true iff the link references a binary file, otherwise false
     */
    private boolean isBinaryFile( String link )
    {
	if ( link.endsWith(".au") ||
	    link.endsWith(".wav") ||
	    link.endsWith(".aiff") ||
	    link.endsWith(".aifc") ||
	    link.endsWith(".mpg") ||
  	    link.endsWith(".mpeg") ||
	    link.endsWith(".voc") ||
	    link.endsWith(".ra") ||
	    link.endsWith(".ram") ||
	    link.endsWith(".gif") ||
	    link.endsWith(".jpg") ||
            link.endsWith(".jpeg") ||
	    link.endsWith(".tif") ||
	    link.endsWith(".tiff") ||
	    link.endsWith(".bmp") ||
	    link.endsWith(".mov") ||
	    link.endsWith(".moov") ||
	    link.endsWith(".qt") )
	    return true;

 	return false;
    }

    
}