#! /usr/bin/python """ $Id: pretty.py,v 1.42 2012/01/30 09:30:20 timbl Exp $ Printing of N3 and RDF formulae 20003-8-20 split offf from llyn.py This is or was http://www.w3.org/2000/10/swap/pretty.py """ import types import string import diag from diag import progress, verbosity, tracking from term import Literal, XMLLiteral, Symbol, Fragment, AnonymousNode, \ AnonymousVariable, FragmentNil, AnonymousUniversal, \ Term, CompoundTerm, List, EmptyList, NonEmptyList, N3Set from formula import Formula, StoredStatement from RDFSink import Logic_NS, RDFSink, forSomeSym, forAllSym from RDFSink import CONTEXT, PRED, SUBJ, OBJ, PARTS, ALL4, \ ANONYMOUS, SYMBOL, LITERAL, LITERAL_DT, LITERAL_LANG, XMLLITERAL from RDFSink import N3_nil, N3_first, N3_rest, OWL_NS, N3_Empty, N3_List, \ List_NS from RDFSink import RDF_NS_URI from RDFSink import RDF_type_URI cvsRevision = "$Revision: 1.42 $" # Magic resources we know about from RDFSink import RDF_type_URI, DAML_sameAs_URI STRING_NS_URI = "http://www.w3.org/2000/10/swap/string#" META_NS_URI = "http://www.w3.org/2000/10/swap/meta#" INTEGER_DATATYPE = "http://www.w3.org/2001/XMLSchema#integer" FLOAT_DATATYPE = "http://www.w3.org/2001/XMLSchema#double" prefixchars = "abcdefghijklmnopqustuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ" # Utility functions for the object-free abstract notation interface def auPair(x): """Encode as object-free form for unparser interface""" if isinstance(x, XMLLiteral): return (XMLLITERAL, x.dom) if isinstance(x, Literal): if x.datatype: return LITERAL_DT, (str(x), x.datatype.uriref()) # could be XMLLit if x.lang: return LITERAL_LANG, (x.string, x.lang) return (LITERAL, x.string) if isinstance(x, AnonymousUniversal): if x.uri: return (SYMBOL, x.uriref()) if isinstance(x, AnonymousNode): return (ANONYMOUS, x.uriref()) else: return (SYMBOL, x.uriref()) def auxPairs(t): return(auPair(t[CONTEXT]), auPair(t[PRED]), auPair(t[SUBJ]), auPair(t[OBJ]), ) class TooMuchRecursion(Exception): pass class Serializer: """A serializer to serialize the formula F into the given abstract syntax sink """ def __init__(self, F, sink, flags="", sorting=0): self.context = F # assert F.canonical is not None, "Formula printed must be canonical" self.store = F.store self.sink = sink self.defaultNamespace = None self.flags = flags self.sorting = sorting self._inContext ={} self._loopCheck = {} self._inLoop = {} self._tooDeep = {} self._occurringAs = [{}, {}, {}, {}] self._topology_returns = {} def selectDefaultPrefix(self, printFunction): """ Symbol whose fragments have the most occurrences. we suppress the RDF namespace itself because the XML syntax has problems with it being default as it is used for attributes. This also outputs the prefixes.""" if "d" in self.flags: self.defaultNamespace = None self.dumpPrefixes() return dummySink = self.sink.dummyClone() dummySerializer = Serializer(self.context, sink=dummySink, flags=self.flags+"d", sorting=self.sorting) printFunction(dummySerializer) best = 0 mp = None counts = dummySink.namespaceCounts() for r, count in counts.items(): if verbosity() > 25: progress(" Count is %3i for %s" %(count, r)) if (r != RDF_NS_URI and count > 0 and (count > best or (count == best and mp > r))) : # Must be repeatable for retests best = count mp = r if verbosity() > 20: progress("# Most popular Namespace is %s with %i" % ( mp, best)) self.defaultNamespace = mp # Make up prefixes for things which don't have them: for r, count in counts.items(): if count > 1 and r != mp: if self.store.prefixes.get(r, None) is None: p = r if p[-1] in "/#": p = p[:-1] slash = p.rfind("/") if slash >= 0: p = p[slash+1:] i = 0 while i < len(p): if p[i] in prefixchars: i = i + 1 else: break p = p[:i] if len(p) <6 and self.store.namespaces.get(p, None)is None: # and p[:3]!='xml': pref = p else: p = p[:5] for l in (3, 2, 4, 1, 5): if self.store.namespaces.get(p[:l], None) is None: # and p[:l][:3]!='xml': pref = p[:l] break else: n = 2 while 1: pref = p[:3]+`n` if self.store.namespaces.get(pref, None) is None: break n = n + 1 self.store.bind(pref, r) if verbosity() > 50: progress("Generated @prefix %s: <%s>." % (pref, r)) if self.defaultNamespace is not None: self.sink.setDefaultNamespace(self.defaultNamespace) # progress("&&&& Counts: ", counts) prefixes = self.store.namespaces.keys() # bind in same way as input did FYI prefixes.sort() # For repeatability of test results for pfx in prefixes: r = self.store.namespaces[pfx] try: count = counts[r] if count > 0: self.sink.bind(pfx, r) except KeyError: pass return def dumpPrefixes(self): if self.defaultNamespace is not None: sink.setDefaultNamespace(self.defaultNamespace) prefixes = self.store.namespaces.keys() # bind in same way as input did FYI prefixes.sort() for pfx in prefixes: uri = self.store.namespaces[pfx] self.sink.bind(pfx, uri) def _listsWithinLists(self, L, lists): if L not in lists: lists.append(L) for i in L: if isinstance(i, NonEmptyList) or isinstance(i, N3Set): self._listsWithinLists(i, lists) def dumpLists(self): "Dump lists out as first and rest. Not used in pretty." listList = {} context = self.context sink = self.sink lists = [] for s in context.statements: for x in s.predicate(), s.subject(), s.object(): if isinstance(x, NonEmptyList) or isinstance(x, N3Set): self._listsWithinLists(x, lists) for l in lists: if isinstance(l, N3Set): a = context.newBlankNode() ll = [mm for mm in l] #I hate sorting things ll.sort(Term.compareAnyTerm) list = self.store.newList(ll) self._outputStatement(sink, (context, self.store.forSome, context, a)) l._list = list l._node = a else: list = l while not isinstance(list, EmptyList): if list not in listList: self._outputStatement(sink, (context, self.store.forSome, context, list)) listList[list] = 1 list = list.rest listList = {} for l in lists: list = l if isinstance(l, N3Set): list = l._list self._outputStatement(sink, (context, self.store.owlOneOf, l._node, list)) while (not isinstance(list, EmptyList)) and list not in listList: if isinstance(list.first, N3Set): self._outputStatement(sink, (context, self.store.first, list, list.first._node)) else: self._outputStatement(sink, (context, self.store.first, list, list.first)) self._outputStatement(sink, (context, self.store.rest, list, list.rest)) listList[list] = 1 list = list.rest def dumpChronological(self): "Fast as possible. Only dumps data. No formulae or universals." context = self.context sink = self.sink sink.startDoc() self.dumpPrefixes() self.dumpVariables(context, sink, sorting=0, dataOnly=1) uu = context.universals() self.dumpLists() def fixSet(x): try: return x._node except AttributeError: return x for s in context.statements: for p in SUBJ, PRED, OBJ: x = s[p] if isinstance(x, Formula) or x in uu: break else: self._outputStatement(sink, [fixSet(x) for x in s.quad]) sink.endDoc() def _outputStatement(self, sink, quad, aWorks = 1): if isinstance(quad[1], Literal): raise ValueError("Cannot have a literal as a predicate. This makes no sense, %s" % `quad[1]`) if isinstance(quad[1], Formula): raise ValueError("Cannot have a formula as a predicate. This makes no sense") sink.makeStatement(auxPairs(quad), aIsPossible=aWorks) def dumpVariables(self, context, sink, sorting=1, pretty=0, dataOnly=0): """Dump the forAlls and the forSomes at the top of a formula""" uv = list(context.universals()) ev = list(context.existentials()) if sorting: uv.sort(Term.compareAnyTerm) ev.sort(Term.compareAnyTerm) if not dataOnly: for v in uv: self._outputStatement(sink, (context, self.store.forAll, context, v)) for v in ev: aWorks = 0 if pretty: _anon, _incoming = self._topology(v, context) if not _anon: self._outputStatement(sink, (context, self.store.forSome, context, v), \ canItbeABNode(context, v)) else: # not pretty, no formulae, can always use _:a form self._outputStatement(sink, (context, self.store.forSome, context, v), 1) def dumpBySubject(self, sorting=1): """ Dump one formula only by order of subject except forSome's first for n3=a mode""" context = self.context uu = context.universals().copy() sink = self.sink self._scan(context) sink.startDoc() self.selectDefaultPrefix(Serializer.dumpBySubject) self.dumpVariables(context, sink, sorting) self.dumpLists() ss = context.statements[:] ss.sort() def fixSet(x): try: return x._node except AttributeError: return x for s in ss: for p in SUBJ, PRED, OBJ: x = s[p] if isinstance(x, Formula) or x in uu: break else: self._outputStatement(sink, [fixSet(x) for x in s.quad]) if 0: # Doesn't work as ther ei snow no list of bnodes rs = self.store.resources.values() if sorting: rs.sort(Term.compareAnyTerm) for r in rs : # First the bare resource statements = context.statementsMatching(subj=r) if sorting: statements.sort(StoredStatement.comparePredObj) for s in statements : self._outputStatement(sink, s.quad) if not isinstance(r, Literal): fs = r.fragments.values() if sorting: fs.sort for f in fs : # then anything in its namespace statements = context.statementsMatching(subj=f) if sorting: statements.sort(StoredStatement.comparePredObj) for s in statements: self._outputStatement(sink, s.quad) sink.endDoc() # # Pretty printing # # An intersting alternative is to use the reverse syntax to the max, which # makes the DLG an undirected labelled graph. s and o above merge. The only think which # then prevents us from dumping the graph without new bnode ids is the presence of cycles. # # Blank nodes can be represented using the implicit syntax [] or rdf/xml equivalent # instead of a dummy identifier iff # - they are blank nodes, ie are existentials whose id has been generated, and # - the node only occurs directly in one formula in the whole thing to be printed, and # - the node occurs at most once as a object or list element within that formula # We used to work this out on the fly, but it is faster to build an index of the # whole formula to be printed first. # # Note when we scan a list we do it in the context of the formula in which we found # it. It may occcur in many formulae. def _scanObj(self, context, x): "Does this appear in just one context, and if so counts how many times as object" z = self._inContext.get(x, None) if z == "many": return # forget it if z is None: self._inContext[x] = context elif z is not context: self._inContext[x] = "many" return if isinstance(x, NonEmptyList) or isinstance(x, N3Set): for y in x: self._scanObj(context, y) if isinstance(x, AnonymousVariable) or (isinstance(x, Fragment) and x.generated()): y = self._occurringAs[OBJ].get(x, 0) + 1 self._occurringAs[OBJ][x] = y if verbosity() > 98: progress( "scan: %s, a %s, now has %i occurrences as %s" %(x, x.__class__,y,"CPSOqqqqq"[y])) # else: # if x is None: raise RuntimeError("Weird - None in a statement?") # progress("&&&&&&&&& %s has class %s " %(`z`, `z.__class__`)) def _scan(self, x, context=None): # progress("Scanning ", x, " &&&&&&&&") # assert self.context._redirections.get(x, None) is None, "Should not be redirected: "+`x` if verbosity() > 98: progress("scanning %s a %s in context %s" %(`x`, `x.__class__`,`context`), x.generated(), self._inContext.get(x, "--")) if isinstance(x, NonEmptyList) or isinstance(x, N3Set): for y in x: self._scanObj(context, y) if isinstance(x, Formula): for s in x.statements: for p in PRED, SUBJ, OBJ: y = s[p] if (isinstance(y, AnonymousVariable) or (isinstance(y, Fragment) and y.generated())): z = self._inContext.get(y, None) if z == "many": continue # forget it if z is None: self._inContext[y] = x elif z is not x: self._inContext[y] = "many" continue z = self._occurringAs[p].get(y, 0) self._occurringAs[p][y] = z + 1 # progress("&&&&&&&&& %s now occurs %i times as %s" %(`y`, z+1, "CPSO"[p])) # else: # progress("&&&&&&&&& yyyy %s has class %s " %(`y`, `y.__class__`)) if x is not y: self._scan(y, x) self._breakloops(x) def _breakloops(self, context): _done = {} _todo = list(self._occurringAs[SUBJ]) _todo.sort(Term.compareAnyTerm) for x in _todo: if x in _done: continue if not (isinstance(x, AnonymousVariable) and not ((isinstance(x, Fragment) and x.generated()))): _done[x] = True continue _done[x] = x a = True y = x while a is not None: a = context._index.get((None, None, y), None) if a is None or len(a) == 0: a = None if a is not None: #print a y = a[0][SUBJ] beenHere = _done.get(y, None) if beenHere is x: self._inLoop[y] = 1 a = None elif beenHere is not None: a = None elif not (isinstance(y, AnonymousVariable) and not ((isinstance(y, Fragment) and y.generated()))): _done[y] = True a = None else: _done[y] = x def _topology(self, x, context): """ Can be output as an anonymous node in N3. Also counts incoming links. Output tuple parts: 1. True iff can be represented as anonymous node in N3, [] or {} 2. Number of incoming links: >0 means occurs as object or pred, 0 means as only as subject. 1 means just occurs once >1 means occurs too many times to be anon Returns number of incoming links (1 or 2) including forSome link or zero if self can NOT be represented as an anonymous node. Paired with this is whether this is a subexpression. """ # This function takes way too long. My attempts to speed it up using a try / except # loop were clearly misguided, because this function does very little as is. # why does this take .08 seconds per function call to do next to nothing? ## try: ## return self._topology_returns[x] ## except KeyError: ## pass # progress("&&&&&&&&& ", `self`, self._occurringAs) _isExistential = x in context.existentials() # _isExistential = context.existentialDict.get(x,0) # return (0, 2) _loop = context.any(subj=x, obj=x) # does'nt count as incomming _asPred = self._occurringAs[PRED].get(x, 0) _asObj = self._occurringAs[OBJ].get(x, 0) _inLoop = self._inLoop.get(x, 0) _tooDeep = self._tooDeep.get(x, 0) if isinstance(x, Literal): _anon = 0 # Never anonymous, always just use the string elif isinstance(x, Formula): _anon = 2 # always anonymous, represented as itself elif isinstance(x, List): if isinstance(x, EmptyList): _anon = 0 # Never anonymous, always just use the string else: _anon = 2 # always anonymous, represented as itself _isExistential = 1 elif isinstance(x, N3Set): _anon = 2 _isExistential = 1 elif not x.generated(): _anon = 0 # Got a name, not anonymous else: # bnode ctx = self._inContext.get(x, "weird") _anon = ctx == "weird" or (ctx is context and _asObj < 2 and _asPred == 0 and _inLoop == 0 and _tooDeep == 0 and (not _loop) and _isExistential) if verbosity() > 97: progress( "Topology %s in %s is: ctx=%s,anon=%i obj=%i, pred=%i loop=%s ex=%i "%( `x`, `context`, `ctx`, _anon, _asObj, _asPred, _loop, _isExistential)) return ( _anon, _asObj+_asPred ) if verbosity() > 98: progress( "Topology %s in %s is: anon=%i obj=%i, pred=%i loop=%s ex=%i "%( `x`, `context`, _anon, _asObj, _asPred, _loop, _isExistential)) ## self._topology_returns[x] = ( _anon, _asObj+_asPred ) return ( _anon, _asObj+_asPred ) def dumpNested(self): """ Iterates over all URIs ever seen looking for statements """ context = self.context # assert context.canonical is not None self._scan(context) self.sink.startDoc() self.selectDefaultPrefix(Serializer.dumpNested) self.dumpFormulaContents(context, self.sink, sorting=1, equals=1) self.sink.endDoc() def tmDumpNested(self): """ """ context = self.context assert context.canonical is not None self._scan(context) self.tm.start() self._dumpFormula(context) self.tm.end() def _dumpNode(self, node): tm = self.tm _anon, _incoming = self._topology(node, context) if isinstance(node, List): tm.startList() [self._dumpNode(x) for x in node] tm.endList() elif isinstance(node, N3Set): pass elif isinstance(node, formula): tm.startFormula() self._dumpFormula(node) tm.endFormula() elif _anon: tm.startAnonymous() self._dumpSubject(context, node) tm.endAnonymous() elif isinstance(node, Literal): tm.addLiteral(node) elif isinstance(node, Symbol): tm.addSymbol(node) else: pass def _dumpFormula(self, node): pass def _dumpSubject(self, formula, node): pass def _dumpPredicate(self, formula, subject, node): pass def dumpFormulaContents(self, context, sink, sorting, equals=0): """ Iterates over statements in formula, bunching them up into a set for each subject. """ allStatements = context.statements[:] if equals: for x, y in context._redirections.items(): if not x.generated() and x not in context.variables(): allStatements.append(StoredStatement( (context, context.store.sameAs, x, y))) allStatements.sort() # context.statements.sort() # @@ necessary? self.dumpVariables(context, sink, sorting, pretty=1) # statements = context.statementsMatching(subj=context) # context is subject # if statements: # progress("@@ Statement with context as subj?!", statements,) # self._dumpSubject(context, context, sink, sorting, statements) currentSubject = None statements = [] for s in allStatements: con, pred, subj, obj = s.quad if subj is con: continue # Done them above if currentSubject is None: currentSubject = subj if subj != currentSubject: self._dumpSubject(currentSubject, context, sink, sorting, statements) statements = [] currentSubject = subj statements.append(s) if currentSubject is not None: self._dumpSubject(currentSubject, context, sink, sorting, statements) def _dumpSubject(self, subj, context, sink, sorting, statements=[]): """ Dump the infomation about one top level subject This outputs arcs leading away from a node, and where appropriate recursively descends the tree, by dumping the object nodes (and in the case of a compact list, the predicate (rest) node). It does NOTHING for anonymous nodes which don't occur explicitly as subjects. The list of statements must be sorted if sorting is true. """ _anon, _incoming = self._topology(subj, context) # Is anonymous? if _anon and _incoming == 1 and not isinstance(subj, Formula): return # Forget it - will be dealt with in recursion if isinstance(subj, List): li = subj else: li = None if isinstance(subj, N3Set): #I hate having to sort things se = [mm for mm in subj] se.sort(Term.compareAnyTerm) li = self.store.newList(se) else: se = None if isinstance(subj, Formula) and subj is not context: sink.startFormulaSubject(auPair(subj)) self.dumpFormulaContents(subj, sink, sorting) sink.endFormulaSubject(auPair(subj)) # Subject is now set up # continue to do arcs elif _anon and (_incoming == 0 or (li is not None and not isinstance(li, EmptyList))): # Will be root anonymous node - {} or [] or () if subj is context: pass else: # Could have alternative syntax here if sorting: statements.sort(StoredStatement.comparePredObj) # @@ Needed now Fs are canonical? if se is not None: a = self.context.newBlankNode() sink.startAnonymousNode(auPair(a)) self.dumpStatement(sink, (context, self.store.owlOneOf, a, li), sorting) for s in statements: # "[] color blue." might be nicer. @@@ Try it? m = s.quad[0:2] + (a, s.quad[3]) self.dumpStatement(sink, m, sorting) sink.endAnonymousNode() return elif li is not None and not isinstance(li, EmptyList): for s in statements: p = s.quad[PRED] if p is not self.store.first and p is not self.store.rest: if verbosity() > 90: progress("Is list, has values for", `p`) break # Something to print (later) else: if subj.generated(): return # Nothing. if "l" not in self.flags: sink.startListSubject(auPair(subj)) for ele in subj: self.dumpStatement(sink, (context, self.store.li, subj, ele), sorting) sink.endListSubject(auPair(subj)) for s in statements: p = s.quad[PRED] if p is not self.store.first and p is not self.store.rest: self.dumpStatement(sink, s.quad, sorting) # Dump the rest outside the () else: # List but explicitly using first and rest sink.startAnonymousNode(auPair(subj)) self.dumpStatement(sink, (context, self.store.first, subj, subj.first), sorting) self.dumpStatement(sink, (context, self.store.rest, subj, subj.rest), sorting) for s in statements: self.dumpStatement(sink, s.quad, sorting) sink.endAnonymousNode() return else: if verbosity() > 90: progress("%s Not list, has property values." % `subj`) sink.startAnonymousNode(auPair(subj)) for s in statements: # "[] color blue." might be nicer. @@@ Try it? try: self.dumpStatement(sink, s.quad, sorting) except TooMuchRecursion: pass sink.endAnonymousNode() return # arcs as subject done if sorting: statements.sort(StoredStatement.comparePredObj) for s in statements: self.dumpStatement(sink, s.quad, sorting) def dumpStatement(self, sink, triple, sorting): "Dump one statement, including structure within object" context, pre, sub, obj = triple if (sub is obj and not isinstance(sub, CompoundTerm)) \ or (isinstance(obj, EmptyList)) \ or isinstance(obj, Literal): self._outputStatement(sink, triple) # Do 1-loops simply return if isinstance(obj, Formula): sink.startFormulaObject(auxPairs(triple)) self.dumpFormulaContents(obj, sink, sorting) sink.endFormulaObject(auPair(pre), auPair(sub)) return if isinstance(obj, NonEmptyList): if verbosity()>99: progress("List found as object of dumpStatement " + `obj` + context.debugString()) collectionSyntaxOK = ("l" not in self.flags) # if "x" in self.flags: # Xml can't serialize literal in collection # for ele in obj: # if isinstance(ele, Literal): # collectionSyntaxOK = 0 # break if collectionSyntaxOK: sink.startListObject(auxPairs(triple)) for ele in obj: self.dumpStatement(sink, (context, self.store.li, obj, ele), # @@@ THIS IS WRONG IN XML !!! sorting) sink.endListObject(auPair(sub), auPair(pre)) else: sink.startAnonymous(auxPairs(triple)) self.dumpStatement(sink, (context, self.store.first, obj, obj.first), sorting) self.dumpStatement(sink, (context, self.store.rest, obj, obj.rest), sorting) sink.endAnonymous(auPair(sub), auPair(pre)) return if isinstance(obj, N3Set): a = self.context.newBlankNode() tempobj = [mm for mm in obj] #I hate sorting things - yosi tempobj.sort(Term.compareAnyTerm) tempList = self.store.newList(tempobj) sink.startAnonymous(auxPairs((triple[CONTEXT], triple[PRED], triple[SUBJ], a))) self.dumpStatement(sink, (context, self.store.owlOneOf, a, tempList), sorting) sink.endAnonymous(auPair(sub), auPair(pre)) return _anon, _incoming = self._topology(obj, context) if _anon and _incoming == 1: # Embedded anonymous node in N3 sink.startAnonymous(auxPairs(triple)) ss = context.statementsMatching(subj=obj) if sorting: ss.sort(StoredStatement.comparePredObj) for t in ss: self.dumpStatement(sink, t.quad, sorting) sink.endAnonymous(sub.asPair(), pre.asPair()) return self._outputStatement(sink, triple) BNodePossibles = None def canItbeABNode(formula, symbol): def returnFunc(): if BNodePossibles is not None: return symbol in BNodePossibles for quad in formula.statements: for s in PRED, SUBJ, OBJ: if isinstance(quad[s], Formula): if BNodePossibles is None: BNodePossible = quad[s].occurringIn( formula.existentials()) else: BNodePossible.update(quad[s].occurringIn( formula.existentials())) return symbol in BNodePossibles return returnFunc #ends