from OFS.SimpleItem import SimpleItem from Products.PageTemplates.PageTemplateFile import PageTemplateFile import xml.etree.ElementTree as ET import re import logging import urllib import base64 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 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) def getPlacesOnPage(self, docinfo=None, pn=None): """Returns list of GIS places of page pn""" docpath = docinfo.get('textURLPath',None) if not docpath: return None places=[] text=self.getServerData("xpath.xql", "document=%s&xpath=//place&pn=%s"%(docpath,pn)) dom = ET.fromstring(text) result = dom.findall(".//resultPage/place") for l in result: id = l.get("id") name = l.text place = {'id': id, 'name': name} places.append(place) return places 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 # pageHeaderTitle elif dc == 'pageHeaderTitle': pageinfo['pageHeaderTitle'] = div.text # numFigureEntries elif dc == 'countFigureEntries': docinfo['numFigureEntries'] = getInt(div.text) # numTocEntries elif dc == 'countTocEntries': # WTF: s1 = int(s)/30+1 docinfo['numTocEntries'] = getInt(div.text) # numPlaces elif dc == 'countPlaces': docinfo['numPlaces'] = 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 #logging.debug("processPageInfo: pageinfo=%s"%repr(pageinfo)) return 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 ideally 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 characterNormalization = pageinfo.get('characterNormalization', None) moreTextParam = '' selfurl = docinfo['viewerUrl'] if not mode: # default is dict mode = 'text' modes = mode.split(',') # check for multiple layers if len(modes) > 1: logging.debug("getTextPage: more than one mode=%s"%mode) # search mode if 'search' in modes: # add highlighting highlightQuery = pageinfo.get('highlightQuery', None) sn = pageinfo.get('sn', None) if highlightQuery and sn: moreTextParam +="&highlightQuery=%s&sn=%s"%(urllib.quote(highlightQuery),sn) # remove mode modes.remove('search') # other modes don't combine if 'dict' in modes: # dict is called textPollux in the backend textmode = 'textPollux' elif len(modes) == 0: # text is default mode textmode = 'text' else: # just take first mode textmode = modes[0] textParam = "document=%s&mode=%s&pn=%s&characterNormalization=%s"%(docpath,textmode,pn,characterNormalization) textParam += moreTextParam # 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
pagediv = None # ElementTree 1.2 in Python 2.6 can't do div[@class='pageContent'] # so we look at the second level divs alldivs = dom.findall("div") for div in alldivs: dc = div.get('class') # page content div if dc == 'pageContent': pagediv = div break # plain text mode if textmode == "text": # get full url assuming documentViewer is parent selfurl = self.getLink() if pagediv is not None: links = pagediv.findall(".//a") for l in links: href = l.get('href') if href and href.startswith('#note-'): href = href.replace('#note-',"%s#note-"%selfurl) l.set('href', href) return serialize(pagediv) # text-with-links mode elif textmode == "textPollux": if pagediv is not None: viewerurl = docinfo['viewerUrl'] selfurl = self.getLink() # check all a-tags links = pagediv.findall(".//a") for l in links: href = l.get('href') if href: # is link with href if href.startswith('http://mpdl-proto.mpiwg-berlin.mpg.de/mpdl/interface/lt/wordInfo.xql'): # is dictionary link - change href (keeping parameters) l.set('href', href.replace('http://mpdl-proto.mpiwg-berlin.mpg.de/mpdl/interface/lt/wordInfo.xql','%s/template/viewer_wordinfo'%viewerurl)) # add target to open new page l.set('target', '_blank') # TODO: is this needed? if href.startswith('http://mpdl-proto.mpiwg-berlin.mpg.de/mpdl/lt/lemma.xql'): selfurl = self.absolute_url() l.set('href', href.replace('http://mpdl-proto.mpiwg-berlin.mpg.de/mpdl/lt/lemma.xql','%s/head_main_lemma'%selfurl)) l.set('target', '_blank') l.set('onclick',"popupWin = window.open(this.href, 'InfoWindow', 'menubar=no, location,width=500,height=600,top=180, left=700, toolbar=no, scrollbars=1'); return false;") l.set('ondblclick', 'popupWin.focus();') if href.startswith('#note-'): # note link l.set('href', href.replace('#note-',"%s#note-"%selfurl)) return serialize(pagediv) # xml mode elif textmode == "xml": if pagediv is not None: return serialize(pagediv) # pureXml mode elif textmode == "pureXml": if pagediv is not None: return serialize(pagediv) # gis mode elif textmode == "gis": if pagediv is not None: # check all a-tags links = pagediv.findall(".//a") # add our URL as backlink selfurl = self.getLink() doc = base64.b64encode(selfurl) for l in links: href = l.get('href') if href: if href.startswith('http://mappit.mpiwg-berlin.mpg.de'): l.set('href', re.sub(r'doc=[\w+/=]+', 'doc=%s'%doc, href)) l.set('target', '_blank') return serialize(pagediv) return None def getSearchResults(self, mode, query=None, pageinfo=None, docinfo=None): """loads list of search results and stores XML in docinfo""" logging.debug("getSearchResults mode=%s query=%s"%(mode, query)) if mode == "none": return docinfo if 'resultSize_%s_%s'%(mode,query) in docinfo: # cached result return docinfo docpath = docinfo['textURLPath'] # fetch full results params = {'document': docpath, 'mode': 'text', 'queryType': mode, 'query': query, 'queryResultPageSize': 1000, 'queryResultPN': 1, 'characterNormalization': pageinfo.get('characterNormalization', 'reg')} pagexml = self.getServerData("doc-query.xql",urllib.urlencode(params)) #pagexml = self.getServerData("doc-query.xql","document=%s&mode=%s&queryType=%s&query=%s&queryResultPageSize=%s&queryResultPN=%s&s=%s&viewMode=%s&characterNormalization=%s&highlightElementPos=%s&highlightElement=%s&highlightQuery=%s"%(docpath, 'text', queryType, urllib.quote(query), pagesize, pn, s, viewMode,characterNormalization, highlightElementPos, highlightElement, urllib.quote(highlightQuery))) dom = ET.fromstring(pagexml) # page content is in
pagediv = None # ElementTree 1.2 in Python 2.6 can't do div[@class='queryResultPage'] alldivs = dom.findall("div") for div in alldivs: dc = div.get('class') # page content div if dc == 'queryResultPage': pagediv = div elif dc == 'queryResultHits': docinfo['resultSize_%s_%s'%(mode,query)] = getInt(div.text) if pagediv is not None: # store XML in docinfo docinfo['resultXML_%s_%s'%(mode,query)] = ET.tostring(pagediv, 'UTF-8') return docinfo def getResultsPage(self, mode="text", query=None, pn=None, start=None, size=None, pageinfo=None, docinfo=None): """returns single page from the table of contents""" logging.debug("getResultsPage mode=%s, pn=%s"%(mode,pn)) # check for cached TOC #TODO: cache only one search if not docinfo.has_key('resultXML_%s_%s'%(mode,query)): self.getSearchResults(mode=mode, query=query, pageinfo=pageinfo, docinfo=docinfo) resultxml = docinfo.get('resultXML_%s_%s'%(mode,query), None) if not resultxml: logging.error("getResultPage: unable to find resultXML") return "Error: no result!" if size is None: size = pageinfo.get('searchResultPageSize', 20) if start is None: start = (pn - 1) * size fullresult = ET.fromstring(resultxml) if fullresult is not None: # paginate first = start len = size del fullresult[:first] del fullresult[len:] tocdivs = fullresult # check all a-tags links = tocdivs.findall(".//a") for l in links: href = l.get('href') if href: # take pn from href m = re.match(r'page-fragment\.xql.*pn=(\d+)', href) if m is not None: # and create new url (assuming parent is documentViewer) #TODO: add highlighting params url = self.getLink('pn', m.group(1)) l.set('href', url) else: logging.warning("getResultPage: Problem with link=%s"%href) # fix two-divs-per-row with containing div # newtoc = ET.Element('div', {'class':'queryResultPage'}) # for (d1,d2) in zip(tocdivs[::2],tocdivs[1::2]): # e = ET.Element('div',{'class':'tocline'}) # e.append(d1) # e.append(d2) # newtoc.append(e) return serialize(tocdivs) return "ERROR: no results!" def getToc(self, mode="text", docinfo=None): """loads table of contents and stores XML in docinfo""" logging.debug("getToc mode=%s"%mode) if mode == "none": return docinfo if 'tocSize_%s'%mode in docinfo: # cached toc return docinfo docpath = docinfo['textURLPath'] # we need to set a result set size pagesize = 1000 pn = 1 if mode == "text": queryType = "toc" else: queryType = mode # number of entries in toc tocSize = 0 tocDiv = None # fetch full toc pagexml = self.getServerData("doc-query.xql","document=%s&queryType=%s&queryResultPageSize=%s&queryResultPN=%s"%(docpath,queryType, pagesize, pn)) dom = ET.fromstring(pagexml) # page content is in
pagediv = None # ElementTree 1.2 in Python 2.6 can't do div[@class='queryResultPage'] alldivs = dom.findall("div") for div in alldivs: dc = div.get('class') # page content div if dc == 'queryResultPage': pagediv = div elif dc == 'queryResultHits': docinfo['tocSize_%s'%mode] = getInt(div.text) if pagediv is not None: # store XML in docinfo docinfo['tocXML_%s'%mode] = ET.tostring(pagediv, 'UTF-8') return docinfo def getTocPage(self, mode="text", pn=None, start=None, size=None, pageinfo=None, docinfo=None): """returns single page from the table of contents""" logging.debug("getTocPage mode=%s, pn=%s"%(mode,pn)) if mode == "text": queryType = "toc" else: queryType = mode # check for cached TOC if not docinfo.has_key('tocXML_%s'%mode): self.getToc(mode=mode, docinfo=docinfo) tocxml = docinfo.get('tocXML_%s'%mode, None) if not tocxml: logging.error("getTocPage: unable to find tocXML") return "Error: no table of contents!" if size is None: size = pageinfo.get('tocPageSize', 30) if start is None: start = (pn - 1) * size fulltoc = ET.fromstring(tocxml) if fulltoc is not None: # paginate first = (start - 1) * 2 len = size * 2 del fulltoc[:first] del fulltoc[len:] tocdivs = fulltoc # check all a-tags links = tocdivs.findall(".//a") for l in links: href = l.get('href') if href: # take pn from href m = re.match(r'page-fragment\.xql.*pn=(\d+)', href) if m is not None: # and create new url (assuming parent is documentViewer) url = self.getLink('pn', m.group(1)) l.set('href', url) else: logging.warning("getTocPage: Problem with link=%s"%href) # fix two-divs-per-row with containing div newtoc = ET.Element('div', {'class':'queryResultPage'}) for (d1,d2) in zip(tocdivs[::2],tocdivs[1::2]): e = ET.Element('div',{'class':'tocline'}) e.append(d1) e.append(d2) newtoc.append(e) return serialize(newtoc) return "ERROR: no table of contents!" def manage_changeMpdlXmlTextServer(self,title="",serverUrl="http://mpdl-text.mpiwg-berlin.mpg.de/mpdl/interface/",timeout=40,RESPONSE=None): """change settings""" self.title=title self.timeout = timeout self.serverUrl = serverUrl if RESPONSE is not None: RESPONSE.redirect('manage_main') # management methods def manage_addMpdlXmlTextServerForm(self): """Form for adding""" pt = PageTemplateFile("zpt/manage_addMpdlXmlTextServer", globals()).__of__(self) return pt() def manage_addMpdlXmlTextServer(self,id,title="",serverUrl="http://mpdl-text.mpiwg-berlin.mpg.de/mpdl/interface/",timeout=40,RESPONSE=None): #def manage_addMpdlXmlTextServer(self,id,title="",serverUrl="http://mpdl-text.mpiwg-berlin.mpg.de:30030/mpdl/interface/",timeout=40,RESPONSE=None): """add zogiimage""" newObj = MpdlXmlTextServer(id,title,serverUrl,timeout) self.Destination()._setObject(id, newObj) if RESPONSE is not None: RESPONSE.redirect('manage_main')