private Hashtable<String, Integer> extractData(String docUrl, Page page) throws Exception {
Hashtable<String, Integer> businessTable = new Hashtable<String, Integer>();
String currentBusiness = "";
Lexer lexer = new Lexer(page);
while (true) {
Node node = lexer.nextNode();
if (node == null) {
break;
}
if (node instanceof TagNode) {
TagNode tagNode = (TagNode) node;
if (tagNode.getTagName().equals("A")) {
String href = tagNode.getAttribute("href");
if (href != null) {
String absUrl = AbsUrlConstructor.construct(docUrl, href);
Crawler.dispatchUrl(absUrl);
Pattern pBusiness = Pattern.compile("^(http://www.yelp.com/biz/)(\\S)+");
if (pBusiness.matcher(absUrl).matches()) {
currentBusiness = extractBusinessName(href);
if (!businessTable.containsKey(currentBusiness)) {
businessTable.put(currentBusiness, -1);
}
// System.out.println("currentBusiness = "+currentBusiness);
// rating = "4";
// UpdateDatabase(linkID, business, rating, userID);
// System.out.println(business + " added.");
}
}
} else if (tagNode.getTagName().equals("IMG")) {
String c1ass2 = tagNode.getAttribute("class");
if (c1ass2 != null) {
String rating = "";
String[] rate = c1ass2.split("_");
int num = rate.length - 1;
if (!rate[num].equals("loader")) {
rating = rate[num].trim();
if (businessTable.get(currentBusiness) == -1) {
businessTable.put(currentBusiness, Integer.parseInt(rating));
}
}
// System.out.println(linkID + " " + business + " " + rating + " " + userID;
}
}
}
}
return businessTable;
}
private boolean isNotTagAppearingInHead(TagNode node) {
String thisTag = node.getTagName();
if (thisTag.startsWith("!")) return false;
for (String tag : okHeadTags) {
if (thisTag.equals(tag)) {
return false;
}
}
return true;
}
protected boolean checkAllowTag(ParseContext context, TagNode tagNode) {
String tagName = tagNode.getTagName();
// Check the NOSCRIPT tag, if force-noscript is set,
// then skip the NOSCRIPT tags and include contents explicitly
if (tagName.equals("NOSCRIPT")) {
String allPolicies = context.getOraclePolicy();
if ((allPolicies != null) && allPolicies.contains("force-noscript")) {
return false;
}
}
return true;
}
public void handleOpenTagNode(ParseContext pContext, TagNode node) throws IOException {
ReplayParseContext context = (ReplayParseContext) pContext;
if (context.getData(FERRET_DONE_KEY) == null) {
// we haven't emitted yet:
// are we running in post-emit?
if (context.getPhase() == ReplayParseEventDelegator.PHASE_POST_OUTPUT) {
// emit if it is a body tag:
if (node.getTagName().equals("BODY")) {
emit((ReplayParseContext) context, node);
}
} else {
// must be PHASE_PRE_MODIFY: if it's a body tag, emit now:
if (isNotTagAppearingInHead(node)) {
if (node.getTagName().equals(FRAMESET_TAG)) {
// don't put content in pages with a FRAMESET:
context.putData(FERRET_DONE_KEY, "1");
} else {
// and this is a tag that shouldn't be in the HEAD. Emit:
emit((ReplayParseContext) context, node);
}
}
}
}
}
// TODO: This should all be refactored up into an abstract base class with
// default no-op methods, allowing a subclass to only override the ones they
// want...
public void handleNode(ParseContext pContext, Node node) throws IOException {
ReplayParseContext context = (ReplayParseContext) pContext;
if (NodeUtils.isRemarkNode(node)) {
RemarkNode remarkNode = (RemarkNode) node;
remarkNode.setText(jsBlockTrans.transform(context, remarkNode.getText()));
emit(context, null, node, null);
} else if (NodeUtils.isTextNode(node)) {
TextNode textNode = (TextNode) node;
if (context.isInCSS()) {
handleCSSTextNode(context, textNode);
} else if (context.isInScriptText()) {
handleJSTextNode(context, textNode);
} else {
emit(context, null, textNode, null);
// handleContentTextNode(context,textNode);
}
} else if (NodeUtils.isTagNode(node)) {
TagNode tagNode = (TagNode) node;
if (NodeUtils.isOpenTagNodeNamed(tagNode, NodeUtils.SCRIPT_TAG_NAME)) {
handleJSIncludeNode(context, tagNode);
} else if (tagNode.isEndTag()) {
if (tagNode.getTagName().equals("HEAD")) {
context.putData(FERRET_IN_HEAD, null);
}
if (checkAllowTag(pContext, tagNode)) {
emit(context, null, tagNode, null);
}
// handleCloseTagNode(context,tagNode);
} else {
// assume start, possibly empty:
handleOpenTagNode(context, tagNode);
}
} else {
throw new IllegalArgumentException("Unknown node type..");
}
}
private void handleOpenTagNode(ReplayParseContext context, TagNode tagNode) throws IOException {
boolean insertedJsp = context.getData(FERRET_DONE_KEY) != null;
String preEmit = null;
String postEmit = null;
String tagName = tagNode.getTagName();
boolean alreadyInsertedHead = (context.getData(FERRET_HEAD_INSERTED) != null);
if (!alreadyInsertedHead) {
// If we're at the beginning of a <head> tag, and haven't inserted yet,
// insert right AFTER head tag
if (tagName.equals("HEAD")) {
emitHeadInsert(context, tagNode, true);
context.putData(FERRET_IN_HEAD, FERRET_IN_HEAD);
return;
}
// If we're at the beginning of any tag, other than <html>,
// (including <body>) and haven't inserted yet,
// insert right BEFORE the next tag, also continue other default processing
// of the tag
if (!tagName.equals("HTML") && !tagName.equals("!DOCTYPE")) {
emitHeadInsert(context, null, false);
// Don't return continue to further processing
}
}
boolean inHead = (context.getData(FERRET_IN_HEAD) != null);
// Time to insert the JSP header?
// IK added check to avoid inserting inside css or script
if (!insertedJsp && !context.isInCSS() && !context.isInScriptText() && !inHead) {
if (!okHeadTagMap.containsKey(tagName)) {
if (tagName.equals(FRAMESET_TAG)) {
// don't put the insert in framsets:
} else {
if (jspInsertPath != null
&& !context.getJspExec().getUiResults().getWbRequest().isIFrameWrapperContext()) {
String tmp = null;
try {
tmp = context.getJspExec().jspToString(jspInsertPath);
} catch (ServletException e) {
e.printStackTrace();
}
if (tagName.equals(BODY_TAG)) {
// insert it now, *after* the current Tag:
postEmit = tmp;
} else {
// hrm... we are seeing a node that should be in
// the body.. lets emit the jsp now, *before*
// the current Tag:
preEmit = tmp;
}
}
}
context.putData(FERRET_DONE_KEY, "");
}
}
// now do all the usual attribute rewriting:
// this could be slightly optimized by moving tags more likely to occur
// to the front of the if/else if/else if routing...
if (tagName.equals("A")) {
transformAttr(context, tagNode, "HREF", anchorUrlTrans);
} else if (tagName.equals("APPLET")) {
transformAttr(context, tagNode, "CODEBASE", objectEmbedUrlTrans);
transformAttr(context, tagNode, "ARCHIVE", objectEmbedUrlTrans);
} else if (tagName.equals("AREA")) {
transformAttr(context, tagNode, "HREF", anchorUrlTrans);
} else if (tagName.equals("BASE")) {
String orig = tagNode.getAttribute("HREF");
if (orig != null) {
try {
context.setBaseUrl(new URL(orig));
transformAttr(context, tagNode, "HREF", anchorUrlTrans);
} catch (MalformedURLException e) {
e.printStackTrace();
}
}
} else if (tagName.equals("EMBED")) {
transformAttr(context, tagNode, "SRC", objectEmbedUrlTrans);
} else if (tagName.equals("IFRAME")) {
transformAttr(context, tagNode, "SRC", iframeUrlTrans);
} else if (tagName.equals("IMG")) {
transformAttr(context, tagNode, "SRC", imageUrlTrans);
} else if (tagName.equals("INPUT")) {
transformAttr(context, tagNode, "SRC", imageUrlTrans);
} else if (tagName.equals("FORM")) {
transformAttr(context, tagNode, "ACTION", anchorUrlTrans);
} else if (tagName.equals("FRAME")) {
transformAttr(context, tagNode, "SRC", framesetUrlTrans);
} else if (tagName.equals("LINK")) {
if (transformAttrWhere(context, tagNode, "REL", "STYLESHEET", "HREF", cssUrlTrans)) {
// no-op
} else if (transformAttrWhere(
context, tagNode, "REL", "SHORTCUT ICON", "HREF", imageUrlTrans)) {
// no-op
} else {
transformAttr(context, tagNode, "HREF", anchorUrlTrans);
}
} else if (tagName.equals("META")) {
transformAttrWhere(context, tagNode, "HTTP-EQUIV", "REFRESH", "CONTENT", metaRefreshTrans);
transformAttr(context, tagNode, "URL", anchorUrlTrans);
} else if (tagName.equals("OBJECT")) {
transformAttr(context, tagNode, "CODEBASE", objectEmbedUrlTrans);
transformAttr(context, tagNode, "CDATA", objectEmbedUrlTrans);
} else if (tagName.equals("SCRIPT")) {
transformAttr(context, tagNode, "SRC", jsUrlTrans);
} else if (tagName.equals("DIV") || tagName.equals("LI")) {
// HTML5 -- can have data-src or data-uri attributes in any tag!
// Can really be in any tag but for now using most common use cases
// Experimental
transformAttr(context, tagNode, "data-src", objectEmbedUrlTrans);
transformAttr(context, tagNode, "data-uri", objectEmbedUrlTrans);
} else {
if (!checkAllowTag(context, tagNode)) {
return;
}
}
// now, for *all* tags...
transformAttr(context, tagNode, "BACKGROUND", imageUrlTrans);
transformAttr(context, tagNode, "STYLE", cssInlineTrans);
transformAttr(context, tagNode, "onclick", jsBlockTrans);
transformAttr(context, tagNode, "onload", jsBlockTrans);
emit(context, preEmit, tagNode, postEmit);
}
