def crawl(site, trm , depth, linksfile):
from java.net import URL
from org.w3c.tidy import Tidy
pattern = re.compile('href="/wiki/(.*?)"')
f = open(linksfile, 'a+')
#try:
if depth < MAX_DEPTH:
print 'crawling [%s]...' % trm,
print >> f, '[%s]' % trm
td = Tidy()
td.setXmlOut(1)
u = URL(site + trm)
input = BufferedInputStream(u.openStream())
output = ByteArrayOutputStream()
#tidy.setInputEncoding("UTF8")
#tidy.setOutputEncoding("UTF8")
td.parse(input, output)
content = output.toString()
hits = pattern.findall(content)
for hit in hits:
if hit.find(":") == -1:
print >> f, hit
print 'done.'
print >> f, ''
for hit in hits:
if hit.find(":") == -1:
crawl(site, hit, depth + 1, linksfile)
#except:
# print "wrong"
f.close()
def __tidy(self, content):
tidy = Tidy()
tidy.setIndentAttributes(False)
tidy.setIndentContent(False)
tidy.setPrintBodyOnly(True)
tidy.setSmartIndent(False)
tidy.setWraplen(0)
tidy.setXHTML(True)
tidy.setNumEntities(True)
tidy.setShowWarnings(False)
tidy.setQuiet(True)
out = ByteArrayOutputStream()
tidy.parse(IOUtils.toInputStream(content, "UTF-8"), out)
return out.toString("UTF-8")
def cleanData(data):
from org.w3c.tidy import Tidy
tidy = Tidy()
tidy.setXHTML(True)
tidy.setInputEncoding("UTF-8")
tidy.setOutputEncoding("UTF-8")
tidy.setMakeClean(False)
tidy.setDropEmptyParas(False)
tidy.setPrintBodyOnly(True)
tidy.setQuoteAmpersand(True)
tidy.setTrimEmptyElements(False)
inputStream = ByteArrayInputStream(String(data).getBytes("UTF-8"))
outputStream = ByteArrayOutputStream()
tidy.parseDOM(inputStream, outputStream)
return outputStream.toString("UTF-8")
def __getContent(self, oid):
slash = oid.rfind("/")
pid = os.path.splitext(oid[slash+1:])[0] + ".htm"
payload = Services.storage.getObject(oid).getPayload(pid)
tidy = Tidy()
tidy.setIndentAttributes(False)
tidy.setIndentContent(False)
tidy.setPrintBodyOnly(True)
tidy.setSmartIndent(False)
tidy.setWraplen(0)
tidy.setXHTML(False)
tidy.setNumEntities(True)
out = ByteArrayOutputStream()
try:
doc = tidy.parseDOM(payload.getInputStream(), out)
content = out.toString("UTF-8")
content = self.__processMedia(oid, doc, content)
#print "[\n%s\n]" % content
except Exception, e:
print " * blog.py: Failed to get content: %s" % e.getMessage()
def __tidy(self, content):
tidy = Tidy()
tidy.setIndentAttributes(False)
tidy.setIndentContent(False)
tidy.setPrintBodyOnly(True)
tidy.setSmartIndent(False)
tidy.setWraplen(0)
tidy.setXHTML(False)
tidy.setNumEntities(True)
out = ByteArrayOutputStream()
doc = tidy.parseDOM(ByteArrayInputStream(String(content).getBytes()), out)
content = out.toString("UTF-8")
return content, doc