# -*- coding: utf-8 -*- # ------------------------------------------------------------------------------ # Appy is a framework for building applications in the Python language. # Copyright (C) 2007-2011 Gaetan Delannay # # Distributed under the GNU General Public License. # # Thanks to Fabio Marcuzzi and Gauthier Bastien for management of strike and # underline. # ------------------------------------------------------------------------------ import xml.sax, time, random from appy.shared.xml_parser import XmlEnvironment, XmlParser from appy.pod.odf_parser import OdfEnvironment from appy.pod.styles_manager import Style from appy.pod import * # To which ODT tags do HTML tags correspond ? HTML_2_ODT = {'h1':'h', 'h2':'h', 'h3':'h', 'h4':'h', 'h5':'h', 'h6':'h', 'p':'p', 'div': 'p', 'b':'span', 'i':'span', 'strong':'span', 'strike':'span', 'u':'span', 'em': 'span', 'sub': 'span', 'sup': 'span', 'br': 'line-break'} DEFAULT_ODT_STYLES = {'b': 'podBold', 'strong':'podBold', 'i': 'podItalic', 'u': 'podUnderline', 'strike': 'podStrike', 'em': 'podItalic', 'sup': 'podSup', 'sub':'podSub', 'td': 'podCell', 'th': 'podHeaderCell'} INNER_TAGS = ('b', 'strong', 'i', 'u', 'em', 'sup', 'sub', 'span') TABLE_CELL_TAGS = ('td', 'th') OUTER_TAGS = TABLE_CELL_TAGS + ('li',) # The following elements can't be rendered inside paragraphs NOT_INSIDE_P = XHTML_HEADINGS + XHTML_LISTS + ('table',) NOT_INSIDE_P_OR_P = NOT_INSIDE_P + ('p', 'div') NOT_INSIDE_LIST = ('table',) IGNORABLE_TAGS = ('meta', 'title', 'style') # ------------------------------------------------------------------------------ class HtmlElement: '''Every time an HTML element is encountered during the SAX parsing, an instance of this class is pushed on the stack of currently parsed elements.''' elemTypes = {'p':'para', 'div':'para', 'li':'para', 'ol':'list', 'ul':'list'} def __init__(self, elem, attrs): self.elem = elem # Keep "class" attribute (useful for finding the corresponding ODT # style) in some cases. Normally, basic XmlElement class stores attrs, # but for a strange reason those attrs are back to None (probably for # performance reasons they become inaccessible after a while). self.classAttr = None if attrs.has_key('class'): self.classAttr = attrs['class'] self.tagsToReopen = [] # When the HTML element corresponding to self # is completely dumped, if there was a problem related to tags # inclusion, we may need to dump start tags corresponding to # tags that we had to close before dumping this element. This list # contains HtmlElement instances. self.tagsToClose = [] # Before dumping the closing tag corresponding # to self, we may need to close other tags (ie closing a paragraph # before closing a cell). This list contains HtmlElement instances. self.elemType = self.elem if self.elemTypes.has_key(self.elem): self.elemType = self.elemTypes[self.elem] # If a conflict occurs on this element, we will note it. self.isConflictual = False def setConflictual(self): '''Note p_self as conflictual.''' self.isConflictual = True return self def getOdfTag(self, env): '''Gets the raw ODF tag that corresponds to me.''' res = '' if HTML_2_ODT.has_key(self.elem): res += '%s:%s' % (env.textNs, HTML_2_ODT[self.elem]) elif self.elem == 'a': res += '%s:a' % env.textNs elif self.elem in XHTML_LISTS: res += '%s:list' % env.textNs elif self.elem == 'li': res += '%s:list-item' % env.textNs elif self.elem == 'table': res += '%s:table' % env.tableNs elif self.elem == 'thead': res += '%s:table-header-rows' % env.tableNs elif self.elem == 'tr': res += '%s:table-row' % env.tableNs elif self.elem in TABLE_CELL_TAGS: res += '%s:table-cell' % env.tableNs return res def getOdfTags(self, env): '''Gets the start and end tags corresponding to p_self.''' tag = self.getOdfTag(env) if not tag: return (None, None) return ('<%s>' % tag, '' % tag) def getConflictualElements(self, env): '''self was just parsed. In some cases, this element can't be dumped in the result because there are conflictual elements among previously parsed opening elements (p_env.currentElements). For example, if we just dumped a "p", we can't dump a table within the "p". Such constraints do not hold in XHTML code but hold in ODF code.''' if not env.currentElements: return () parentElem = env.currentElements[-1] # Check elements that can't be found within a paragraph if (parentElem.elemType == 'para') and \ (self.elem in NOT_INSIDE_P_OR_P): # Oups, li->p wrongly considered as a conflict. if (parentElem.elem == 'li') and (self.elem in ('p', 'div')): return () return (parentElem.setConflictual(),) # Check inner paragraphs if (parentElem.elem in INNER_TAGS) and (self.elemType == 'para'): res = [parentElem.setConflictual()] if len(env.currentElements) > 1: i = 2 visitParents = True while visitParents: try: nextParent = env.currentElements[-i] i += 1 res.insert(0, nextParent.setConflictual()) if nextParent.elemType == 'para': visitParents = False except IndexError: visitParents = False return res if parentElem.tagsToClose and \ (parentElem.tagsToClose[-1].elemType == 'para') and \ (self.elem in NOT_INSIDE_P): return (parentElem.tagsToClose[-1].setConflictual(),) # Check elements that can't be found within a list if (parentElem.elemType=='list') and (self.elem in NOT_INSIDE_LIST): return (parentElem.setConflictual(),) return () def addInnerParagraph(self, env): '''Dump an inner paragraph inside self (if not already done).''' if not self.tagsToClose: # We did not do it yet env.dumpString('<%s:p' % env.textNs) if self.elem == 'li': itemStyle = env.getCurrentElement(isList=True).elem # ul or ol # Which 'li'-related style must I use? if self.classAttr: odtStyle = env.parser.caller.findStyle( self.elem, classValue=self.classAttr) if odtStyle and (odtStyle.name == 'podItemKeepWithNext'): itemStyle += '_kwn' styleName = env.itemStyles[itemStyle] else: # Check if a style must be applied on 'p' tags odtStyle = env.parser.caller.findStyle('p') if odtStyle: styleName = odtStyle.name else: styleName = env.itemStyles[itemStyle] env.dumpString(' %s:style-name="%s"' % (env.textNs, styleName)) else: # Check if a style must be applied on 'p' tags odtStyle = env.parser.caller.findStyle('p') if odtStyle: env.dumpString(' %s:style-name="%s"' % (env.textNs, odtStyle.name)) env.dumpString('>') self.tagsToClose.append(HtmlElement('p', {})) def dump(self, start, env): '''Dumps the start or end (depending on p_start) tag of this HTML element. We must take care of potential innerTags.''' # Compute the tag in itself tag = '' prefix = '<' if not start: prefix += '/' # Compute tag attributes attrs = '' if start: if self.elemType == 'list': # I must specify the list style attrs += ' %s:style-name="%s"' % ( env.textNs, env.listStyles[self.elem]) if self.elem == 'ol': # I have interrupted a numbered list. I need to continue # the numbering. attrs += ' %s:continue-numbering="true"' % env.textNs else: attrs = env.getOdtAttributes(self) tag = prefix + self.getOdfTag(env) + attrs + '>' # Close/open subTags if any for subElem in self.tagsToClose: subTag = subElem.dump(start, env) if start: tag += subTag else: tag = subTag + tag return tag def __repr__(self): return '' % self.elem # ------------------------------------------------------------------------------ class HtmlTable: '''Represents an HTML table, and also a sub-buffer. When parsing elements corresponding to an HTML table (, , entirely in order to know how much columns are present in the HTML table before dumping the ODF table. So we use this class as a sub-buffer that will be constructed as we parse the HTML table; when encountering the end of the HTML table, we will dump the result of this sub-buffer into the parent buffer, which may be the global buffer or another table buffer.''' def __init__(self, env): elems = str(time.time()).split('.') self.name= 'AppyTable%s%s%d' % (elems[0],elems[1],random.randint(1,100)) self.styleNs = env.ns[OdfEnvironment.NS_STYLE] self.res = u'' # The sub-buffer. self.tempRes = u'' # The temporary sub-buffer, into which we will # dump all table sub-elements, until we encounter the end of the first # row. Then, we will know how much columns are defined in the table; # we will dump columns declarations into self.res and dump self.tempRes # into self.res. self.firstRowParsed = False # Was the first table row completely parsed? self.nbOfColumns = 0 # Are we currently within a table cell? Instead of a boolean, the field # stores an integer. The integer is > 1 if the cell spans more than one # column. self.inCell = 0 # The index, within the current row, of the current cell self.cellIndex = -1 # The size of the content of the currently parsed table cell self.cellContentSize = 0 # The following list stores, for every column, the size of the biggest # content of all its cells. self.columnContentSizes = [] def computeColumnStyles(self, renderer): '''Once the table has been completely parsed, self.columnContentSizes should be correctly filled. Based on this, we can deduce the width of every column and create the corresponding style declarations, in p_renderer.dynamicStyles.''' total = 65000.0 # A number representing the total width of the table # Ensure first that self.columnContentSizes is correct if (len(self.columnContentSizes) != self.nbOfColumns) or \ (None in self.columnContentSizes): # There was a problem while parsing the table. Set every column # with the same width. widths = [int(total/self.nbOfColumns)] * self.nbOfColumns else: widths = [] # Compute the sum of all column content sizes contentTotal = 0 for size in self.columnContentSizes: contentTotal += size contentTotal = float(contentTotal) for size in self.columnContentSizes: width = int((size/contentTotal) * total) widths.append(width) # Compute style declatation corresponding to every column. s = self.styleNs i = 0 for width in widths: i += 1 # Compute the width of this column, relative to "total". decl = '<%s:style %s:name="%s.%d" %s:family="table-column">' \ '<%s:table-column-properties %s:rel-column-width="%d*"' \ '/>' % (s, s, self.name, i, s, s, s, width, s) renderer.dynamicStyles.append(decl.encode('utf-8')) # ------------------------------------------------------------------------------ class XhtmlEnvironment(XmlEnvironment): itemStyles = {'ul': 'podBulletItem', 'ol': 'podNumberItem', 'ul_kwn': 'podBulletItemKeepWithNext', 'ol_kwn': 'podNumberItemKeepWithNext'} listStyles = {'ul': 'podBulletedList', 'ol': 'podNumberedList'} def __init__(self, renderer): XmlEnvironment.__init__(self) self.renderer = renderer self.ns = renderer.currentParser.env.namespaces self.res = u'' self.currentContent = u'' self.currentElements = [] # Stack of currently walked elements self.currentLists = [] # Stack of currently walked lists (ul or ol) self.currentTables = [] # Stack of currently walked tables self.textNs = self.ns[OdfEnvironment.NS_TEXT] self.linkNs = self.ns[OdfEnvironment.NS_XLINK] self.tableNs = self.ns[OdfEnvironment.NS_TABLE] # The following attr will be True when parsing parts of the XHTML that # must be ignored. self.ignore = False def getCurrentElement(self, isList=False): '''Gets the element that is on the top of self.currentElements or self.currentLists.''' res = None if isList: elements = self.currentLists # Stack of list elements only else: elements = self.currentElements # Stack of all elements (including # elements also pushed on other stacks, like lists and tables). if elements: res = elements[-1] return res def anElementIsMissing(self, previousElem, currentElem): res = False if previousElem and (previousElem.elem in OUTER_TAGS) and \ ((not currentElem) or (currentElem.elem in INNER_TAGS)): res = True return res def dumpCurrentContent(self): '''Dumps content that was temporarily stored in self.currentContent into the result.''' contentSize = 0 if self.currentContent.strip(): # Manage missing elements currentElem = self.getCurrentElement() if self.anElementIsMissing(currentElem, None): currentElem.addInnerParagraph(self) # Dump and reinitialize the current content content = self.currentContent.strip('\n\t') contentSize = len(content) for c in content: # We remove leading and trailing carriage returns, but not # whitespace because whitespace may be part of the text to dump. if XML_SPECIAL_CHARS.has_key(c): self.dumpString(XML_SPECIAL_CHARS[c]) else: self.dumpString(c) self.currentContent = u'' # If we are within a table cell, update the total size of cell content. if self.currentTables and self.currentTables[-1].inCell: for table in self.currentTables: table.cellContentSize += contentSize def getOdtAttributes(self, htmlElem, htmlAttrs={}): '''Gets the ODT attributes to dump for p_currentElem. p_htmlAttrs are the parsed attributes from the XHTML p_currentElem.''' odtStyle = self.parser.caller.findStyle(htmlElem.elem, htmlAttrs) styleName = None if odtStyle: styleName = odtStyle.name elif DEFAULT_ODT_STYLES.has_key(htmlElem.elem): styleName = DEFAULT_ODT_STYLES[htmlElem.elem] res = '' if styleName: res += ' %s:style-name="%s"' % (self.textNs, styleName) if (htmlElem.elem in XHTML_HEADINGS) and \ (odtStyle.outlineLevel != None): res += ' %s:outline-level="%d"' % (self.textNs, \ odtStyle.outlineLevel) return res def dumpStyledElement(self, htmlElem, odfTag, attrs): '''Dumps an element that potentially has associated style information.''' self.dumpString('<' + odfTag) self.dumpString(self.getOdtAttributes(htmlElem, attrs)) self.dumpString('>') def getTags(self, elems, start=True): '''This method returns a series of start or end tags (depending on p_start) that correspond to HtmlElement instances in p_elems.''' res = '' for elem in elems: tag = elem.dump(start, self) if start: res += tag else: res = tag + res return res def closeConflictualElements(self, conflictElems): '''This method dumps end tags for p_conflictElems, excepted if those tags would be empty. In this latter case, tags are purely removed from the result.''' startTags = self.getTags(conflictElems, start=True) if self.res.endswith(startTags): # In this case I would dump an empty (series of) tag(s). Instead, I # will remove those tags. self.res = self.res[:-len(startTags)] else: self.dumpString(self.getTags(conflictElems, start=False)) def dumpString(self, s): '''Dumps arbitrary content p_s. If the table stack is not empty, we must dump p_s into the buffer corresponding to the last parsed table. Else, we must dump p_s into the global buffer (self.res).''' if self.currentTables: currentTable = self.currentTables[-1] if (not currentTable.res) or currentTable.firstRowParsed: currentTable.res += s else: currentTable.tempRes += s else: self.res += s def getTagsToReopen(self, conflictElems): '''Normally, tags to reopen are equal to p_conflictElems. But we have a special case. Indeed, if a conflict elem has itself tagsToClose, the last tag to close may not be needed anymore on the tag to reopen, so we remove it.''' conflictElems[-1].tagsToClose = [] return conflictElems def onElementStart(self, elem, attrs): previousElem = self.getCurrentElement() self.dumpCurrentContent() currentElem = HtmlElement(elem, attrs) # Manage conflictual elements conflictElems = currentElem.getConflictualElements(self) if conflictElems: # We must close the conflictual elements, and once the currentElem # will be dumped, we will re-open the conflictual elements. self.closeConflictualElements(conflictElems) currentElem.tagsToReopen = self.getTagsToReopen(conflictElems) # Manage missing elements if self.anElementIsMissing(previousElem, currentElem): previousElem.addInnerParagraph(self) # Add the current element on the stack of walked elements self.currentElements.append(currentElem) if elem in XHTML_LISTS: # Update stack of current lists self.currentLists.append(currentElem) elif elem == 'table': # Update stack of current tables if not self.currentTables: # We are within a table. If no local style mapping is defined # for paragraphs, add a specific style mapping for a better # rendering of cells' content. caller = self.parser.caller map = caller.localStylesMapping if 'p' not in map: map['p'] = Style('Appy_Table_Content', 'paragraph') self.currentTables.append(HtmlTable(self)) elif elem in TABLE_CELL_TAGS: # Determine colspan colspan = 1 if attrs.has_key('colspan'): colspan = int(attrs['colspan']) table = self.currentTables[-1] table.inCell = colspan table.cellIndex += colspan # If we are in the first row of a table, update columns count if not table.firstRowParsed: table.nbOfColumns += colspan return currentElem def onElementEnd(self, elem): res = None self.dumpCurrentContent() currentElem = self.currentElements.pop() if elem in XHTML_LISTS: self.currentLists.pop() elif elem == 'table': table = self.currentTables.pop() # Computes the column styles required by the table table.computeColumnStyles(self.parser.caller.renderer) # Dumps the content of the last parsed table into the parent buffer self.dumpString(table.res) # Remove cell-paragraph from local styles mapping if it was added. map = self.parser.caller.localStylesMapping if not self.currentTables and ('p' in map): mapValue = map['p'] if isinstance(mapValue, Style) and \ (mapValue.name == 'Appy_Table_Content'): del map['p'] elif elem == 'tr': table = self.currentTables[-1] table.cellIndex = -1 if not table.firstRowParsed: table.firstRowParsed = True # First row is parsed. I know the number of columns in the # table: I can dump the columns declarations. for i in range(1, table.nbOfColumns + 1): table.res+= '<%s:table-column %s:style-name="%s.%d"/>' % \ (self.tableNs, self.tableNs, table.name, i) table.res += table.tempRes table.tempRes = u'' elif elem in TABLE_CELL_TAGS: # Update attr "columnContentSizes" of the currently parsed table, # excepted if the cell spans several columns. table = self.currentTables[-1] if table.inCell == 1: sizes = table.columnContentSizes # Insert None values if the list is too small while (len(sizes)-1) < table.cellIndex: sizes.append(None) highest = max(sizes[table.cellIndex], table.cellContentSize, 5) # Put a maximum highest = min(highest, 100) sizes[table.cellIndex] = highest table.inCell = 0 table.cellContentSize = 0 if currentElem.tagsToClose: self.closeConflictualElements(currentElem.tagsToClose) if currentElem.tagsToReopen: res = currentElem.tagsToReopen return currentElem, res # ------------------------------------------------------------------------------ class XhtmlParser(XmlParser): def lowerizeInput(self, elem, attrs=None): '''Because (X)HTML is case insensitive, we may receive input p_elem and p_attrs in lower-, upper- or mixed-case. So here we produce lowercase versions that will be used throughout our parser.''' resElem = elem.lower() resAttrs = attrs if attrs: resAttrs = {} for attrName in attrs.keys(): resAttrs[attrName.lower()] = attrs[attrName] if attrs == None: return resElem else: return resElem, resAttrs def startElement(self, elem, attrs): elem, attrs = self.lowerizeInput(elem, attrs) e = XmlParser.startElement(self, elem, attrs) currentElem = e.onElementStart(elem, attrs) odfTag = currentElem.getOdfTag(e) if HTML_2_ODT.has_key(elem): e.dumpStyledElement(currentElem, odfTag, attrs) elif elem == 'a': e.dumpString('<%s %s:type="simple"' % (odfTag, e.linkNs)) if attrs.has_key('href'): e.dumpString(' %s:href="%s"' % (e.linkNs, attrs['href'])) e.dumpString('>') elif elem in XHTML_LISTS: prologue = '' if len(e.currentLists) >= 2: # It is a list into another list. In this case the inner list # must be surrounded by a list-item element. prologue = '<%s:list-item>' % e.textNs numbering = '' if elem == 'ol': numbering = ' %s:continue-numbering="false"' % e.textNs e.dumpString('%s<%s %s:style-name="%s"%s>' % ( prologue, odfTag, e.textNs, e.listStyles[elem], numbering)) elif elem in ('li', 'thead', 'tr'): e.dumpString('<%s>' % odfTag) elif elem == 'table': # Here we must call "dumpString" only once table = e.currentTables[-1] e.dumpString('<%s %s:name="%s" %s:style-name="podTable">' % \ (odfTag, e.tableNs, table.name, e.tableNs)) elif elem in TABLE_CELL_TAGS: e.dumpString('<%s %s:style-name="%s"' % \ (odfTag, e.tableNs, DEFAULT_ODT_STYLES[elem])) if attrs.has_key('colspan'): e.dumpString(' %s:number-columns-spanned="%s"' % \ (e.tableNs, attrs['colspan'])) e.dumpString('>') elif elem == 'img': style = None if attrs.has_key('style'): style = attrs['style'] imgCode = e.renderer.importDocument(at=attrs['src'], wrapInPara=False, style=style) e.dumpString(imgCode) elif elem in IGNORABLE_TAGS: e.ignore = True def endElement(self, elem): elem = self.lowerizeInput(elem) e = XmlParser.endElement(self, elem) currentElem, elemsToReopen = e.onElementEnd(elem) # Determine the tag to dump startTag, endTag = currentElem.getOdfTags(e) if currentElem.isConflictual: # Compute the start tag, with potential styles applied startTag = e.getTags((currentElem,), start=True) if currentElem.isConflictual and e.res.endswith(startTag): # We will not dump it, it would constitute a silly empty tag. e.res = e.res[:-len(startTag)] else: # Dump the end tag. But dump some additional stuff if required. if elem in XHTML_LISTS: if len(e.currentLists) >= 1: # We were in an inner list. So we must close the list-item # tag that surrounds it. endTag = '%s' % (endTag, e.textNs) if endTag: e.dumpString(endTag) if elem in IGNORABLE_TAGS: e.ignore = False if elemsToReopen: e.dumpString(e.getTags(elemsToReopen, start=True)) def characters(self, content): e = XmlParser.characters(self, content) if not e.ignore: e.currentContent += content # ------------------------------------------------------------------------------- class Xhtml2OdtConverter: '''Converts a chunk of XHTML into a chunk of ODT.''' def __init__(self, xhtmlString, encoding, stylesManager, localStylesMapping, renderer): self.renderer = renderer self.xhtmlString = xhtmlString self.encoding = encoding # Todo: manage encoding that is not utf-8 self.stylesManager = stylesManager self.localStylesMapping = localStylesMapping self.odtChunk = None self.xhtmlParser = XhtmlParser(XhtmlEnvironment(renderer), self) def run(self): self.xhtmlParser.parse(self.xhtmlString) return self.xhtmlParser.env.res def findStyle(self, elem, attrs=None, classValue=None): return self.stylesManager.findStyle(elem, attrs, classValue, self.localStylesMapping) # ------------------------------------------------------------------------------
, etc), we can't dump corresponding ODF elements directly into the global result buffer (XhtmlEnvironment.res). Indeed, when dumping an ODF table, we must dump columns declarations at the beginning of the table. So before dumping rows and cells, we must know how much columns will be present in the table. It means that we must first parse the first