def w_map_to_html(the_map):
"""This callback receives a RDF property (WBEM or WMI) and a map
which represents the CIM links associated to the current object. """
logging.debug("w_map_to_html len=%d", len(the_map))
for url_subj in the_map:
unquoted_subj = lib_util.urllib_unquote(url_subj)
subj_text, subj_entity_graph_class, subj_entity_id = lib_naming.ParseEntityUri(
unquoted_subj)
yield "<tr>"
yield "<td valign='top'><a href='%s'>%s</a></td>" % (str(url_subj),
subj_text)
yield "<td>"
yield "<table border=0>"
for the_prop, url_obj in the_map[url_subj]:
yield "<tr>"
prop_nam = lib_exports.PropToShortPropNam(the_prop)
yield "<td><i>%s</i></td>" % prop_nam
if lib_kbase.IsLiteral(url_obj):
yield "<td>%s</td>" % str(url_obj)
else:
unquoted_obj = lib_util.urllib_unquote(url_obj)
obj_text, obj_entity_graph_class, obj_entity_id = lib_naming.ParseEntityUri(
unquoted_obj)
yield "<td><a href='%s'>%s</a></td>" % (str(url_obj),
obj_text)
yield "</tr>"
yield "</table>"
yield "</td>"
yield "</tr>"
def RdfNodeToDotLabelExtended(obj, prop):
objNam = RdfNodeToDotLabel(obj)
try:
dictOfProps = dictCollapsedObjectLabelsToSubjectLabels[objNam]
except KeyError:
# sys.stderr.write("RdfNodeToDotLabelExtended propNam=%s objNam=%s\n"%(propNam,objNam) )
return objNam
# Let's hope there is only one collapsed property for this node. Otherwise, it means
# that this node would be displayed in two different tables. It happened...
if not prop is None:
propNam = lib_exports.PropToShortPropNam(prop)
try:
# Maybe this property is not collapsed.
subjNam = dictOfProps[propNam]
except KeyError:
prop = None
# Maybe the property is not known, if the node is the subject.
# Or the property is not collapsed.
if prop is None:
# In Python3, keys() is an iterable. No need to create a list.
#propNam = list(dictOfProps.keys())[0]
#propNam = dictOfProps.keys()[0]
for propNam in dictOfProps.keys():
break
# First property available.
subjNam = dictOfProps[propNam]
newObjNam = CollapsedLabel(propNam, subjNam) + ":" + objNam
return newObjNam
def WMapToHtml(theMap, propData):
sys.stderr.write("WMapToHtml len=%d\n" % len(theMap))
for urlSubj in theMap:
(subjText, subjEntityGraphClass,
subjEntityId) = lib_naming.ParseEntityUri(
lib_util.urllib_unquote(urlSubj))
WrtAsUtf("<tr>")
WrtAsUtf("<td valign='top'><a href='%s'>%s</a></td>" %
(str(urlSubj), subjText))
WrtAsUtf("<td>")
WrtAsUtf("<table>")
for theProp, urlObj in theMap[urlSubj]:
WrtAsUtf("<tr>")
propNam = lib_exports.PropToShortPropNam(theProp)
WrtAsUtf("<td><i>%s</i></td>" % propNam)
if lib_kbase.IsLiteral(urlObj):
WrtAsUtf("<td>%s</td>" % (str(urlObj)))
else:
(objText, objEntityGraphClass,
objEntityId) = lib_naming.ParseEntityUri(
lib_util.urllib_unquote(urlObj))
WrtAsUtf("<td><a href='%s'>%s</a></td>" %
(str(urlObj), objText))
WrtAsUtf("</tr>")
WrtAsUtf("</table>")
WrtAsUtf("</td>")
WrtAsUtf("</tr>")
def WMapToHtml(theMap):
DEBUG("WMapToHtml len=%d", len(theMap))
for urlSubj in theMap:
(subjText, subjEntityGraphClass,
subjEntityId) = lib_naming.ParseEntityUri(
lib_util.urllib_unquote(urlSubj))
yield ("<tr>")
yield ("<td valign='top'><a href='%s'>%s</a></td>" %
(str(urlSubj), subjText))
yield ("<td>")
yield ("<table border=0>")
for theProp, urlObj in theMap[urlSubj]:
yield ("<tr>")
propNam = lib_exports.PropToShortPropNam(theProp)
yield ("<td><i>%s</i></td>" % propNam)
if lib_kbase.IsLiteral(urlObj):
yield ("<td>%s</td>" % (str(urlObj)))
else:
(objText, objEntityGraphClass,
objEntityId) = lib_naming.ParseEntityUri(
lib_util.urllib_unquote(urlObj))
yield ("<td><a href='%s'>%s</a></td>" %
(str(urlObj), objText))
yield ("</tr>")
yield ("</table>")
yield ("</td>")
yield ("</tr>")
def rdf_node_to_dot_label_extended(obj, prop):
obj_nam = _rdf_node_to_dot_label(obj)
try:
dict_of_props = dict_collapsed_object_labels_to_subject_labels[obj_nam]
except KeyError:
return obj_nam
# Let's hope there is only one collapsed property for this node. Otherwise, it means
# that this node would be displayed in two different tables. It happened...
if not prop is None:
prop_nam = lib_exports.PropToShortPropNam(prop)
try:
# Maybe this property is not collapsed.
subj_nam = dict_of_props[prop_nam]
except KeyError:
prop = None
# Maybe the property is not known, if the node is the subject.
# Or the property is not collapsed.
if prop is None:
# In Python3, keys() is an iterable. No need to create a list.
for sub_prop_nam in dict_of_props.keys():
break
# First property available.
subj_nam = dict_of_props[sub_prop_nam]
new_obj_nam = collapsed_label(sub_prop_nam, subj_nam) + ":" + obj_nam
return new_obj_nam
def output_rdf_graph_as_json_d3(page_title, error_msg, parameters, grph):
"""
Transforms a RDF graph into a JSON document.
This returns a graph made of Json objects which are suitable for visualisation in the Javascript
interface to Survol, which is based on D3.
"""
# Must be reset to zero between several executions, when run by WSGI.
global _node_json_number
_node_json_number = 0
# It contains a cache because the same nodes may appear several times.
def node_to_json_obj(the_nod):
try:
return node_to_json_obj.dictNod2Json[the_nod]
except KeyError:
json_obj = NodeJson(the_nod)
node_to_json_obj.dictNod2Json[the_nod] = json_obj
return json_obj
node_to_json_obj.dictNod2Json = dict()
links = []
for subj, pred, obj in grph:
# This applies only to entity.py : In rendering based on Json, scripts are not displayed as nodes,
# but in hierarchical menus. The node must not appear at all.
# TODO: Should probably also eliminate pc.property_rdf_data_nolist2 etc ... See lib_client.
if pred == pc.property_script:
logging.debug("continue subj=%s obj=%s", subj, obj)
continue
# Normal data scripts are not accepted. This should apply only to file_directory.py and file_to_mime.py
if not _script_for_json(subj):
continue
if not _script_for_json(obj):
continue
subj_obj = node_to_json_obj(subj)
subj_id = subj_obj.m_survol_url
prop_nam = lib_exports.PropToShortPropNam(pred)
# TODO: BUG: If several nodes for the same properties, only the last one is kept.
if lib_kbase.IsLink(obj):
obj_obj = node_to_json_obj(obj)
obj_id = obj_obj.m_survol_url
links.extend([{
'source': subj_id,
'target': obj_id,
'survol_link_prop': prop_nam
}])
# TODO: Add the name corresponding to the URL, in m_info_dict so that some elements
# of the tooltip would be clickable. On the other hand, one just need to merge
# the nodes relative to the object, by right-clicking.
elif lib_kbase.IsLiteral(obj):
if pred == pc.property_information:
try:
subj_obj.m_info_list.append(str(obj.value))
except UnicodeEncodeError:
# 'ascii' codec can't encode character u'\xf3' in position 17: ordinal not in range(128)
# https://stackoverflow.com/questions/9942594/unicodeencodeerror-ascii-codec-cant-encode-character-u-xa0-in-position-20
subj_obj.m_info_list.append(obj.value.encode('utf-8'))
else:
if isinstance(obj.value, six.integer_types) or isinstance(
obj.value, six.string_types):
subj_obj.m_info_dict[prop_nam] = obj.value
else:
# If the value cannot be serializable to JSON.
subj_obj.m_info_dict[prop_nam] = type(obj.value).__name__
else:
raise Exception(__file__ + " Cannot happen here")
# Now, this creates the nodes sent as json objects.
num_nodes = len(node_to_json_obj.dictNod2Json)
nodes = [None] * num_nodes
for nod in node_to_json_obj.dictNod2Json:
nod_obj = node_to_json_obj.dictNod2Json[nod]
nod_titl = nod_obj.m_label
nod_id = nod_obj.m_index
# The URL must not contain any HTML entities when in a XML or SVG document,
# and therefore must be escaped. Therefore they have to be unescaped when transmitted in JSON.
# This is especially needed for RabbitMQ because the parameter defining its connection name
# has the form: "Url=LOCALHOST:12345,Connection=127.0.0.1:51748 -> 127.0.0.1:5672"
# HTTP_MIME_URL
the_survol_nam = lib_util.survol_unescape(
nod_titl) # MUST UNESCAPE HTML ENTITIES !
# TODO: Use the same object for lookup and Json.
nodes[nod_id] = {
'id': nod_obj.m_survol_url, # Required by D3
'name': the_survol_nam,
# Theoretically, this URL should be HTML unescaped then CGI escaped.
'survol_url': nod_obj.m_survol_url, # Duplicate of 'id'
'survol_universal_alias': nod_obj.m_survol_universal_alias,
'survol_fill': nod_obj.m_color,
'entity_class': nod_obj.
m_class, # TODO: Maybe not needed because also in the URL ?
'survol_info_list': nod_obj.m_info_list,
'survol_info_dict': nod_obj.m_info_dict
}
# This is the graph displayed by D3.
graph = {"page_title": page_title, "nodes": nodes, "links": links}
_write_json_header(json.dumps(graph, indent=2))
def Rdf2Dot(grph, logfil, stream, CollapsedProperties):
fieldsSet = collections.defaultdict(list)
# This maps RDFLIB nodes to DOT label names.
dictRdf2Dot = {}
# This returns the DOT label of a RDFLIB, and creates a new one if necessary.
def RdfNodeToDotLabel(x):
try:
return dictRdf2Dot[x]
except KeyError:
nodelabel = "nd_%d" % len(dictRdf2Dot)
dictRdf2Dot[x] = nodelabel
return nodelabel
# The QName is an abbreviation of URI reference with the namespace function for XML.
# Edge label.
# Transforms "http://primhillcomputers.com/ontologies/ppid" into "ppid"
# TODO: Beware, a CGI parameter might be there. CGIPROP
def qname(x, grph):
try:
q = grph.compute_qname(x)
# q[0] is the shortened namespace "ns"
# Could return q[0] + ":" + q[2]
return q[2]
except:
return x
# Nothing really interesting at the moment, just hardcodes.
return lib_properties.prop_color(prop)
# The input is any Python object.
# This returns a simple object which can be transformed into a string.
# If the input is a container, it returns a HTML table.
def FormatElementAux(val, depth=0):
if val is None:
return ""
try:
int(val)
return val
except:
pass
try:
float(val)
return val
except:
pass
if isinstance(val, dict):
subTable = ""
# TODO: Consider using six.iteritems.
for subKey, subVal in val.items():
subTd = FormatPair(subKey, subVal, depth + 1)
if subTd:
subTable += "<tr>%s</tr>" % subTd
return "<table border='0'>%s</table>" % subTable
# Note: Recursive list are not very visible.
if isinstance(val, (list, tuple)):
# If this is an empty list or tuple.
if not val:
# return "(Empty)"
# Empty set character in UTF8
return "{" + "∅" + "}"
if depth % 2 == 0:
subTable = ""
for subElement in val:
subTd = FormatElement(subElement, depth + 1)
subTable += "<tr>%s</tr>" % subTd
return "<table border='0'>%s</table>" % subTable
else:
subTable = ""
for subElement in val:
subTd = FormatElement(subElement, depth + 1)
subTable += subTd
return "<table border='0'><tr>%s</tr></table>" % subTable
try:
decodVal = json.loads(val)
return FormatElementAux(decodVal, depth + 1)
except ValueError:
# It is a string which cannot be converted to json.
val = cgi.escape(val)
return lib_exports.StrWithBr(val)
except TypeError:
# "Expected a string or buffer"
# It is not a string, so it could be a datetime.datetime
val = cgi.escape(str(val))
return lib_exports.StrWithBr(val)
return "FormatElement failure"
def FormatElement(val, depth=0):
if lib_kbase.IsLink(val):
valTitle = "FormatElement " + ExternalToTitle(val)
valTitleUL = lib_exports.DotUL(valTitle)
return "<td align='left' balign='left' border='0' href='%s'>%s</td>" % (
val, valTitleUL)
resStr = FormatElementAux(val, depth)
return "<td align='left' balign='left' border='0'>%s</td>" % resStr
# Prints a key-value pair as two TD tags, to go in an HTML table.
def FormatPair(key, val, depth=0):
colFirst = "<td align='left' valign='top' border='0'>%s</td>" % lib_exports.DotBold(
key)
colSecond = FormatElement(val, depth + 1)
return colFirst + colSecond
# Display in the DOT node the list of its literal properties.
def FieldsToHtmlVertical(grph, the_fields):
props = {}
idx = 0
# TODO: The sort must put at first, some specific keys.
# For example, sources_top/nmap_run.py, the port number as an int (Not a string)
# Also, filenames, case-sensitive or not.
for (key, val) in sorted(the_fields):
# This should come first, but it does not so we prefix with "----". Hack !
if key == pc.property_information:
# Completely left-aligned. Col span is 2, approximate ratio.
val = lib_exports.StrWithBr(val, 2)
currTd = "<td align='left' balign='left' colspan='2'>%s</td>" % val
elif IsFlatProperty(key):
urlTxt = lib_naming.ParseEntityUri(val)[0]
splitTxt = lib_exports.StrWithBr(urlTxt, 2)
# The text of the link must be underlined.
currTd = '<td href="%s" align="left" colspan="2">%s</td>' % (
val, lib_exports.DotUL(splitTxt))
else:
key_qname = qname(key, grph)
# This assumes: type(val) == 'rdflib.term.Literal'
# sys.stderr.write("FORMAT ELEMENT: %s\n" %(dir(val)))
if lib_kbase.IsLiteral(val):
currTd = FormatPair(key_qname, val.value)
else:
currTd = FormatPair(key_qname, val)
props[idx] = currTd
idx += 1
return props
# Ca liste les labels des objects qui apparaissent dans les blocs,
# et pointent vers le nom du record.
dictCollapsedObjectLabelsToSubjectLabels = {}
# This contains, for each node (subject), the related node (object) linked
# to it with a property to be displayed in tables instead of individual nodes.
dictPropsCollapsedSubjectsToObjectLists = {}
for collapsPropObj in CollapsedProperties:
collapsPropNam = lib_exports.PropToShortPropNam(collapsPropObj)
dictPropsCollapsedSubjectsToObjectLists[
collapsPropNam] = collections.defaultdict(list)
# TODO: (TRANSLATE THIS) Une premiere passe pour batir l'arbre d'une certaine propriete.
# Si pas un DAG, tant pis, ca fera un lien en plus.
# ON voulait batir des records, mais les nodes dans un record ne peuvent pas
# avoir un URL: Donc ca va pas evidemment.
# HTML-LIKE Labels avec PORT et PORTPOS.
# CA VA AUSSI SIMPLIFIER L'AFFICHAGE DES TRUCS ENORMES: Modules, Fichiers etc...
# Et on pourra trier car il y a un ordre.
# Donc ca doit etre facile d'ajouter des proprietes affichees comme ca.
logfil.write(TimeStamp() + " Rdf2Dot: First pass\n")
# New intermediary node created.
def CollapsedLabel(collapsPropNam, subjNam):
return "R_" + collapsPropNam + "_" + subjNam
# Called mainly from entity.py. If S points vers O, transforms "O" => "R_S:O"
# Accordingly we create an edge: "S" => "R_S"
def SubjNamFromCollapsed(collapsPropNam, subjNam):
#sys.stderr.write("ADDING1 subjNam=%s collapsPropNam=%s\n" % (subjNam,collapsPropNam))
collapsedSubjNam = dictCollapsedObjectLabelsToSubjectLabels[subjNam][
collapsPropNam]
#sys.stderr.write("ADDING2 subjNam=%s collapsPropNam=%s\n" % (subjNam,collapsPropNam))
newSubjNam = CollapsedLabel(collapsPropNam,
collapsedSubjNam) + ":" + subjNam
#sys.stderr.write("ADDED collapsedSubjNam=%s newSubjNam=%s collapsPropNam=%s\n" % (collapsedSubjNam,newSubjNam,collapsPropNam))
return newSubjNam
# This is sorted so the result is deterministic. Very small performance impact.
# Any order will do as long as the result is always the same.
sortedGrph = sorted(grph)
# TODO: Loop only on the "collapsed" properties, the ones whose objects must be displayed
# in tables, instead of links - if only they have a single subject. Otherwise it cannot work.
for subj, prop, obj in sortedGrph:
# Objects linked with these properties, are listed in a table, instead of distinct nodes in a graph.
if prop in CollapsedProperties:
# TODO: We lose the property, unfortunately. Should make a map: subject => prop => object ?
subjNam = RdfNodeToDotLabel(subj)
propNam = lib_exports.PropToShortPropNam(prop)
dictPropsCollapsedSubjectsToObjectLists[propNam][subj].append(obj)
# Maybe we already entered it: Not a problem.
objNam = RdfNodeToDotLabel(obj)
# CollapsedProperties can contain only properties which define a tree,
# as visibly the "object" nodes can have one ancestor only.
try:
# TODO: We should check if a node appears in two tables,
# associated to two properties and/or two parent node.
dictCollapsedObjectLabelsToSubjectLabels[objNam][
propNam] = subjNam
except KeyError:
dictCollapsedObjectLabelsToSubjectLabels[objNam] = dict()
dictCollapsedObjectLabelsToSubjectLabels[objNam][
propNam] = subjNam
# For getting the node of an object, as it might be in a table.
def RdfNodeToDotLabelExtended(obj, prop):
objNam = RdfNodeToDotLabel(obj)
try:
dictOfProps = dictCollapsedObjectLabelsToSubjectLabels[objNam]
except KeyError:
# sys.stderr.write("RdfNodeToDotLabelExtended propNam=%s objNam=%s\n"%(propNam,objNam) )
return objNam
# Let's hope there is only one collapsed property for this node. Otherwise, it means
# that this node would be displayed in two different tables. It happened...
if not prop is None:
propNam = lib_exports.PropToShortPropNam(prop)
try:
# Maybe this property is not collapsed.
subjNam = dictOfProps[propNam]
except KeyError:
prop = None
# Maybe the property is not known, if the node is the subject.
# Or the property is not collapsed.
if prop is None:
# In Python3, keys() is an iterable. No need to create a list.
#propNam = list(dictOfProps.keys())[0]
#propNam = dictOfProps.keys()[0]
for propNam in dictOfProps.keys():
break
# First property available.
subjNam = dictOfProps[propNam]
newObjNam = CollapsedLabel(propNam, subjNam) + ":" + objNam
return newObjNam
# Now we know that we have seen all nodes in a collapsed property.
for subj, prop, obj in sortedGrph:
if prop in CollapsedProperties:
continue
# Maybe the subject node belongs to a table, but the property is not known.
subjNam = RdfNodeToDotLabelExtended(subj, None)
if lib_kbase.IsLink(obj):
prp_col = lib_properties.prop_color(prop)
# TODO: GENERALIZE THIS TO ALL COMMUTATIVE PROPERTIES.
# THAT IS: PROPERTIES WHOSE TRIPLES ARE MERGED WHEN
# WE HAVE AT THE SAME TIME: (Subj,Prop,Obj) and (Obj,Prop,Subj).
# WHEN THIS HAPPENS, THE ARROW MUST BE BIDIRECTIONAL.
# TODO: All commutative relation have bidirectional arrows.
# At the moment, only one property can be bidirectional.
if prop == pc.property_socket_end:
# BEWARE, MAYBE THIS IS A PORT INTO A TABLE. SO IT HAS TO BE PREFIXED BY THE RECORD NAME.
objNam = RdfNodeToDotLabelExtended(obj, prop)
if (obj, prop, subj) in grph:
if subjNam < objNam:
stream.write(
pattEdgeBiDir %
(subjNam, objNam, prp_col, qname(prop, grph)))
else:
# One connection only: We cannot see the other.
stream.write(pattEdgeOrien %
(subjNam, objNam, prp_col, qname(prop, grph)))
elif prop in [
pc.property_rdf_data_nolist1, pc.property_rdf_data_nolist2,
pc.property_rdf_data_nolist3
]:
# TODO: Il suffit de tester si obj est un url de la forme "entity.py" ???
# HTML and images urls can be "flattened" because the nodes have no descendants.
# Do not create a node for this.
# MIME links displayed in the same column as sub-directory.
# Also, it might be enough to test if the object has the form "entity.py" because it has no descendant.
# TODO: CGIPROP: Can it have several html or sub-rdf ?? It is necessary !
fieldsSet[subj].append((prop, obj))
else:
objNam = RdfNodeToDotLabelExtended(obj, prop)
# C est la que si subjNam est dans une liste de dictCollapsedSubjectsToObjectLists,
# il faut rajouter devant, le nom du record, c est a dire SON subjNam + "_table_rdf_data:".
try:
# Syntax with colon required by DOT.
propNam = lib_exports.PropToShortPropNam(prop)
subjNam = SubjNamFromCollapsed(propNam, subjNam)
except KeyError:
# sys.stderr.write("PASS subjNam=%s objNam=%s\n"%(subjNam,objNam))
pass
stream.write(pattEdgeOrien %
(subjNam, objNam, prp_col, qname(prop, grph)))
elif obj == None:
# No element created in nodes[]
fieldsSet[subj].append((prop, "Null"))
else:
# For Literals. No element created in nodes[]
# Literals can be processed according to their type.
# Some specific properties cannot have children so they can be stored as literals?
# Les proprietes comme "pid", on devrait plutot afficher le lien vers le process, dans la table ???
# Les URLs de certaines proprietes sont affichees en colonnes.
# Ou bien pour ces proprietes, on recree un entity.py ??
fieldsSet[subj].append((prop, obj))
logfil.write(TimeStamp() +
" Rdf2Dot: Replacing vectors: CollapsedProperties=%d.\n" %
(len(CollapsedProperties)))
# Now, replaces each vector by a single object containg a HTML table.
# TODO: Unfortunately, the prop is lost, which implies that all children are mixed together.
def ProcessCollapsedProperties(propNam):
dictCollapsedSubjectsToObjectLists = dictPropsCollapsedSubjectsToObjectLists[
propNam]
logfil.write(TimeStamp() +
" Rdf2Dot: dictCollapsedSubjectsToObjectLists=%d.\n" %
(len(dictCollapsedSubjectsToObjectLists)))
for subjUrl, nodLst in lib_util.six_iteritems(
dictCollapsedSubjectsToObjectLists):
subjNam = RdfNodeToDotLabel(subjUrl)
subjNamTab = CollapsedLabel(propNam, subjNam)
try:
# TODO: Cette logique ajoute parfois un niveau de noeud inutile. Plus exactement, ca duplique un noeud.
# Ou plus exactement, le noed est represente par deux objects graphiques:
# * Un qui a les scripts.
# * Un autre qui a la liste HTML qu on fabrique.
# => Peut-on imaginer de melanger les deux ??
# Dans WritePatterns: Ajouter le nom du noeud au label.
subjNam = SubjNamFromCollapsed(propNam, subjNam)
except KeyError:
pass
# Point from the subject to the table containing the objects.
stream.write(pattEdgeOrien %
(subjNam, subjNamTab, "GREEN", propNam))
(labText, subjEntityGraphicClass,
entity_id) = lib_naming.ParseEntityUri(subjUrl)
# Probleme avec les champs:
# Faire une premiere passe et reperer les fields, detecter les noms des colonnes, leur attribuer ordre et indice.
# Seconde passe pour batir les lignes.
# Donc on ordonne toutes les colonnes.
# Pour chaque field: les prendre dans le sens du header et quand il y a un trou, colonne vide.
# Inutile de trier les field, mais il d'abord avoir une liste complete des champs, dans le bon sens.
# CA SUPPOSE QUE DANS FIELDSSET LES KEYS SONT UNIQUES.
# SI ON NE PEUT PAS, ALORS ON METTRA DES LISTES. MAIS CETTE CONTRAINTE SIMPLIFIE L'AFFICHAGE.
# DOMMAGE QU ON SCANNE LES OBJETS DEUX FOIS UNIQUEMENT POUR AVOIR LES NOMS DES CHAMPS !!!!!!!!!!!!!
# TODO: HEURISTIQUE: ON pourrait s'arreter aux dix premiers. Ou bien faire le tri avant ?
# On bien prendre les colonnes de la premiere ligne, et recommencer si ca ne marche pas.
# Unique columns of the descendant of this subject.
rawFieldsKeys = set()
for obj in nodLst:
# One table per node.
rawFieldsKeys.update(fld[0] for fld in fieldsSet[obj])
# sys.stderr.write("rawFieldsKeys BEFORE =%s\n" % str(rawFieldsKeys) )
# Mandatory properties must come at the beginning of the columns of the header, with first indices.
# BUG: Si on retire html de cette liste alors qu il y a des valeurs, colonnes absentes.
# S il y a du html ou du RDF, on veut que ca vienne en premier.
fieldsKeysOrdered = []
for fldPriority in FlatPropertertiesList:
try:
# Must always be appended. BUT IF THERE IS NO html_data, IS IT WORTH ?
# TODO: Remove if not HTML and no sub-rdf. CGIPROP
# If the property is never used, exception then next property.
rawFieldsKeys.remove(fldPriority)
fieldsKeysOrdered.append(fldPriority)
except KeyError:
pass
# This one is always removed because its content is concatenated at the first column.
for fldToRemove in [pc.property_information]:
try:
rawFieldsKeys.remove(fldToRemove)
except KeyError:
pass
# Appends rest of properties, sorted.
fieldsKeys = fieldsKeysOrdered + sorted(rawFieldsKeys)
# sys.stderr.write("fieldsKeys=%s\n" % str(fieldsKeys) )
# This assumes that the header columns are sorted.
keyIndices = {
nameKey: indexKey
for (indexKey, nameKey) in enumerate(fieldsKeys, 1)
}
numberKeys = len(keyIndices) + 1
# Apparently, no embedded tables.
dictHtmlLines = dict()
for objUri in nodLst:
# One table per node.
subObjId = RdfNodeToDotLabel(objUri)
# Beware "\L" which should not be replaced by "<TABLE>" but this is not the right place.
subNodUri = objUri.replace('&', '&')
try:
(subObjNam, subEntityGraphicClass,
subEntityId) = lib_naming.ParseEntityUriShort(objUri)
except UnicodeEncodeError:
sys.stderr.write("UnicodeEncodeError error:%s\n" %
(objUri))
(subObjNam, subEntityGraphicClass,
subEntityId) = ("Utf problem1", "Utf problem2",
"Utf problem3")
# sys.stderr.write("subEntityGraphicClass=%s\n"%subEntityGraphicClass)
# If this is a script, always displayed on white, even if reletd to a specific entity.
# THIS IS REALLY A SHAME BECAUSE WE JUST NEED THE ORIGINAL PROPERTY.
if objUri.find("entity.py") < 0:
objColor = "#FFFFFF"
else:
objColor = lib_patterns.EntityClassToColor(
subEntityGraphicClass)
# This lighter cololor for the first column.
objColorLight = lib_patterns.ColorLighter(objColor)
# Some colors a bit clearer ? Or take the original color of the class ?
td_bgcolor_plain = '<td BGCOLOR="%s" ' % objColor
td_bgcolor_light = '<td BGCOLOR="%s" ' % objColorLight
td_bgcolor = td_bgcolor_plain
# Some columns might not have a value. The first column is for the key.
columns = [td_bgcolor + " ></td>"] * numberKeys
# Just used for the vertical order of lines, one line per object.
title = ""
# TODO: CGIPROP. This is not a dict, the same key can appear several times ?
for (key, val) in fieldsSet[objUri]:
if key == pc.property_information:
# This can be a short string only.
title += val
continue
# TODO: This is hard-coded.
if IsFlatProperty(key):
# In fact, it might also be an internal URL with "entity.py"
if lib_kbase.IsLiteral(val):
if isinstance(val.value, (list, tuple)):
strHtml = FormatElementAux(val.value)
sys.stderr.write("val.value=%s\n" % strHtml)
tmpCell = td_bgcolor + 'align="left">%s</td>' % strHtml
else:
tmpCell = td_bgcolor + 'align="left">%s</td>' % val.value
else:
valTitle = lib_naming.ParseEntityUri(val)[0]
valTitleUL = lib_exports.DotUL(valTitle)
tmpCell = td_bgcolor + 'href="%s" align="left" >%s</td>' % (
val, valTitleUL)
else:
try:
float(val)
tmpCell = td_bgcolor + 'align="right">%s</td>' % val
except:
# Wraps the string if too long. Can happen only with a literal.
tmpCell = td_bgcolor + 'align="left">%s</td>' % lib_exports.StrWithBr(
val)
idxKey = keyIndices[key]
columns[idxKey] = tmpCell
if title:
title_key = title
else:
title_key = subObjNam
# Maybe the first column is a literal ?
if subEntityId != "PLAINTEXTONLY":
# WE SHOULD PROBABLY ESCAPE HERE TOO.
columns[
0] = td_bgcolor_light + 'port="%s" href="%s" align="LEFT" >%s</td>' % (
subObjId, subNodUri, title_key)
else:
subNodUri = cgi.escape(subNodUri)
columns[
0] = td_bgcolor_light + 'port="%s" align="LEFT" >%s</td>' % (
subObjId, subNodUri)
# Several scripts might have the same help text, so add a number.
# "Title" => "Title"
# "Title" => "Title/2"
# "Title" => "Title/3" etc...
# Beware that it is quadratic with the number of scripts with identical info.
title_idx = 2
title_uniq = title_key
while title_uniq in dictHtmlLines:
title_uniq = "%s/%d" % (title_key, title_idx)
title_idx += 1
# TODO: L'ordre est base sur les chaines mais devrait etre base sur le contenu. Exemple:
# TODO: "(TUT_UnixProcess) Handle=10" vient avant "(TUT_UnixProcess) Handle=2"
# TODO: title_uniq devrait etre plutot la liste des proprietes.
# TODO: By clicking on the column names, we could change the order.
dictHtmlLines[title_uniq] = "".join(columns)
# Replace the first column by more useful information.
numNodLst = len(nodLst)
# TODO: Compute this once for all.
eltNam = subEntityGraphicClass.split("/")[-1]
if not eltNam:
# TODO: This is not the right criteria. Must select if we are listing scripts.
eltNam = "script"
eltNamPlural = lib_grammar.ToPlural(eltNam, numNodLst)
txtElements = "%d %s" % (numNodLst, eltNamPlural)
header = '<td border="1">' + lib_exports.DotBold(
txtElements) + "</td>"
# TODO: Replace each column name with a link which sorts the line based on this column.
# The order of columns could be specified with an extra cgi argument with the columns names.
for key in fieldsKeys:
columnTitle = qname(key, grph)
columnTitle = columnTitle.replace("_", " ").capitalize()
header += "<td border='1'>" + lib_exports.DotBold(
columnTitle) + "</td>"
# With an empty key, it comes first when sorting.
dictHtmlLines[""] = header
# MAYBE SHOULD BE DONE TWICE !!!!! SEE ALSO ELSEWHERE !!!!
subjUrlClean = subjUrl.replace('&', '&')
# ATTENTION: La forme du record est celle du sujet.
# ca veut donc dire qu'on va avoir la meme couleur pour des objets de types
# differents s'ils sont dans la meme relation avec un sujet identique ?
numFields = len(fieldsKeys) + 1
# The label might be truncated
if subjEntityGraphicClass:
helpText = "List of " + subjEntityGraphicClass + " objects in " + labText
else:
helpText = "List of scripts in " + labText
# TODO: Le title and the content are not necessarily of the same class.
# labTextWithBr is the first line of the table containing nodes linked with the
# same property. Unfortunately we have lost this property.
labText = lib_exports.TruncateInSpace(labText, 30)
labTextWithBr = lib_exports.StrWithBr(labText)
labTextWithBr += ": " + propNam
if entity_id == "PLAINTEXTONLY":
subjUrlClean = ""
# This color is the table's contour.
lib_patterns.WritePatterned(stream, subjEntityGraphicClass,
subjNamTab, helpText, '"#000000"',
subjUrlClean, numFields, labTextWithBr,
dictHtmlLines)
# TODO: Eviter les repetitions de la meme valeur dans une colonne en comparant d une ligne a l autre.
# TODO: Si une cellule est identique jusqu a un delimiteur, idem, remplacer par '"'.
if CollapsedProperties:
for collapsedProp in CollapsedProperties:
collapsedPropNam = lib_exports.PropToShortPropNam(collapsedProp)
ProcessCollapsedProperties(collapsedPropNam)
logfil.write(TimeStamp() +
" Rdf2Dot: Display remaining nodes. dictRdf2Dot=%d\n" %
len(dictRdf2Dot))
# Now, display the normal nodes, which are not displayed in tables.
for objRdfNode, objLabel in lib_util.six_iteritems(dictRdf2Dot):
# TODO: Avoids this lookup.
if objLabel in dictCollapsedObjectLabelsToSubjectLabels:
continue
objPropsAsHtml = FieldsToHtmlVertical(grph, fieldsSet[objRdfNode])
labHRef = objRdfNode.replace('&', '&')
try:
# TODO: Probleme ici: La chaine est deja codee pour HTML ce qui en rend le parsing different
# TODO: ... de celui d'un URL deja decode. DOMMAGE: On quote puis unquote !!!
(labText, objEntityGraphClass,
entity_id) = lib_naming.ParseEntityUri(
lib_util.urllib_unquote(objRdfNode))
except UnicodeEncodeError:
sys.stderr.write("UnicodeEncodeError error:%s\n" % (objRdfNode))
# WritePatterned receives an list of strings similar to "<td>jhh</td><td>jhh</td><td>jhh</td>"
# This function adds <tr> and </tr> on both sides.
# This avoids concatenations.
# Ampersand are intentionnally doubled, because later on they are replaced twice.
# That is, interpreted twice as HTML entities.
# This might be temporary until we replace CGI arguments by genuine WMI Monikers.
labTextNoAmp = labText.replace("&amp;", " ")
labTextNoAmp = labTextNoAmp.strip()
labTextClean = lib_exports.StrWithBr(labTextNoAmp)
# Two columns because it encompasses the key and the value.
if objEntityGraphClass:
helpText = labTextNoAmp + " is a " + objEntityGraphClass
else:
if labTextClean.startswith("http"):
helpText = "External URL " + labTextNoAmp
else:
helpText = "Script " + labTextNoAmp
# This color is the object's contour.
lib_patterns.WritePatterned(stream, objEntityGraphClass, objLabel,
helpText, '"#000000"', labHRef, 2,
labTextClean, objPropsAsHtml)
logfil.write(TimeStamp() + " Rdf2Dot: Leaving\n")
stream.write("}\n")
def _rdf_graph_to_dot_stream(grph, logfil, stream, collapsed_properties, commutative_properties):
"""Used for transforming into SVG format.
If from entity.py, collapsed_properties = pc.property_directory,pc.property_script """
fields_set = collections.defaultdict(list)
# This maps RDFLIB nodes to DOT label names.
dict_rdf2_dot = {}
# This returns the DOT label of a RDFLIB, and creates a new one if necessary.
def _rdf_node_to_dot_label(x):
try:
return dict_rdf2_dot[x]
except KeyError:
nodelabel = "nd_%d" % len(dict_rdf2_dot)
dict_rdf2_dot[x] = nodelabel
return nodelabel
# The input is any Python object.
# This returns a simple object which can be transformed into a string.
# If the input is a container, it returns a HTML table.
def _format_element_aux(val, depth=0):
if val is None:
return ""
try:
int(val)
return val
except:
pass
try:
float(val)
return val
except:
pass
if isinstance(val, dict):
sub_table = ""
# TODO: Consider using six.iteritems.
for sub_key, sub_val in val.items():
sub_td = _format_pair(sub_key, sub_val, depth + 1)
if sub_td:
sub_table += "<tr>%s</tr>" % sub_td
return "<table border='0'>%s</table>" % sub_table
# Note: Recursive list are not very visible.
if isinstance(val, (list, tuple)):
# If this is an empty list or tuple.
if not val:
# Empty set character in UTF8
return "{" + "∅" + "}"
if depth % 2 == 0:
sub_table = ""
for sub_element in val:
sub_td = _format_element(sub_element, depth + 1)
sub_table += "<tr>%s</tr>" % sub_td
return "<table border='0'>%s</table>" % sub_table
else:
sub_table = ""
for sub_element in val:
sub_td = _format_element(sub_element, depth + 1)
sub_table += sub_td
return "<table border='0'><tr>%s</tr></table>" % sub_table
try:
decod_val = json.loads(val)
return _format_element_aux(decod_val, depth + 1)
except ValueError:
# It is a string which cannot be converted to json.
val = lib_util.html_escape(val)
return _str_with_br(val)
except TypeError:
# "Expected a string or buffer"
# It is not a string, so it could be a datetime.datetime
val = lib_util.html_escape(str(val))
return _str_with_br(val)
return "_format_element failure"
def _format_element(val, depth=0):
if lib_kbase.IsLink(val):
val_title = "_format_element " + external_url_to_title(val)
val_title_ul = _dot_ul(val_title)
return "<td align='left' balign='left' border='0' href='%s'>%s</td>" % (val, val_title_ul)
res_str = _format_element_aux(val, depth)
return "<td align='left' balign='left' border='0'>%s</td>" % res_str
# Prints a key-value pair as two TD tags, to go in an HTML table.
def _format_pair(key, val, depth=0):
col_first = "<td align='left' valign='top' border='0'>%s</td>" % _dot_bold(key)
col_second = _format_element(val, depth+1)
return col_first + col_second
# Display in the DOT node the list of its literal properties.
def fields_to_html_vertical(grph, the_fields):
props = {}
idx = 0
# TODO: The sort must put at first, some specific keys.
# For example, sources_top/nmap_run.py, the port number as an int (Not a string)
# Also, filenames, case-sensitive or not.
for key, val in sorted(the_fields):
# This should come first, but it does not so we prefix with "----". Hack !
if key == pc.property_information:
# Completely left-aligned. Col span is 2, approximate ratio.
val = _str_with_br(val, 2)
curr_td = "<td align='left' balign='left' colspan='2'>%s</td>" % val
elif is_flat_property(key) :
url_txt = lib_naming.ParseEntityUri(val)[0]
split_txt = _str_with_br(url_txt, 2)
# The text of the link must be underlined.
curr_td = '<td href="%s" align="left" colspan="2">%s</td>' % (val, _dot_ul(split_txt))
else:
key_qname = lib_kbase.qname(key, grph)
# This assumes: type(val) == 'rdflib.term.Literal'
if lib_kbase.IsLiteral(val):
curr_td = _format_pair(key_qname, val.value)
else:
curr_td = _format_pair(key_qname, val)
props[idx] = curr_td
idx += 1
return props
# This lists the labels of objects which appear in the blocks,
# and point to the name of records.
dict_collapsed_object_labels_to_subject_labels = {}
# This contains, for each node (subject), the related node (object) linked
# to it with a property to be displayed in tables instead of individual nodes.
dict_props_collapsed_subjects_to_object_lists = {}
for collaps_prop_obj in collapsed_properties:
# TODO: Les arguments CGI de l'URL ils servent peut-etre a trier.
# Donc on doit les garder pour trier une fois qu'on a rassemble.
collaps_prop_nam = lib_exports.PropToShortPropNam(collaps_prop_obj)
dict_props_collapsed_subjects_to_object_lists[collaps_prop_nam] = collections.defaultdict(list)
logfil.write(lib_util.TimeStamp() + " Rdf2Dot: First pass\n")
def collapsed_label(collaps_prop_nam, subj_nam):
"""New intermediary node created."""
assert collaps_prop_nam.find("#") < 0
return "R_" + collaps_prop_nam + "_" + subj_nam
# Called mainly from entity.py. If S points vers O, transforms "O" => "R_S:O"
# Accordingly we create an edge: "S" => "R_S"
def subj_nam_from_collapsed(collaps_prop_nam, subj_nam):
collapsed_subj_nam = dict_collapsed_object_labels_to_subject_labels[subj_nam][collaps_prop_nam]
new_subj_nam = collapsed_label(collaps_prop_nam, collapsed_subj_nam) + ":" + subj_nam
return new_subj_nam
# This is sorted so the result is deterministic. Very small performance impact.
# Any order will do as long as the result is always the same for the same URL, if the content is identical.
sorted_grph = sorted(grph)
# TODO: Loop only on the "collapsed" properties, the ones whose objects must be displayed
# in tables, instead of links - if only they have a single subject. Otherwise it cannot work.
for subj, prop, obj in sorted_grph:
# Objects linked with these properties, are listed in a table, instead of distinct nodes in a graph.
if prop in collapsed_properties:
# TODO: We lose the property, unfortunately. Should make a map: subject => prop => object ?
subj_nam = _rdf_node_to_dot_label(subj)
prop_nam = lib_exports.PropToShortPropNam(prop)
dict_props_collapsed_subjects_to_object_lists[prop_nam][subj].append(obj)
# Maybe we already entered it: Not a problem.
obj_nam = _rdf_node_to_dot_label(obj)
# collapsed_properties can contain only properties which define a tree,
# as visibly the "object" nodes can have one ancestor only.
try:
# TODO: We should check if a node appears in two tables,
# associated to two properties and/or two parent node.
dict_collapsed_object_labels_to_subject_labels[obj_nam][prop_nam] = subj_nam
except KeyError:
dict_collapsed_object_labels_to_subject_labels[obj_nam] = {prop_nam: subj_nam}
# For getting the node of an object, as it might be in a table.
def rdf_node_to_dot_label_extended(obj, prop):
obj_nam = _rdf_node_to_dot_label(obj)
try:
dict_of_props = dict_collapsed_object_labels_to_subject_labels[obj_nam]
except KeyError:
return obj_nam
# Let's hope there is only one collapsed property for this node. Otherwise, it means
# that this node would be displayed in two different tables. It happened...
if not prop is None:
prop_nam = lib_exports.PropToShortPropNam(prop)
try:
# Maybe this property is not collapsed.
subj_nam = dict_of_props[prop_nam]
except KeyError:
prop = None
# Maybe the property is not known, if the node is the subject.
# Or the property is not collapsed.
if prop is None:
# In Python3, keys() is an iterable. No need to create a list.
for sub_prop_nam in dict_of_props.keys():
break
# First property available.
subj_nam = dict_of_props[sub_prop_nam]
new_obj_nam = collapsed_label(sub_prop_nam, subj_nam) + ":" + obj_nam
return new_obj_nam
# Now we know that we have seen all nodes in a collapsed property.
for subj, prop, obj in sorted_grph:
if prop in collapsed_properties:
continue
# Maybe the subject node belongs to a table, but the property is not known.
subj_nam = rdf_node_to_dot_label_extended(subj, None)
if lib_kbase.IsLink(obj):
prp_col = lib_properties.prop_color(prop)
# TODO: REMOVE THIS HARDCODE, GENERALIZE THIS TO ALL COMMUTATIVE PROPERTIES,
# TODO: PROPERTIES WHOSE TRIPLES ARE MERGED WHEN SIMULTANEOUSLY: (Subj,Prop,Obj) and (Obj,Prop,Subj).
# TODO: WHEN THIS HAPPENS, THE ARROW MUST BE BIDIRECTIONAL. Commutative triples have bidirectional arrows.
# TODO: Look for "commutative_property" and lib_properties.add_property_metadata_to_graph()
# TODO: At the moment, only one property can be bidirectional: property_socket_end
# if prop == pc.property_socket_end:
if prop in commutative_properties:
# BEWARE, MAYBE THIS IS A PORT INTO A TABLE. SO IT HAS TO BE PREFIXED BY THE RECORD NAME.
obj_nam = rdf_node_to_dot_label_extended(obj, prop)
if (obj, prop, subj) in grph :
if subj_nam < obj_nam:
stream.write(
_pattern_edge_bidirect
% (subj_nam, obj_nam, prp_col, lib_kbase.qname(prop, grph)))
else:
# One connection only: We cannot see the other.
stream.write(
_pattern_edge_oriented
% (subj_nam, obj_nam, prp_col, lib_kbase.qname(prop, grph)))
elif prop in _flat_properties_list:
# HTML and images urls can be "flattened" because the nodes have no descendants.
# Do not create a node for this.
# MIME links displayed in the same column as sub-directory.
# Also, it might be enough to test if the object has the form "entity.py" because it has no descendant.
# TODO: CGIPROP: Can it have several html or sub-rdf ?? It is necessary !
fields_set[subj].append((prop, obj))
else:
obj_nam = rdf_node_to_dot_label_extended(obj, prop)
# If subj_nam is in a list of dict_collapsed_subjects_to_object_lists,
# one must add at front, the record name, i.e. its subj_nam + "_table_rdf_data:".
try:
# Syntax with colon required by DOT.
prop_nam = lib_exports.PropToShortPropNam(prop)
subj_nam = subj_nam_from_collapsed(prop_nam, subj_nam)
except KeyError:
pass
stream.write(_pattern_edge_oriented % (subj_nam, obj_nam, prp_col, lib_kbase.qname(prop, grph)))
elif obj == None:
# No element created in nodes[]
fields_set[subj].append((prop, "Null"))
else:
# For Literals. No element created in nodes[]
# Literals can be processed according to their type.
# Some specific properties cannot have children so they can be stored as literals?
fields_set[subj].append((prop, obj))
logfil.write(lib_util.TimeStamp() + " Rdf2Dot: Replacing vectors.\n")
logfil.write(lib_util.TimeStamp() + " Rdf2Dot: Number of collapsed properties=%d.\n" % len(collapsed_properties))
# Now, replaces each vector by a single object containg a HTML table.
# TODO: Unfortunately, the prop is lost, which implies that all children are mixed together.
def _process_collapsed_properties(prop_nam):
dict_collapsed_subjects_to_object_lists = dict_props_collapsed_subjects_to_object_lists[prop_nam]
logfil.write(lib_util.TimeStamp()+" Rdf2Dot: dict_collapsed_subjects_to_object_lists=%d.\n"
% len(dict_collapsed_subjects_to_object_lists))
for subj_url, nod_lst in six.iteritems(dict_collapsed_subjects_to_object_lists):
subj_nam = _rdf_node_to_dot_label(subj_url)
subj_nam_tab = collapsed_label(prop_nam, subj_nam)
try:
# TODO: This logic adds an extra level of node: Try to flatten the tree.
subj_nam = subj_nam_from_collapsed(prop_nam, subj_nam)
except KeyError:
pass
# This points from the subject to the table containing the objects.
# TODO: This color should be a parameter.
stream.write(_pattern_edge_oriented % (subj_nam, subj_nam_tab, "GREEN", prop_nam))
lab_text, subj_entity_graphic_class, entity_id = lib_naming.ParseEntityUri(subj_url)
# At the moment, two passes are necessary:
# * A first pass to create the total list of fields, because they might be a bit different
# from one record to the other. The column names of these fields get an unique index number
# and can therefore be sorted.
# * A second pass uses these result, to display the lines.
#
# This could be faster by assuming that the first ten or twenty columns have all the fields.
# We could then start the second pass, and if an undetected column is found,
# then restart from the beginning, as it is done now.
# Unique columns of the descendant of this subject.
raw_fields_keys = set()
for obj in nod_lst:
# One table per node.
raw_fields_keys.update(fld[0] for fld in fields_set[obj])
# Mandatory properties must come at the beginning of the columns of the header, with first indices.
fields_keys_ordered = []
for fld_priority in _flat_properties_list:
try:
# Must always be appended. BUT IF THERE IS NO html_data, IS IT WORTH ?
# If the property is never used, exception then next property.
raw_fields_keys.remove(fld_priority)
fields_keys_ordered.append(fld_priority)
except KeyError:
pass
# This one is always removed because its content is concatenated at the first column.
for fld_to_remove in [pc.property_information]:
try:
raw_fields_keys.remove(fld_to_remove)
except KeyError:
pass
# Appends rest of properties which are the column names, alphabetically sorted.
fields_keys = fields_keys_ordered + sorted(raw_fields_keys)
# This assumes that the header columns are sorted by alphabetical order.
key_indices = {name_key: index_key for (index_key, name_key) in enumerate(fields_keys, 1)}
number_keys = len(key_indices)+1
# Apparently, no embedded tables.
dict_html_lines = dict()
for obj_uri in nod_lst:
# One table per node.
sub_obj_id = _rdf_node_to_dot_label(obj_uri)
# Beware "\L" which should not be replaced by "<TABLE>" but this is not the right place.
sub_nod_uri = obj_uri.replace('&', '&')
try:
sub_obj_nam, sub_entity_graphic_class, sub_entity_id = lib_naming.parse_entity_uri_short(obj_uri)
except UnicodeEncodeError:
logging.warning("UnicodeEncodeError error:%s", obj_uri)
sub_obj_nam, sub_entity_graphic_class, sub_entity_id = ("Utf err 1", "Utf err 2", "Utf err 3")
# If this is a script, always displayed on white, even if related to a specific entity.
# THIS IS REALLY A SHAME BECAUSE WE JUST NEED THE ORIGINAL PROPERTY.
if obj_uri.find("entity.py") < 0:
obj_color_plain = "#FFFFFF"
else:
obj_color_plain = lib_patterns.EntityClassToColor(sub_entity_graphic_class)
# This lighter color for the first column.
obj_color_light = lib_patterns.color_lighter(obj_color_plain)
# Some colors a bit clearer ? Or take the original color of the class ?
td_bgcolor_plain = '<td BGCOLOR="%s" ' % obj_color_plain
td_bgcolor_light = '<td BGCOLOR="%s" ' % obj_color_light
# Some columns might not have a value. The first column is for the key.
html_columns = [td_bgcolor_plain + " ></td>"] * number_keys
# Just used for the vertical order of lines, one line per object.
concatenated_info_values = ""
for key, val in fields_set[obj_uri]:
# TODO: This property is by default the sorting key:
# TODO: This can be a parameter for lists of classes <MY_Class>
# TODO: ... by adding triplets of the form: (<MY_Class>, sorting_key, pc.property_information)
if key == pc.property_information:
# This can be a short string only.
# Instead of concatenation, consider a list, or use an unique delimiter.
concatenated_info_values += val
# If there is a key, it overrides
sub_entity_id = "NOT_" + "PLAINTEXTONLY" # val
continue
# TODO: This is hard-coded.
if is_flat_property(key) :
# In fact, it might also be an internal URL with "entity.py"
if lib_kbase.IsLiteral(val):
if isinstance(val.value, (list, tuple)):
str_html = _format_element_aux(val.value)
tmp_cell = 'align="left">%s</td>' % str_html
else:
tmp_cell = 'align="left">%s</td>' % val.value
else:
# This displays objects in a table: The top-level object must be
# in the same host, so there is no need to display a long label.
val_title = lib_naming.parse_entity_uri_short(val)[0]
assert isinstance(val_title, (six.text_type, six.binary_type))
# This could probably be replaced by "str"
# There might be non-ascii characters such as accents etc...
try:
val_title.encode('ascii')
except UnicodeEncodeError:
val_title = "Not ascii"
val_title_ul = _dot_ul(val_title)
tmp_cell = 'href="%s" align="left" >%s</td>' % (val, val_title_ul)
else:
try:
float(val)
tmp_cell = 'align="right">%s</td>' % val
except:
# Wraps the string if too long. Can happen only with a literal.
tmp_cell = 'align="left">%s</td>' % _str_with_br(val)
idx_key = key_indices[key]
html_columns[idx_key] = td_bgcolor_plain + tmp_cell
if concatenated_info_values:
title_key = concatenated_info_values
else:
title_key = sub_obj_nam
# Maybe the first column is a literal, and not an object ?
if sub_entity_id != "PLAINTEXTONLY":
# TODO: WE SHOULD PROBABLY ESCAPE HERE TOO.
# For example, this displays the column labelled with pc.property_information
tmp_col_0 = 'port="%s" href="%s" align="LEFT" >%s</td>' % (sub_obj_id, sub_nod_uri, title_key)
else:
sub_nod_uri = lib_util.html_escape(sub_nod_uri)
# For example, this displays the title of another table: Typically sub-scripts.
# The title itself is not an URL.
tmp_col_0 = 'port="%s" align="LEFT" >%s</td>' % (sub_obj_id, sub_nod_uri)
html_columns[0] = td_bgcolor_light + tmp_col_0
# concatenated_info_values
# Several scripts might have the same help text, so add a number.
# "Title" => "Title"
# "Title" => "Title/2"
# "Title" => "Title/3" etc...
# Beware that it is quadratic with the number of scripts with identical info.
title_idx = 2
title_key_uniq = title_key
while title_key_uniq in dict_html_lines:
title_key_uniq = "%s/%d" % (title_key, title_idx)
title_idx += 1
# TODO: The sorting order is based on these strings but should rather be based on the content.
# TODO: For example, "(TUT_UnixProcess) Handle=10" comes before "(TUT_UnixProcess) Handle=2".
# TODO: This is later sorted by the function lib_util.natural_sort_list.
# TODO: Or: title_key_uniq should rather be replaced by the list of properties, for example.
# TODO: By clicking on the column names, we could change the order.
# TODO: Another possibility is to have a "key" metadata which would replace title_key_uniq.
dict_html_lines[title_key_uniq] = "".join(html_columns)
# Replace the first column by more useful information.
num_nod_lst = len(nod_lst)
# WBEM and WMI classes have the syntax: "ns1/ns2/ns3:class" and the class it self can have base classes.
# Survol classes have the syntax: "dir/dir/dir/class": This considers that namespaces are not really
# necessary and can be replaced by classes. Also, there is a one-to-one match between the class inheritance
# tree and its directory.
# If Survol had to be started from scratch, there would be one Python class per survol class,
# and they would be stored in the top dir "root/cimv2" ... it is not too late !
#
# This strips the upper directories: "mysql/instance" or "oracle/table", if this is a Survol class
elt_nam = sub_entity_graphic_class.split("/")[-1]
# This strips the namespace: "root/cimv2:CIM_LogicalElement", if this is a WBEM or WMI class.
elt_nam = elt_nam.split(":")[-1]
if not elt_nam:
# TODO: This is not the right criteria. Must select if we are listing scripts.
elt_nam = "script"
elt_nam_plural = lib_grammar.to_plural(elt_nam, num_nod_lst)
txt_elements = "%d %s" % (num_nod_lst, elt_nam_plural)
header = '<td border="1">%s</td>' % _dot_bold(txt_elements)
# TODO: Replace each column name with a link which sorts the line based on this column.
# The order of columns could be specified with an extra cgi argument with the columns names.
for key in fields_keys:
column_title = lib_kbase.qname(key, grph)
column_title = column_title.replace("_"," ").capitalize()
header += "<td border='1'>%s</td>" % _dot_bold(column_title)
# With an empty key, it comes first when sorting.
dict_html_lines[""] = header
# MAYBE SHOULD BE DONE TWICE !!!!! SEE ALSO ELSEWHERE !!!!
subj_url_clean = subj_url.replace('&', '&')
# BEWARE: The shape and the color of this HTML table is from the subjects,
# because the elements can be of different classes, even if the share the same predicate.
# TODO: Each row should have its own color according to its class.
num_fields = len(fields_keys)+1
# The rows of this HTML table could belong to different classes:
# What the shared is the predicate. Hence, the predicate, property name is used as a title.
prop_nam_plural = lib_grammar.to_plural(prop_nam, None)
help_text = "List of " + prop_nam_plural + " in " + lab_text
# TODO: The title and the content are not necessarily of the same class.
# lab_text_with_br is the first line of the table containing nodes linked with the
# same property. Unfortunately we have lost this property.
lab_text = _truncate_in_space(lab_text, 30)
lab_text_with_br = _str_with_br(lab_text)
lab_text_with_br += ": " + prop_nam
# No object with this script.
if entity_id == "PLAINTEXTONLY":
subj_url_clean = ""
# This color is the table's contour.
lib_patterns.WritePatterned(
stream,
subj_entity_graphic_class,
subj_nam_tab,
help_text,
'"#000000"',
subj_url_clean,
num_fields,
lab_text_with_br,
dict_html_lines)
# TODO: Sometimes, the same value is repeated in contiguous celles of the sames columns.
# TODO: This could be avoided with the character '"': One just need to compare the values
# TODO: ... consecutive cells of the same column.
# TODO: One can even do that if the first N words of a following cell are identical.
if collapsed_properties :
for collapsed_prop in collapsed_properties:
collapsed_prop_nam = lib_exports.PropToShortPropNam(collapsed_prop)
_process_collapsed_properties(collapsed_prop_nam)
logfil.write(lib_util.TimeStamp() + " Rdf2Dot: Display remaining nodes. dict_rdf2_dot=%d\n" % len(dict_rdf2_dot))
# Now, display the normal nodes, which are not displayed in tables.
for obj_rdf_node, obj_label in six.iteritems(dict_rdf2_dot):
# TODO: Avoids this lookup.
if obj_label in dict_collapsed_object_labels_to_subject_labels:
continue
obj_props_as_html = fields_to_html_vertical(grph, fields_set[obj_rdf_node])
lab_href = obj_rdf_node.replace('&', '&')
try:
# TODO: The chain is already encoded for HTML, so the parsing is different
# TODO: ... of an URL already encoded. They are quoted then unquoted.
lab_text, obj_entity_graph_class, entity_id = lib_naming.ParseEntityUri(
lib_util.urllib_unquote(obj_rdf_node))
except UnicodeEncodeError:
logging.warning("UnicodeEncodeError error:%s", obj_rdf_node)
# WritePatterned receives an list of strings similar to "<td>jhh</td><td>jhh</td><td>jhh</td>"
# This function adds <tr> and </tr> on both sides.
# This avoids concatenations.
# Ampersand are intentionally doubled, because later on they are replaced twice.
# That is, interpreted twice as HTML entities.
# This might be temporary until we replace CGI arguments by genuine WMI Monikers.
lab_text_no_amp = lab_text.replace("&amp;", " ")
lab_text_no_amp = lab_text_no_amp.strip()
lab_text_clean = _str_with_br(lab_text_no_amp)
# Two columns because it encompasses the key and the value.
if obj_entity_graph_class:
help_text = lab_text_no_amp
if not help_text:
help_text = "Top-level script"
# This condition is for WMI and WBEM where the name of the node is also a class or a namespace.
# This is a bit convoluted, and just for nicer display.
# "root/cimv2 (WBEM subclasses) at http://vps516494.ovh.net:5988 is a root/cimv2:"
# "wmi_namespace is a wmi_namespace"
elif not lab_text_no_amp.startswith(obj_entity_graph_class.replace(":", " ")):
if obj_entity_graph_class:
# "is a" or "is an"
the_article = lib_grammar.indefinite_article(obj_entity_graph_class)
help_text += " is %s %s" % (the_article, obj_entity_graph_class)
else:
if lab_text_clean.startswith("http"):
help_text = "External URL " + lab_text_no_amp
else:
help_text = "Script " + lab_text_no_amp
# This color is the object's contour.
lib_patterns.WritePatterned(
stream,
obj_entity_graph_class,
obj_label,
help_text,
'"#000000"',
lab_href,
2,
lab_text_clean,
obj_props_as_html)
logfil.write(lib_util.TimeStamp() + " Rdf2Dot: Leaving\n")
stream.write("}\n")