--- documentViewer/MpdlXmlTextServer.py 2011/03/10 14:03:31 1.231 +++ documentViewer/MpdlXmlTextServer.py 2011/08/05 09:24:42 1.238.2.8 @@ -1,14 +1,62 @@ - from OFS.SimpleItem import SimpleItem from Products.PageTemplates.PageTemplateFile import PageTemplateFile + from Ft.Xml import EMPTY_NAMESPACE, Parse +from Ft.Xml.Domlette import NonvalidatingReader +import Ft.Xml.Domlette +import cStringIO + +import xml.etree.ElementTree as ET -import md5 -import sys +import re import logging import urllib -import documentViewer -from documentViewer import getTextFromNode, serializeNode + +from SrvTxtUtils import getInt, getText, getHttpData + +def serialize(node): + """returns a string containing an XML snippet of node""" + s = ET.tostring(node, 'UTF-8') + # snip off XML declaration + if s.startswith('') + return s[i+3:] + + return s + + +def getTextFromNode(node): + """get the cdata content of a node""" + if node is None: + return "" + # ET: +# text = node.text or "" +# for e in node: +# text += gettext(e) +# if e.tail: +# text += e.tail + + # 4Suite: + nodelist=node.childNodes + text = "" + for n in nodelist: + if n.nodeType == node.TEXT_NODE: + text = text + n.data + + return text + +def serializeNode(node, encoding="utf-8"): + """returns a string containing node as XML""" + #s = ET.tostring(node) + + # 4Suite: + stream = cStringIO.StringIO() + Ft.Xml.Domlette.Print(node, stream=stream, encoding=encoding) + s = stream.getvalue() + stream.close() + + return s + class MpdlXmlTextServer(SimpleItem): """TextServer implementation for MPDL-XML eXist server""" @@ -20,9 +68,7 @@ class MpdlXmlTextServer(SimpleItem): manage_changeMpdlXmlTextServerForm = PageTemplateFile("zpt/manage_changeMpdlXmlTextServer", globals()) - #def __init__(self,id,title="",serverUrl="http://mpdl-proto.mpiwg-berlin.mpg.de/mpdl/interface/", serverName=None, timeout=40): - def __init__(self,id,title="",serverUrl="http://mpdl-test.mpiwg-berlin.mpg.de:30030/mpdl/interface/", serverName=None, timeout=40): - + def __init__(self,id,title="",serverUrl="http://mpdl-text.mpiwg-berlin.mpg.de/mpdl/interface/", serverName=None, timeout=40): """constructor""" self.id=id self.title=title @@ -34,15 +80,17 @@ class MpdlXmlTextServer(SimpleItem): def getHttpData(self, url, data=None): """returns result from url+data HTTP request""" - return documentViewer.getHttpData(url,data,timeout=self.timeout) + return getHttpData(url,data,timeout=self.timeout) def getServerData(self, method, data=None): """returns result from text server for method+data""" url = self.serverUrl+method - return documentViewer.getHttpData(url,data,timeout=self.timeout) + return getHttpData(url,data,timeout=self.timeout) - def getSearch(self, pageinfo=None, docinfo=None, lemma=None): + # WTF: what does this really do? can it be integrated in getPage? + def getSearch(self, pageinfo=None, docinfo=None): """get search list""" + logging.debug("getSearch()") docpath = docinfo['textURLPath'] url = docinfo['url'] pagesize = pageinfo['queryPageSize'] @@ -54,10 +102,10 @@ class MpdlXmlTextServer(SimpleItem): viewMode= pageinfo['viewMode'] tocMode = pageinfo['tocMode'] characterNormalization = pageinfo['characterNormalization'] - optionToggle = pageinfo['optionToggle'] + #optionToggle = pageinfo['optionToggle'] tocPN = pageinfo['tocPN'] selfurl = self.absolute_url() - data = self.getServerData("doc-query.xql","document=%s&mode=%s&queryType=%s&query=%s&queryResultPageSize=%s&queryResultPN=%s&sn=%s&viewMode=%s&characterNormalization=%s&optionToggle=%s&highlightQuery=%s"%(docpath, 'text', queryType, urllib.quote(query), pagesize, pn, sn, viewMode,characterNormalization,optionToggle ,urllib.quote(highlightQuery))) + data = self.getServerData("doc-query.xql","document=%s&mode=%s&queryType=%s&query=%s&queryResultPageSize=%s&queryResultPN=%s&sn=%s&viewMode=%s&characterNormalization=%s&highlightQuery=%s"%(docpath, 'text', queryType, urllib.quote(query), pagesize, pn, sn, viewMode,characterNormalization, urllib.quote(highlightQuery))) pagexml = data.replace('?document=%s'%str(docpath),'?url=%s'%url) pagedom = Parse(pagexml) @@ -85,7 +133,7 @@ class MpdlXmlTextServer(SimpleItem): href = hrefNode.nodeValue if href.startswith('page-fragment.xql'): selfurl = self.absolute_url() - pagexml=href.replace('mode=text','mode=texttool&viewMode=%s&queryType=%s&query=%s&queryResultPageSize=%s&queryResultPN=%s&tocMode=%s&searchPN=%s&tocPN=%s&optionToggle=%s&characterNormalization=%s'%(viewMode,queryType,urllib.quote(query),pagesize,pn,tocMode,pn,tocPN,optionToggle,characterNormalization)) + pagexml=href.replace('mode=text','mode=texttool&viewMode=%s&queryType=%s&query=%s&queryResultPageSize=%s&queryResultPN=%s&tocMode=%s&searchPN=%s&tocPN=%s&characterNormalization=%s'%(viewMode,queryType,urllib.quote(query),pagesize,pn,tocMode,pn,tocPN, characterNormalization)) hrefNode.nodeValue = pagexml.replace('page-fragment.xql','%s'%selfurl) #logging.debug("PUREXML :%s"%(serializeNode(pagenode))) return serializeNode(pagenode) @@ -100,13 +148,13 @@ class MpdlXmlTextServer(SimpleItem): href = hrefNode.nodeValue if href.startswith('page-fragment.xql'): selfurl = self.absolute_url() - pagexml=href.replace('mode=text','mode=texttool&viewMode=%s&queryType=%s&query=%s&queryResultPageSize=%s&queryResultPN=%s&tocMode=%s&searchPN=%s&tocPN=%s&optionToggle=%s&characterNormalization=%s'%(viewMode,queryType,urllib.quote(query),pagesize,pn,tocMode,pn,tocPN,optionToggle,characterNormalization)) + pagexml=href.replace('mode=text','mode=texttool&viewMode=%s&queryType=%s&query=%s&queryResultPageSize=%s&queryResultPN=%s&tocMode=%s&searchPN=%s&tocPN=%s&characterNormalization=%s'%(viewMode,queryType,urllib.quote(query),pagesize,pn,tocMode,pn,tocPN,characterNormalization)) hrefNode.nodeValue = pagexml.replace('page-fragment.xql','%s'%selfurl) if href.startswith('../lt/lemma.xql'): - hrefNode.nodeValue = href.replace('../lt/lemma.xql','%s/template/head_main_lemma_New'%(selfurl)) + hrefNode.nodeValue = href.replace('../lt/lemma.xql','%s/template/head_main_query'%(selfurl)) l.setAttributeNS(None, 'target', '_blank') l.setAttributeNS(None, 'onClick',"popupWin = window.open(this.href, 'contacts', 'location,width=500,height=600,top=180, left=400, scrollbars=1'); return false;") - l.setAttributeNS(None, 'onClick', 'popupWin.focus();') + l.setAttributeNS(None, 'onDblclick', 'popupWin.focus();') pagedivs = pagedom.xpath("//div[@class='queryResultMorphExpansion']") return serializeNode(pagenode) if (queryType=="ftIndex")or(queryType=="ftIndexMorph"): @@ -118,17 +166,17 @@ class MpdlXmlTextServer(SimpleItem): hrefNode = l.getAttributeNodeNS(None, u"href") if hrefNode: href = hrefNode.nodeValue - hrefNode.nodeValue=href.replace('mode=text','mode=texttool&viewMode=%s&tocMode=%s&tocPN=%s&pn=%s&optionToggle=%s&characterNormalization=%s'%(viewMode,tocMode,tocPN,pn,optionToggle,characterNormalization)) + hrefNode.nodeValue=href.replace('mode=text','mode=texttool&viewMode=%s&tocMode=%s&tocPN=%s&pn=%s&characterNormalization=%s'%(viewMode,tocMode,tocPN,pn,characterNormalization)) if href.startswith('../lt/lex.xql'): - hrefNode.nodeValue = href.replace('../lt/lex.xql','%s/template/head_main_voc'%selfurl) + hrefNode.nodeValue = href.replace('../lt/lex.xql','%s/template/head_main_lex'%selfurl) l.setAttributeNS(None, 'target', '_blank') l.setAttributeNS(None, 'onClick',"popupWin = window.open(this.href, 'contacts', 'location,width=500,height=600,top=180, left=400, scrollbars=1'); return false;") - l.setAttributeNS(None, 'onClick', 'popupWin.focus();') + l.setAttributeNS(None, 'onDblclick', 'popupWin.focus();') if href.startswith('../lt/lemma.xql'): - hrefNode.nodeValue = href.replace('../lt/lemma.xql','%s/template/head_main_lemma'%selfurl) + hrefNode.nodeValue = href.replace('../lt/lemma.xql','%s/template/head_main_lemma'%(selfurl)) l.setAttributeNS(None, 'target', '_blank') l.setAttributeNS(None, 'onClick',"popupWin = window.open(this.href, 'contacts', 'location,width=500,height=600,top=180, left=400, scrollbars=1'); return false;") - l.setAttributeNS(None, 'onClick', 'popupWin.focus();') + l.setAttributeNS(None, 'onDblclick', 'popupWin.focus();') return serializeNode(pagenode) return "no text here" @@ -145,12 +193,12 @@ class MpdlXmlTextServer(SimpleItem): hrefList=[] myList= "" text=self.getServerData("xpath.xql", "document=%s&xpath=%s&pn=%s"%(docinfo['textURLPath'],xpath,pn)) - dom = Parse(text) - result = dom.xpath("//result/resultPage/place") + dom = ET.fromstring(text) + result = dom.findall(".//result/resultPage/place") for l in result: - hrefNode= l.getAttributeNodeNS(None, u"id") - href= hrefNode.nodeValue + href = l.get("id") hrefList.append(href) + # WTF: what does this do? myList = ",".join(hrefList) #logging.debug("getGisPlaces :%s"%(myList)) return myList @@ -165,253 +213,233 @@ class MpdlXmlTextServer(SimpleItem): hrefList=[] myList="" text=self.getServerData("xpath.xql", "document=%s&xpath=%s"%(docinfo['textURLPath'],xpath)) - dom =Parse(text) - result = dom.xpath("//result/resultPage/place") + dom = ET.fromstring(text) + result = dom.findall(".//result/resultPage/place") for l in result: - hrefNode = l.getAttributeNodeNS(None, u"id") - href= hrefNode.nodeValue + href = l.get("id") hrefList.append(href) + # WTF: what does this do? myList = ",".join(hrefList) #logging.debug("getALLGisPlaces :%s"%(myList)) return myList + def processPageInfo(self, dom, docinfo, pageinfo): + """processes page info divs from dom and stores in docinfo and pageinfo""" + # assume first second level div is pageMeta + alldivs = dom.find("div") + + if alldivs is None or alldivs.get('class', '') != 'pageMeta': + logging.error("processPageInfo: pageMeta div not found!") + return + + for div in alldivs: + dc = div.get('class') + + # pageNumberOrig + if dc == 'pageNumberOrig': + pageinfo['pageNumberOrig'] = div.text + + # pageNumberOrigNorm + elif dc == 'pageNumberOrigNorm': + pageinfo['pageNumberOrigNorm'] = div.text + + # pageNumberOrigNorm + elif dc == 'countFigureEntries': + docinfo['numFigureEntries'] = getInt(div.text) + + # pageNumberOrigNorm + elif dc == 'countTocEntries': + # WTF: s1 = int(s)/30+1 + docinfo['numTocEntries'] = getInt(div.text) + + # pageHeaderTitle + elif dc == 'pageHeaderTitle': + docinfo['pageHeaderTitle'] = div.text + + # numTextPages + elif dc == 'countPages': + np = getInt(div.text) + if np > 0: + docinfo['numTextPages'] = np + if docinfo.get('numPages', 0) == 0: + # seems to be text-only - update page count + docinfo['numPages'] = np + pageinfo['end'] = min(pageinfo['end'], np) + pageinfo['numgroups'] = int(np / pageinfo['groupsize']) + if np % pageinfo['groupsize'] > 0: + pageinfo['numgroups'] += 1 + + #logging.debug("processPageInfo: pageinfo=%s"%repr(pageinfo)) + return + - def getTextPage(self, mode="text_dict", pn=1, docinfo=None, pageinfo=None): + def getTextPage(self, mode="text", pn=1, docinfo=None, pageinfo=None): """returns single page from fulltext""" + logging.debug("getTextPage mode=%s, pn=%s"%(mode,pn)) + # check for cached text -- but this shouldn't be called twice + if pageinfo.has_key('textPage'): + logging.debug("getTextPage: using cached text") + return pageinfo['textPage'] + docpath = docinfo['textURLPath'] - path = docinfo['textURLPath'] - url = docinfo.get('url',None) - name = docinfo.get('name',None) - pn =pageinfo['current'] - sn = pageinfo['sn'] - optionToggle =pageinfo ['optionToggle'] - highlightQuery = pageinfo['highlightQuery'] - #mode = pageinfo ['viewMode'] - tocMode = pageinfo['tocMode'] - characterNormalization=pageinfo['characterNormalization'] - tocPN = pageinfo['tocPN'] - selfurl = self.absolute_url() - if mode == "text_dict": + # just checking + if pageinfo['current'] != pn: + logging.warning("getTextPage: current!=pn!") + + # stuff for constructing full urls + url = docinfo['url'] + urlmode = docinfo['mode'] + sn = pageinfo.get('sn', None) + highlightQuery = pageinfo.get('highlightQuery', None) + tocMode = pageinfo.get('tocMode', None) + tocPN = pageinfo.get('tocPN',None) + characterNormalization = pageinfo.get('characterNormalization', None) + selfurl = docinfo['viewerUrl'] + + if mode == "dict" or mode == "text_dict": + # dict is called textPollux in the backend textmode = "textPollux" + elif not mode: + # default is text + textmode = "text" else: textmode = mode - textParam = "document=%s&mode=%s&pn=%s&characterNormalization=%s&optionToggle=%s"%(docpath,textmode,pn,characterNormalization,optionToggle) - if highlightQuery is not None: + textParam = "document=%s&mode=%s&pn=%s&characterNormalization=%s"%(docpath,textmode,pn,characterNormalization) + if highlightQuery: textParam +="&highlightQuery=%s&sn=%s"%(urllib.quote(highlightQuery),sn) + # fetch the page pagexml = self.getServerData("page-fragment.xql",textParam) - dom = Parse(pagexml) - #original Pages - pagedivs = dom.xpath("//div[@class='pageNumberOrig']") - """if pagedivs == dom.xpath("//div[@class='pageNumberOrig']"): - if len(pagedivs)>0: - docinfo['pageNumberOrig']= getTextFromNode(pagedivs[0]) - logging.debug("ORIGINAL PAGE: %s"%(docinfo['pageNumberOrig'])) - - #original Pages Norm - pagedivs = dom.xpath("//div[@class='pageNumberOrigNorm']") - if pagedivs == dom.xpath("//div[@class='pageNumberOrigNorm']"): - if len(pagedivs)>0: - docinfo['pageNumberOrigNorm']= getTextFromNode(pagedivs[0]) - logging.debug("ORIGINAL PAGE NORM: %s"%(docinfo['pageNumberOrigNorm'])) - """ - #figureEntries - pagedivs = dom.xpath("//div[@class='countFigureEntries']") - if pagedivs == dom.xpath("//div[@class='countFigureEntries']"): - if len(pagedivs)>0: - docinfo['countFigureEntries'] = getTextFromNode(pagedivs[0]) - s = getTextFromNode(pagedivs[0]) - if s=='0': - try: - docinfo['countFigureEntries'] = int(s) - except: - docinfo['countFigureEntries'] = 0 - else: - s1 = int(s)/30+1 - try: - docinfo['countFigureEntries'] = int(s1) - except: - docinfo['countFigureEntries'] = 0 - - #allPlaces - pagedivs = dom.xpath("//div[@class='countPlaces']") - if pagedivs == dom.xpath("//div[@class='countPlaces']"): - if len(pagedivs)>0: - docinfo['countPlaces']= getTextFromNode(pagedivs[0]) - s = getTextFromNode(pagedivs[0]) - try: - docinfo['countPlaces'] = int(s) - except: - docinfo['countPlaces'] = 0 - - #tocEntries - pagedivs = dom.xpath("//div[@class='countTocEntries']") - if pagedivs == dom.xpath("//div[@class='countTocEntries']"): - if len(pagedivs)>0: - docinfo['countTocEntries'] = int(getTextFromNode(pagedivs[0])) - s = getTextFromNode(pagedivs[0]) - if s=='0': - try: - docinfo['countTocEntries'] = int(s) - except: - docinfo['countTocEntries'] = 0 - else: - s1 = int(s)/30+1 - try: - docinfo['countTocEntries'] = int(s1) - except: - docinfo['countTocEntries'] = 0 - - #numTextPages - pagedivs = dom.xpath("//div[@class='countPages']") - if pagedivs == dom.xpath("//div[@class='countPages']"): - if len(pagedivs)>0: - docinfo['numPages'] = getTextFromNode(pagedivs[0]) - s = getTextFromNode(pagedivs[0]) - - try: - docinfo['numPages'] = int(s) - #logging.debug("PAGE NUMBER: %s"%(s)) - - np = docinfo['numPages'] - pageinfo['end'] = min(pageinfo['end'], np) - pageinfo['numgroups'] = int(np / pageinfo['groupsize']) - if np % pageinfo['groupsize'] > 0: - pageinfo['numgroups'] += 1 - except: - docinfo['numPages'] = 0 - - else: - #no full text -- init to 0 - docinfo['pageNumberOrig'] = 0 - docinfo['countFigureEntries'] = 0 - docinfo['countPlaces'] = 0 - docinfo['countTocEntries'] = 0 - docinfo['numPages'] = 0 - docinfo['pageNumberOrigNorm'] = 0 - #return docinfo + dom = ET.fromstring(pagexml) + # extract additional info + self.processPageInfo(dom, docinfo, pageinfo) + # page content is in