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 re import logging import urllib 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""" meta_type="MPDL-XML TextServer" manage_options=( {'label':'Config','action':'manage_changeMpdlXmlTextServerForm'}, )+SimpleItem.manage_options manage_changeMpdlXmlTextServerForm = PageTemplateFile("zpt/manage_changeMpdlXmlTextServer", globals()) 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 self.timeout = timeout if serverName is None: self.serverUrl = serverUrl else: self.serverUrl = "http://%s/mpdl/interface/"%serverName def getHttpData(self, url, data=None): """returns result from url+data HTTP request""" 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 getHttpData(url,data,timeout=self.timeout) # 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'] pn = pageinfo.get('searchPN',1) sn = pageinfo['sn'] highlightQuery = pageinfo['highlightQuery'] query =pageinfo['query'] queryType =pageinfo['queryType'] viewMode= pageinfo['viewMode'] tocMode = pageinfo['tocMode'] characterNormalization = pageinfo['characterNormalization'] #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&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) """ pagedivs = pagedom.xpath("//div[@class='queryResultHits']") if (pagedivs == pagedom.xpath("//div[@class='queryResultHits']")): if len(pagedivs)>0: docinfo['queryResultHits'] = int(getTextFromNode(pagedivs[0])) s = getTextFromNode(pagedivs[0]) s1 = int(s)/10+1 try: docinfo['queryResultHits'] = int(s1) logging.debug("SEARCH ENTRIES: %s"%(s1)) except: docinfo['queryResultHits'] = 0 """ if (queryType=="fulltext")or(queryType=="xpath")or(queryType=="xquery")or(queryType=="fulltextMorphLemma"): pagedivs = pagedom.xpath("//div[@class='queryResultPage']") if len(pagedivs)>0: pagenode=pagedivs[0] links=pagenode.xpath("//a") for l in links: hrefNode = l.getAttributeNodeNS(None, u"href") if hrefNode: 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&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) if (queryType=="fulltextMorph"): pagedivs = pagedom.xpath("//div[@class='queryResult']") if len(pagedivs)>0: pagenode=pagedivs[0] links=pagenode.xpath("//a") for l in links: hrefNode = l.getAttributeNodeNS(None, u"href") if hrefNode: 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&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_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, 'onDblclick', 'popupWin.focus();') pagedivs = pagedom.xpath("//div[@class='queryResultMorphExpansion']") return serializeNode(pagenode) if (queryType=="ftIndex")or(queryType=="ftIndexMorph"): pagedivs= pagedom.xpath("//div[@class='queryResultPage']") if len(pagedivs)>0: pagenode=pagedivs[0] links=pagenode.xpath("//a") for l in links: 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&characterNormalization=%s'%(viewMode,tocMode,tocPN,pn,characterNormalization)) if href.startswith('../lt/lex.xql'): 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, 'onDblclick', 'popupWin.focus();') if href.startswith('../lt/lemma.xql'): 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, 'onDblclick', 'popupWin.focus();') return serializeNode(pagenode) return "no text here" def getGisPlaces(self, docinfo=None, pageinfo=None): """ Show all Gis Places of whole Page""" xpath='//place' docpath = docinfo.get('textURLPath',None) if not docpath: return None url = docinfo['url'] selfurl = self.absolute_url() pn = pageinfo['current'] hrefList=[] myList= "" text=self.getServerData("xpath.xql", "document=%s&xpath=%s&pn=%s"%(docinfo['textURLPath'],xpath,pn)) dom = ET.fromstring(text) result = dom.findall(".//result/resultPage/place") for l in result: href = l.get("id") hrefList.append(href) # WTF: what does this do? myList = ",".join(hrefList) #logging.debug("getGisPlaces :%s"%(myList)) return myList def getAllGisPlaces (self, docinfo=None, pageinfo=None): """Show all Gis Places of whole Book """ xpath ='//echo:place' docpath =docinfo['textURLPath'] url = docinfo['url'] selfurl =self.absolute_url() pn =pageinfo['current'] hrefList=[] myList="" text=self.getServerData("xpath.xql", "document=%s&xpath=%s"%(docinfo['textURLPath'],xpath)) dom = ET.fromstring(text) result = dom.findall(".//result/resultPage/place") for l in result: 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""" # process all toplevel divs alldivs = dom.findall(".//div") pagediv = None for div in alldivs: dc = div.get('class') # page content div if dc == 'pageContent': pagediv = div # pageNumberOrig elif dc == 'pageNumberOrig': pageinfo['pageNumberOrig'] = div.text # pageNumberOrigNorm elif dc == 'pageNumberOrigNorm': pageinfo['pageNumberOrigNorm'] = div.text # pageNumberOrigNorm elif dc == 'countFigureEntries': docinfo['countFigureEntries'] = getInt(div.text) # pageNumberOrigNorm elif dc == 'countTocEntries': # WTF: s1 = int(s)/30+1 docinfo['countTocEntries'] = getInt(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 return def getTextPage(self, mode="text_dict", 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'] # 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 == "text_dict": # text_dict is called textPollux in the backend textmode = "textPollux" else: textmode = mode 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 = ET.fromstring(pagexml) # extract additional info self.processPageInfo(dom, docinfo, pageinfo) # page content is in