X-Git-Url: http://wamblee.org/gitweb/?a=blobdiff_plain;f=crawler%2Fbasic%2Fsrc%2Forg%2Fwamblee%2Fcrawler%2FAbstractPageRequest.java;h=28482d7fbd096fb8641db821cfcbaab339841726;hb=bc261b857facb7111e9d6ae68da1f5cc2400d21d;hp=0c47430bd87b49af267bd82437c3e9dd4c1282ea;hpb=f917d1e82b25e297fc485c7ca8de4a6f4cc641cd;p=utils diff --git a/crawler/basic/src/org/wamblee/crawler/AbstractPageRequest.java b/crawler/basic/src/org/wamblee/crawler/AbstractPageRequest.java index 0c47430b..28482d7f 100644 --- a/crawler/basic/src/org/wamblee/crawler/AbstractPageRequest.java +++ b/crawler/basic/src/org/wamblee/crawler/AbstractPageRequest.java @@ -17,9 +17,7 @@ package org.wamblee.crawler; import java.io.ByteArrayOutputStream; -import java.io.File; import java.io.IOException; -import java.io.PrintStream; import javax.xml.transform.OutputKeys; import javax.xml.transform.Transformer; @@ -31,11 +29,9 @@ import javax.xml.transform.stream.StreamResult; import org.apache.commons.httpclient.Header; import org.apache.commons.httpclient.HttpClient; -import org.apache.commons.httpclient.HttpException; import org.apache.commons.httpclient.HttpMethod; import org.apache.commons.httpclient.HttpStatus; import org.apache.commons.httpclient.NameValuePair; -import org.apache.commons.httpclient.URIException; import org.apache.commons.httpclient.methods.GetMethod; import org.apache.commons.logging.Log; import org.apache.commons.logging.LogFactory; @@ -43,9 +39,8 @@ import org.apache.xml.serialize.OutputFormat; import org.apache.xml.serialize.XMLSerializer; import org.w3c.dom.Document; import org.w3c.tidy.Tidy; -import org.wamblee.io.FileResource; -import org.wamblee.xml.DOMUtility; -import org.wamblee.xml.XSLT; +import org.wamblee.xml.DomUtils; +import org.wamblee.xml.XslTransformer; /** * General support claas for all kinds of requests. @@ -61,32 +56,44 @@ public abstract class AbstractPageRequest implements PageRequest { private int _maxDelay; private NameValuePair[] _params; + + private NameValuePair[] _headers; private String _xslt; - - private PrintStream _os; + + private XslTransformer _transformer; /** - * Constructs the request. - * @param aMaxTries Maximum retries to perform. - * @param aMaxDelay Maximum delay before executing a request. - * @param aParams Request parameters to use. - * @param aXslt XSLT used to convert the response. - * @param aOs Output stream for logging (if null then no logging is done). + * Constructs the request. + * + * @param aMaxTries + * Maximum retries to perform. + * @param aMaxDelay + * Maximum delay before executing a request. + * @param aParams + * Request parameters to use. + * @param aHeaders + * Request headers to use. + * @param aXslt + * XSLT used to convert the response. */ protected AbstractPageRequest(int aMaxTries, int aMaxDelay, - NameValuePair[] aParams, String aXslt, PrintStream aOs) { + NameValuePair[] aParams, NameValuePair[] aHeaders, String aXslt, XslTransformer aTransformer) { if (aParams == null) { throw new IllegalArgumentException("aParams is null"); } + if (aHeaders == null) { + throw new IllegalArgumentException("aHeaders is null"); + } if (aXslt == null) { throw new IllegalArgumentException("aXslt is null"); } _maxTries = aMaxTries; _maxDelay = aMaxDelay; _params = aParams; + _headers = aHeaders; _xslt = aXslt; - _os = aOs; + _transformer = aTransformer; } /* @@ -99,23 +106,43 @@ public abstract class AbstractPageRequest implements PageRequest { } /** - * Gets the parameters for the request. - * @return Request parameters. + * Gets the parameters for the request. + * + * @return Request parameters. */ protected NameValuePair[] getParameters() { return _params; } + + /** + * Gets the headers for the request. + * @return Request headers. + */ + protected NameValuePair[] getHeaders() { + return _headers; + } /** - * Executes the request with a random delay and with a maximum number of - * retries. - * @param aClient HTTP client to use. - * @param aMethod Method representing the request. - * @return XML document describing the response. - * @throws TransformerException In case transformation of the HTML to XML fails. + * Executes the request with a random delay and with a maximum number of + * retries. + * + * @param aClient + * HTTP client to use. + * @param aMethod + * Method representing the request. + * @return XML document describing the response. + * @throws IOException + * In case of IO problems. + * @throws TransformerException + * In case transformation of the HTML to XML fails. */ protected Document executeMethod(HttpClient aClient, HttpMethod aMethod) - throws TransformerException { + throws IOException, TransformerException { + + for (NameValuePair header: getHeaders()) { + aMethod.setRequestHeader(header.getName(), header.getValue()); + } + int triesLeft = _maxTries; while (triesLeft > 0) { triesLeft--; @@ -131,36 +158,38 @@ public abstract class AbstractPageRequest implements PageRequest { } /** - * Executes the request without doing any retries in case XSLT transformation - * fails. - * @param aClient HTTP client to use. - * @param aMethod Method to execute. - * @return XML document containing the result. - * @throws TransformerException In case transformation of the result to XML fails. + * Executes the request without doing any retries in case XSLT + * transformation fails. + * + * @param aClient + * HTTP client to use. + * @param aMethod + * Method to execute. + * @return XML document containing the result. + * @throws IOException + * In case of IO problems. + * @throws TransformerException + * In case transformation of the result to XML fails. */ protected Document executeMethodWithoutRetries(HttpClient aClient, - HttpMethod aMethod) throws TransformerException { + HttpMethod aMethod) throws IOException, TransformerException { try { aMethod = executeWithRedirects(aClient, aMethod); byte[] xhtmlData = getXhtml(aMethod); - - Document transformed = new XSLT().transform(xhtmlData, - new FileResource(new File(_xslt))); - _os.println("Transformed result is: "); + + Document transformed = _transformer.transform(xhtmlData, + _transformer.resolve(_xslt)); + ByteArrayOutputStream os = new ByteArrayOutputStream(); Transformer transformer = TransformerFactory.newInstance() .newTransformer(); transformer.setParameter(OutputKeys.INDENT, "yes"); transformer.setParameter(OutputKeys.METHOD, "xml"); transformer.transform(new DOMSource(transformed), new StreamResult( - _os)); - + os)); + LOG.debug("Transformed result is \n" + os.toString()); return transformed; - } catch (HttpException e) { - throw new RuntimeException(e.getMessage(), e); - } catch (IOException e) { - throw new RuntimeException(e.getMessage(), e); } catch (TransformerConfigurationException e) { - throw new RuntimeException(e.getMessage(), e); + throw new TransformerException("Transformer configuration problem", e); } finally { // Release the connection. aMethod.releaseConnection(); @@ -168,43 +197,41 @@ public abstract class AbstractPageRequest implements PageRequest { } /** - * Gets the result of the HTTP method as an XHTML document. - * @param aMethod Method to invoke. - * @return XHTML as a byte array. - * @throws URIException In case of poblems with the URI - * @throws IOException In case of problems obtaining the XHTML. + * Gets the result of the HTTP method as an XHTML document. + * + * @param aMethod + * Method to invoke. + * @return XHTML as a byte array. + * @throws IOException + * In case of problems obtaining the XHTML. */ - private byte[] getXhtml(HttpMethod aMethod) throws URIException, IOException { - // Transform the HTML into wellformed XML. - Tidy tidy = new Tidy(); - tidy.setXHTML(true); - tidy.setQuiet(true); - tidy.setShowWarnings(false); - if (_os != null) { - _os.println("Content of '" + aMethod.getURI() + "'"); - _os.println(); - } - // We write the jtidy output to XML since the DOM tree it produces is - // not namespace aware and namespace awareness is required by XSLT. - // An alternative is to configure namespace awareness of the XML parser - // in a system wide way. - Document w3cDoc = tidy.parseDOM(aMethod.getResponseBodyAsStream(), - _os); - DOMUtility.removeDuplicateAttributes(w3cDoc); - - ByteArrayOutputStream xhtml = new ByteArrayOutputStream(); - XMLSerializer serializer = new XMLSerializer(xhtml, new OutputFormat()); - serializer.serialize(w3cDoc); - xhtml.flush(); - if (_os != null) { - _os.println(); - } - return xhtml.toByteArray(); - } + private byte[] getXhtml(HttpMethod aMethod) throws IOException { + // Transform the HTML into wellformed XML. + Tidy tidy = new Tidy(); + tidy.setXHTML(true); + tidy.setQuiet(true); + tidy.setShowWarnings(false); + + // We write the jtidy output to XML since the DOM tree it produces is + // not namespace aware and namespace awareness is required by XSLT. + // An alternative is to configure namespace awareness of the XML parser + // in a system wide way. + ByteArrayOutputStream os = new ByteArrayOutputStream(); + Document w3cDoc = tidy.parseDOM(aMethod.getResponseBodyAsStream(), os); + DomUtils.removeDuplicateAttributes(w3cDoc); + LOG.debug("Content of response is \n" + os.toString()); + + ByteArrayOutputStream xhtml = new ByteArrayOutputStream(); + XMLSerializer serializer = new XMLSerializer(xhtml, new OutputFormat()); + serializer.serialize(w3cDoc); + xhtml.flush(); + + return xhtml.toByteArray(); + } /** * Sleeps for a random time but no more than the maximum delay. - * + * */ private void delay() { try { @@ -215,12 +242,16 @@ public abstract class AbstractPageRequest implements PageRequest { } /** - * Executes the request and follows redirects if needed. - * @param aClient HTTP client to use. - * @param aMethod Method to use. - * @return Final HTTP method used (differs from the parameter passed in in case - * of redirection). - * @throws IOException In case of network problems. + * Executes the request and follows redirects if needed. + * + * @param aClient + * HTTP client to use. + * @param aMethod + * Method to use. + * @return Final HTTP method used (differs from the parameter passed in in + * case of redirection). + * @throws IOException + * In case of network problems. */ private HttpMethod executeWithRedirects(HttpClient aClient, HttpMethod aMethod) throws IOException { @@ -238,11 +269,11 @@ public abstract class AbstractPageRequest implements PageRequest { Header header = aMethod.getResponseHeader(REDIRECT_HEADER); aMethod = new GetMethod(header.getValue()); return executeWithRedirects(aClient, aMethod); // TODO protect - // against infinite - // recursion. + // against infinite + // recursion. } default: { - throw new RuntimeException("Method failed: " + throw new IOException("Method failed: " + aMethod.getStatusLine()); } }