def _bind_type(self, obj):
"""
If object has a type (publication_type, organization_type, event_type etc.)
this method assigns type from the database enum to this object. If the type
doesn't exist, method returns False. If everything went OK, returns True.
If object doesn't have a type, raises DatabaseError.
The object itself has to be inserted into database before binding type!
(it fact, it has to contain attribute ID). If not, method raises Exception.
"""
if obj['id'] is None:
raise DatabaseError("Object has to contain attribute ID. Insert it into db first.")
if 'type' in obj:
type_name = obj['type']['type']
q = FluentSQLQuery()
q.select("id").from_table(obj['type']._table_name).where("type=", type_name)
q()
res = q.fetch_one()
if res is None:
return False
type_id = res['id']
q.cleanup()
q.update(obj._table_name, {"type_id": type_id}).where("id=", obj['id'])
q()
return q.count() == 1
else:
raise RRSDatabaseEntityError("Entity doesn't have a type to bind to.")
def _bind_type(self, obj):
"""
If object has a type (publication_type, organization_type, event_type etc.)
this method assigns type from the database enum to this object. If the type
doesn't exist, method returns False. If everything went OK, returns True.
If object doesn't have a type, raises DatabaseError.
The object itself has to be inserted into database before binding type!
(it fact, it has to contain attribute ID). If not, method raises Exception.
"""
if obj["id"] is None:
raise DatabaseError("Object has to contain attribute ID. Insert it into db first.")
if "type" in obj:
type_name = obj["type"]["type"]
q = FluentSQLQuery()
q.select("id").from_table(obj["type"]._table_name).where("type=", type_name)
q()
res = q.fetch_one()
if res is None:
return False
type_id = res["id"]
q.cleanup()
q.update(obj._table_name, {"type_id": type_id}).where("id=", obj["id"])
q()
return q.count() == 1
else:
raise RRSDatabaseEntityError("Entity doesn't have a type to bind to.")
def __init__(self, config, importer_kwargs): #data ziskana z api self.rPublication = None #objekt typu RRSPublication, ktery po naplneni budeme importovat do db self.publication = None #nastaveni pro importer self.importer_kwargs = importer_kwargs #sleeper range self.LimitMin = 0.1 self.LimitMax = 0.2 #objekt pro vytvareni sql dotazu self.q = FluentSQLQuery() #researchr API self.researchrClass = ResearchrClass() #nejvyssi vrstva, pro nacteni objektu podle id self.rrsdb = RRSDatabase() #normalizator self.norm = Normalize() #importer self.importer = RRSXMLImporter(self.importer_kwargs)
def checkIfImport(key):
"""
Check if publication is in database.
@type name: string
@param name: Key of publication.
"""
q = FluentSQLQuery()
q.select("id").from_table("publication")
q.where("researchr_key=", key)
q()
data = q.fetch_one()
return data
def _bind_location_to_woeid(self, location):
"""
Binds location to woeid number.
@returns True if woeid found, False otherwise
"""
if type(location) is not RRSLocation:
raise TypeError("location has to be type RRSLocation")
if not 'woeid' in location:
q = FluentSQLQuery()
for attr in ("address", "city", "country", "name"):
if location[attr] is None: continue
q.cleanup()
q.select("woeid").from_table("geoplanet").where("name=", location[attr])
q()
res = q.fetch_one()
if res is None: continue
location['woeid'] = res[0]
return True
return False
return True
def _bind_location_to_woeid(self, location):
"""
Binds location to woeid number.
@returns True if woeid found, False otherwise
"""
if type(location) is not RRSLocation:
raise TypeError("location has to be type RRSLocation")
if not "woeid" in location:
q = FluentSQLQuery()
for attr in ("address", "city", "country", "name"):
if location[attr] is None:
continue
q.cleanup()
q.select("woeid").from_table("geoplanet").where("name=", location[attr])
q()
res = q.fetch_one()
if res is None:
continue
location["woeid"] = res[0]
return True
return False
return True
def lookup(self, obj, level=None):
"""
More sophisticated RRSDatabase.contains(). This method doesnt call
RRSDatabase.contains() explicitly, it checks other entities and tries
to find relationship between them. This method uses list of lookup rules.
@returns True if found (the object now carries the ID)
False if not found
"""
if level is None:
level = self.lookup_level
if level < 0:
return
if not isinstance(obj, _RRSDatabaseEntity):
raise TypeError('lookup() method can be called only on database '\
'entity objects.')
if obj._table_name.endswith("_meta"):
raise RRSDatabaseEntityError('lookup() method cannot be called on meta-tables.')
q = FluentSQLQuery()
# LEVEL 0 rules
try:
lvl_zero_rules = self._lookup_rules.get_rules(type(obj), 0)
except KeyError:
if self.logger is not None:
self.logger.error("Level 0 rules for '%s' not found." % obj._table_name)
return False
for rule in lvl_zero_rules:
attr_present = [item for item in rule.entities if item in obj]
# if there are no such attrubutes or not the requested count of them,
# continue to the next rule
if rule.reqcount > len(attr_present):
continue
self._db.refresh()
for cnt in reversed(range(rule.reqcount, len(attr_present)+1)):
for attr_comb in combinations(attr_present, cnt):
# now select them
q.cleanup()
q.select("id").from_table(obj._table_name)
for attr in attr_comb:
try:
q.where("%s=" % attr, obj[attr])
except FluentSQLQueryError:
q.and_("%s=" % attr, obj[attr])
q()
res = q.fetch_all()
if q.count() > 1: # there shouln't be more results than one
self.logger.warning("There are more than one identical "\
"%ss. List of ID's: %s" % (obj._table_name, str([x[0] for x in res])))
if not res or res is None:
continue
obj['id'] = res[0][0]
return True
# LEVEL 1 rules
try:
lvl_one_rules = self._lookup_rules.get_rules(type(obj), 1)
except KeyError:
if self.logger is not None:
self.logger.error("Level 1 rules for '%s' not found." % obj._table_name)
return False
# returns type of entity mapped in ent_id_map
def getetype(ent_id_map, ent):
for k in ent_id_map.keys():
e, et = k
if ent == e:
return et
# these are objects which really are present in the entity
for rule in lvl_one_rules:
ent_present = [item for item in rule.entities if item in obj]
# if there are no such entities or not the requested count of them,
# continue to the next rule
if rule.reqcount > len(ent_present):
continue
# get all those identifiers
ent_id_map = {}
for ent_name in ent_present:
target = obj[ent_name]
if type(target) is list and target:
# list of relationship objects
key = (ent_name, type(target[0]))
ent_id_map[key] = []
for rel_obj in target:
assert len(rel_obj.get_entities()) > 0
e = rel_obj.get_entities()[0]
if self.lookup(e, level-1):
if not key in ent_id_map:
ent_id_map[key] = []
ent_id_map[key].append(e)
if not ent_id_map[key]:
del ent_id_map[key]
elif isinstance(target, _RRSDatabaseEntity):
# this is FK - @target is RRS*** object
if self.lookup(target, level-1):
ent_id_map[(ent_name, type(target))] = [target]
else:
ent_id_map[(ent_name, type(target))] = [target]
# if we did not found as much as the rules requests, continue
if rule.reqcount > len(ent_id_map):
continue
# try to catch some data from the minimum count of requested entities
# to match, probably 2
# if this select spits out too many results (>100), the reqcount level 2
# is omitted and the process starts again from 3.
# There has to be a flag, which indicates, that the level 2
# requested entities returned too many results
next_reqcount_lvl = False
ent_keys = [x[0] for x in ent_id_map.keys()]
for cnt in range(rule.reqcount, len(ent_id_map)+1):
next_reqcount_lvl = False
for entity_comb in combinations(ent_keys, cnt):
if next_reqcount_lvl: break
self._db.refresh() # re-create cursors to drop the loaded data
# construct the query
q.cleanup()
tg_tbl = obj._table_name
from_lst = [tg_tbl]
q.select("%s.id" % tg_tbl)
# recognition of the same table in the query
tablecounter = 1
for ent in entity_comb:
etype = getetype(ent_id_map, ent)
# now we have key to the object -> ent_id_map[(ent, etype)]
# @ent is instance of RRS****** - 1:N relationship
# the object contains id of this entity
if issubclass(etype, _RRSDatabaseEntity):
o = ent_id_map[(ent, etype)][0]
try:
q.where("%s.%s_id=" % (tg_tbl, ent), o['id'])
except FluentSQLQueryError:
q.and_("%s.%s_id=" % (tg_tbl, ent), o['id'])
# @ent is fake junction table - it means, that it's
# the second side of 1:N relationship - N:1.
elif issubclass(etype, _RRSDbEntityRelationship) and etype._fake_table:
# TODO
return False
# @ent is true junction table - this M:N relationship.
elif issubclass(etype, _RRSDbEntityRelationship) and not etype._fake_table:
j_tbl_uniq_as = None
# storage of all acronyms iof junction tables
j_tbl_uniq_as_list = []
o = None
# join together all the found entities - for example:
# given publication, two persons (authors), both found
# in db so create query which selects ID of publication
# which has both - the first AND the second person.
for o in ent_id_map[(ent, etype)]:
j_tbl_uniq_as = "%s%s" % (etype._table_name, tablecounter)
j_tbl_uniq_as_list.append(j_tbl_uniq_as)
e_tbl_uniq_as = "%s%s" % (o._table_name, tablecounter)
# add table to the list of tables we are joining together
from_lst.append("%s AS %s" % (etype._table_name, j_tbl_uniq_as))
from_lst.append("%s AS %s" % (o._table_name, e_tbl_uniq_as))
try:
q.where("%s.id=" % e_tbl_uniq_as, o['id'])
except FluentSQLQueryError:
q.and_("%s.id=" % e_tbl_uniq_as, o['id'])
q.and_("%s.%s_id=" % (j_tbl_uniq_as, o._table_name), "%s.id" % e_tbl_uniq_as, True)
tablecounter += 1
# add the condition that all the junction table ID's of
# the entity we are looking for has to be the same - we
# are looking not for union, but intersection of them
for i in range(0, len(j_tbl_uniq_as_list)):
try:
j1 = j_tbl_uniq_as_list[i]
j2 = j_tbl_uniq_as_list[i+1]
q.and_("%s.%s_id=" % (j1, tg_tbl), "%s.%s_id" % (j2, tg_tbl), True)
except IndexError:
break
# bind junction table.entity_id to id of entity we are looking for
q.and_("%s.%s_id=" % (j_tbl_uniq_as, tg_tbl), "%s.id" % tg_tbl, True)
# @ent is attribute (int, basestring ect.)
else:
attr = ent_id_map[(ent, etype)][0]
try:
q.where("%s.%s=" % (tg_tbl, ent), attr)
except FluentSQLQueryError:
q.and_("%s.%s=" % (tg_tbl, ent), attr)
q.from_table(from_lst)
q()
search_sql_query = q._sql
# now if the total count of probably identical files is higher
# than 100, we need to specify it more, so we jump to next
# request count level (probably 1->2 or 2->3).
if q.count() > 100:
next_reqcount_lvl = True
continue
res = q.fetch_all()
if not res:
continue
elif len(res) == 1:
obj['id'] = res[0][0]
self.logger.info("Found exactly one result for lookup: %s, params: %s, found ID: %s, SQL: %s" % \
(obj._table_name, str(entity_comb), obj['id'], search_sql_query))
return True
else:
# do some magic stuff here
# intelligenty compare the attribute of all returned results
# and choose the most similar
q.cleanup()
id_list = [x[0] for x in res]
attrunion = set(["id"])
lvl_zero_rules = self._lookup_rules.get_rules(type(obj), 0)
# make a list of attributes needed to acomplish the rules
# (these are all which are present in rules)
for rule in lvl_zero_rules:
attrunion = attrunion.union(set(rule.entities))
# construct query which loads all needed attributes of all returned ID's
q.select(list(attrunion)).from_table(obj._table_name)
for _id in id_list:
try:
q.where("id=", _id)
except FluentSQLQueryError:
q.or_("id=", _id)
q() # perform the query
loaded_data = q.fetch_all()
similarity = {}
# every rule tell us what attributes have to be similar
# (or identical)
for rule in lvl_zero_rules:
attrs = [item for item in rule.entities if item in obj]
if rule.reqcount > len(attrs):
continue
# count every row's similarity (the result is sum of
# similarities of their attributes)
sim_lst = {}
for d in loaded_data:
row_similarity = 0.0
for attr in attrs:
if attr not in d or d[attr] is None or attr not in obj:
continue
if (d['id'], attr) not in similarity:
s = SequenceMatcher(None, d[attr], obj[attr])
similarity[(d['id'], attr)] = s.ratio()
row_similarity += similarity[(d['id'], attr)]
sim_lst[row_similarity] = d['id']
# get the most similar row to the object
obj['id'] = sim_lst[max(sim_lst.keys())]
self.logger.info("Found more than one result for lookup: %s, params: %s, "\
"Choosen ID: %s, SQL: %s" % (obj._table_name, str(entity_comb), obj['id'], search_sql_query))
return True
def _bind_entity_to_name(self, namedentity, source_module):
"""
This method creates connection between entity and it's name, which is stored
in other database table. These tables are:
- person vs person_name
- organization vs organization_name
- event vs event_name
@returns ID of the row in the name-table.
"""
ACRONYM = 'acronym' # to be easily changed to abbreviation or whatever needed..
TITLE = 'title' # will be name? or what?
if not isinstance(namedentity, _RRSDatabaseEntity):
raise TypeError("Named object has to be instance of subclass of _RRSDatabaseEntity")
if not 'id' in namedentity:
raise DatabaseError("Named object has to contain ID!")
q = FluentSQLQuery()
if namedentity._table_name == "person":
# act like person and handle person_name
# this is slightly different because there is N:N relationship
# create new person name object
pname = RRSPerson_name()
for attr in ('first_name', 'middle_name', 'last_name', 'full_name'):
if attr in namedentity:
pname[attr] = namedentity[attr]
# create relationship object
rel_obj = RRSRelationshipPersonPerson_name()
rel_obj.set_entity(pname)
namedentity['person_name'] = rel_obj
# look for this name in database
if self.lookup(pname):
# it is in db yet, just check if rel exists
q.select("person_id").from_table(("j__person__person_name"))
q.where("person_id=", namedentity['id']).and_("person_name_id=", pname['id'])
q()
if not q.count():
# if the relationship doesn't exist, create new one
self._rrsdb.relationship("person_name", rel_obj)
elif q.count() > 1:
self.logger.warning("There are more than one relationship "\
"entries in table 'j__person__person_name"\
" between person.id=%s and person_name.id=%s" \
% (namedentity['id'], pname['id']))
else:
# insert new person_name and create the relationship
self._rrsdb.insert(pname, self.module)
self._rrsdb.relationship("person_name", rel_obj)
# get the reference out of which is this name extracted and assign
# the person name to the reference (j__person_name__reference)
try:
refe = namedentity['publication'][0].get_entities()[0]['reference_reference'][0].get_entities()[0]
except (KeyError, TypeError, IndexError):
pass
else:
pname_ref_rel = RRSRelationshipPerson_nameReference()
pname_ref_rel.set_entity(refe)
pname['reference'] = pname_ref_rel
try:
self._rrsdb.relationship('reference', pname_ref_rel)
except DatabaseError:
self._queue.wait(WQEntry(self._rrsdb.relationship, ('reference', pname_ref_rel)))
elif namedentity._table_name in ("event", "organization"):
if TITLE not in namedentity:
# this violates constraint... raise exception?? Or return false?
return False
name_tbl = "%s_name" % namedentity._table_name
# if there in the database is no title like this, insert it
q.select(("id", "%s_id" % namedentity._table_name, ACRONYM, TITLE)).from_table(name_tbl)
q.where("%s=" % TITLE, namedentity[TITLE])
if ACRONYM in namedentity:
q.or_("%s=" % ACRONYM, namedentity[ACRONYM])
q()
if q.count():
# check the parent id if it matches
for row in q.fetch_all():
if namedentity['id'] == row[1]:
# if it matched on acronym, check the titles if they are the same
if row[TITLE] != namedentity[TITLE]:
# if not, check the rest and maybe add new row into table
continue
# add the missing acronym if needed
if row[ACRONYM] is None and namedentity[ACRONYM] is not None:
# update the row
q.cleanup()
q.update(name_tbl, {ACRONYM: namedentity[ACRONYM]})
q.where("id=", row['id'])
q()
return row['id']
# if nothing matched, insert new name
name_obj = self._table_to_class_map[name_tbl]()
for attr in (TITLE, ACRONYM):
if attr in namedentity:
name_obj[attr] = namedentity[attr]
if name_obj.empty():
return False
name_obj[namedentity._table_name] = namedentity
self._rrsdb.insert(name_obj, source_module)
return name_obj['id']
else:
raise RRSDatabaseEntityError("%s is not a named entity." % type(namedentity))
self._db.refresh()
def lookup(self, obj, level=None):
"""
More sophisticated RRSDatabase.contains(). This method doesnt call
RRSDatabase.contains() explicitly, it checks other entities and tries
to find relationship between them. This method uses list of lookup rules.
@returns True if found (the object now carries the ID)
False if not found
"""
if level is None:
level = self.lookup_level
if level < 0:
return
if not isinstance(obj, _RRSDatabaseEntity):
raise TypeError("lookup() method can be called only on database " "entity objects.")
if obj._table_name.endswith("_meta"):
raise RRSDatabaseEntityError("lookup() method cannot be called on meta-tables.")
q = FluentSQLQuery()
# LEVEL 0 rules
try:
lvl_zero_rules = self._lookup_rules.get_rules(type(obj), 0)
except KeyError:
if self.logger is not None:
self.logger.error("Level 0 rules for '%s' not found." % obj._table_name)
return False
for rule in lvl_zero_rules:
attr_present = [item for item in rule.entities if item in obj]
# if there are no such attrubutes or not the requested count of them,
# continue to the next rule
if rule.reqcount > len(attr_present):
continue
self._db.refresh()
for cnt in reversed(range(rule.reqcount, len(attr_present) + 1)):
for attr_comb in combinations(attr_present, cnt):
# now select them
q.cleanup()
q.select("id").from_table(obj._table_name)
for attr in attr_comb:
try:
q.where("%s=" % attr, obj[attr])
except FluentSQLQueryError:
q.and_("%s=" % attr, obj[attr])
q()
res = q.fetch_all()
if q.count() > 1: # there shouln't be more results than one
self.logger.warning(
"There are more than one identical "
"%ss. List of ID's: %s" % (obj._table_name, str([x[0] for x in res]))
)
if not res or res is None:
continue
obj["id"] = res[0][0]
return True
# LEVEL 1 rules
try:
lvl_one_rules = self._lookup_rules.get_rules(type(obj), 1)
except KeyError:
if self.logger is not None:
self.logger.error("Level 1 rules for '%s' not found." % obj._table_name)
return False
# returns type of entity mapped in ent_id_map
def getetype(ent_id_map, ent):
for k in ent_id_map.keys():
e, et = k
if ent == e:
return et
# these are objects which really are present in the entity
for rule in lvl_one_rules:
ent_present = [item for item in rule.entities if item in obj]
# if there are no such entities or not the requested count of them,
# continue to the next rule
if rule.reqcount > len(ent_present):
continue
# get all those identifiers
ent_id_map = {}
for ent_name in ent_present:
target = obj[ent_name]
if type(target) is list and target:
# list of relationship objects
key = (ent_name, type(target[0]))
ent_id_map[key] = []
for rel_obj in target:
assert len(rel_obj.get_entities()) > 0
e = rel_obj.get_entities()[0]
if self.lookup(e, level - 1):
if not key in ent_id_map:
ent_id_map[key] = []
ent_id_map[key].append(e)
if not ent_id_map[key]:
del ent_id_map[key]
elif isinstance(target, _RRSDatabaseEntity):
# this is FK - @target is RRS*** object
if self.lookup(target, level - 1):
ent_id_map[(ent_name, type(target))] = [target]
else:
ent_id_map[(ent_name, type(target))] = [target]
# if we did not found as much as the rules requests, continue
if rule.reqcount > len(ent_id_map):
continue
# try to catch some data from the minimum count of requested entities
# to match, probably 2
# if this select spits out too many results (>100), the reqcount level 2
# is omitted and the process starts again from 3.
# There has to be a flag, which indicates, that the level 2
# requested entities returned too many results
next_reqcount_lvl = False
ent_keys = [x[0] for x in ent_id_map.keys()]
for cnt in range(rule.reqcount, len(ent_id_map) + 1):
next_reqcount_lvl = False
for entity_comb in combinations(ent_keys, cnt):
if next_reqcount_lvl:
break
self._db.refresh() # re-create cursors to drop the loaded data
# construct the query
q.cleanup()
tg_tbl = obj._table_name
from_lst = [tg_tbl]
q.select("%s.id" % tg_tbl)
# recognition of the same table in the query
tablecounter = 1
for ent in entity_comb:
etype = getetype(ent_id_map, ent)
# now we have key to the object -> ent_id_map[(ent, etype)]
# @ent is instance of RRS****** - 1:N relationship
# the object contains id of this entity
if issubclass(etype, _RRSDatabaseEntity):
o = ent_id_map[(ent, etype)][0]
try:
q.where("%s.%s_id=" % (tg_tbl, ent), o["id"])
except FluentSQLQueryError:
q.and_("%s.%s_id=" % (tg_tbl, ent), o["id"])
# @ent is fake junction table - it means, that it's
# the second side of 1:N relationship - N:1.
elif issubclass(etype, _RRSDbEntityRelationship) and etype._fake_table:
# TODO
return False
# @ent is true junction table - this M:N relationship.
elif issubclass(etype, _RRSDbEntityRelationship) and not etype._fake_table:
j_tbl_uniq_as = None
# storage of all acronyms iof junction tables
j_tbl_uniq_as_list = []
o = None
# join together all the found entities - for example:
# given publication, two persons (authors), both found
# in db so create query which selects ID of publication
# which has both - the first AND the second person.
for o in ent_id_map[(ent, etype)]:
j_tbl_uniq_as = "%s%s" % (etype._table_name, tablecounter)
j_tbl_uniq_as_list.append(j_tbl_uniq_as)
e_tbl_uniq_as = "%s%s" % (o._table_name, tablecounter)
# add table to the list of tables we are joining together
from_lst.append("%s AS %s" % (etype._table_name, j_tbl_uniq_as))
from_lst.append("%s AS %s" % (o._table_name, e_tbl_uniq_as))
try:
q.where("%s.id=" % e_tbl_uniq_as, o["id"])
except FluentSQLQueryError:
q.and_("%s.id=" % e_tbl_uniq_as, o["id"])
q.and_("%s.%s_id=" % (j_tbl_uniq_as, o._table_name), "%s.id" % e_tbl_uniq_as, True)
tablecounter += 1
# add the condition that all the junction table ID's of
# the entity we are looking for has to be the same - we
# are looking not for union, but intersection of them
for i in range(0, len(j_tbl_uniq_as_list)):
try:
j1 = j_tbl_uniq_as_list[i]
j2 = j_tbl_uniq_as_list[i + 1]
q.and_("%s.%s_id=" % (j1, tg_tbl), "%s.%s_id" % (j2, tg_tbl), True)
except IndexError:
break
# bind junction table.entity_id to id of entity we are looking for
q.and_("%s.%s_id=" % (j_tbl_uniq_as, tg_tbl), "%s.id" % tg_tbl, True)
# @ent is attribute (int, basestring ect.)
else:
attr = ent_id_map[(ent, etype)][0]
try:
q.where("%s.%s=" % (tg_tbl, ent), attr)
except FluentSQLQueryError:
q.and_("%s.%s=" % (tg_tbl, ent), attr)
q.from_table(from_lst)
q()
search_sql_query = q._sql
# now if the total count of probably identical files is higher
# than 100, we need to specify it more, so we jump to next
# request count level (probably 1->2 or 2->3).
if q.count() > 100:
next_reqcount_lvl = True
continue
res = q.fetch_all()
if not res:
continue
elif len(res) == 1:
obj["id"] = res[0][0]
self.logger.info(
"Found exactly one result for lookup: %s, params: %s, found ID: %s, SQL: %s"
% (obj._table_name, str(entity_comb), obj["id"], search_sql_query)
)
return True
else:
# do some magic stuff here
# intelligenty compare the attribute of all returned results
# and choose the most similar
q.cleanup()
id_list = [x[0] for x in res]
attrunion = set(["id"])
lvl_zero_rules = self._lookup_rules.get_rules(type(obj), 0)
# make a list of attributes needed to acomplish the rules
# (these are all which are present in rules)
for rule in lvl_zero_rules:
attrunion = attrunion.union(set(rule.entities))
# construct query which loads all needed attributes of all returned ID's
q.select(list(attrunion)).from_table(obj._table_name)
for _id in id_list:
try:
q.where("id=", _id)
except FluentSQLQueryError:
q.or_("id=", _id)
q() # perform the query
loaded_data = q.fetch_all()
similarity = {}
# every rule tell us what attributes have to be similar
# (or identical)
for rule in lvl_zero_rules:
attrs = [item for item in rule.entities if item in obj]
if rule.reqcount > len(attrs):
continue
# count every row's similarity (the result is sum of
# similarities of their attributes)
sim_lst = {}
for d in loaded_data:
row_similarity = 0.0
for attr in attrs:
if attr not in d or d[attr] is None or attr not in obj:
continue
if (d["id"], attr) not in similarity:
s = SequenceMatcher(None, d[attr], obj[attr])
similarity[(d["id"], attr)] = s.ratio()
row_similarity += similarity[(d["id"], attr)]
sim_lst[row_similarity] = d["id"]
# get the most similar row to the object
obj["id"] = sim_lst[max(sim_lst.keys())]
self.logger.info(
"Found more than one result for lookup: %s, params: %s, "
"Choosen ID: %s, SQL: %s"
% (obj._table_name, str(entity_comb), obj["id"], search_sql_query)
)
return True
def _bind_entity_to_name(self, namedentity, source_module):
"""
This method creates connection between entity and it's name, which is stored
in other database table. These tables are:
- person vs person_name
- organization vs organization_name
- event vs event_name
@returns ID of the row in the name-table.
"""
ACRONYM = "acronym" # to be easily changed to abbreviation or whatever needed..
TITLE = "title" # will be name? or what?
if not isinstance(namedentity, _RRSDatabaseEntity):
raise TypeError("Named object has to be instance of subclass of _RRSDatabaseEntity")
if not "id" in namedentity:
raise DatabaseError("Named object has to contain ID!")
q = FluentSQLQuery()
if namedentity._table_name == "person":
# act like person and handle person_name
# this is slightly different because there is N:N relationship
# create new person name object
pname = RRSPerson_name()
for attr in ("first_name", "middle_name", "last_name", "full_name"):
if attr in namedentity:
pname[attr] = namedentity[attr]
# create relationship object
rel_obj = RRSRelationshipPersonPerson_name()
rel_obj.set_entity(pname)
namedentity["person_name"] = rel_obj
# look for this name in database
if self.lookup(pname):
# it is in db yet, just check if rel exists
q.select("person_id").from_table(("j__person__person_name"))
q.where("person_id=", namedentity["id"]).and_("person_name_id=", pname["id"])
q()
if not q.count():
# if the relationship doesn't exist, create new one
self._rrsdb.relationship("person_name", rel_obj)
elif q.count() > 1:
self.logger.warning(
"There are more than one relationship "
"entries in table 'j__person__person_name"
" between person.id=%s and person_name.id=%s" % (namedentity["id"], pname["id"])
)
else:
# insert new person_name and create the relationship
self._rrsdb.insert(pname, self.module)
self._rrsdb.relationship("person_name", rel_obj)
# get the reference out of which is this name extracted and assign
# the person name to the reference (j__person_name__reference)
try:
refe = namedentity["publication"][0].get_entities()[0]["reference_reference"][0].get_entities()[0]
except (KeyError, TypeError, IndexError):
pass
else:
pname_ref_rel = RRSRelationshipPerson_nameReference()
pname_ref_rel.set_entity(refe)
pname["reference"] = pname_ref_rel
try:
self._rrsdb.relationship("reference", pname_ref_rel)
except DatabaseError:
self._queue.wait(WQEntry(self._rrsdb.relationship, ("reference", pname_ref_rel)))
elif namedentity._table_name in ("event", "organization"):
if TITLE not in namedentity:
# this violates constraint... raise exception?? Or return false?
return False
name_tbl = "%s_name" % namedentity._table_name
# if there in the database is no title like this, insert it
q.select(("id", "%s_id" % namedentity._table_name, ACRONYM, TITLE)).from_table(name_tbl)
q.where("%s=" % TITLE, namedentity[TITLE])
if ACRONYM in namedentity:
q.or_("%s=" % ACRONYM, namedentity[ACRONYM])
q()
if q.count():
# check the parent id if it matches
for row in q.fetch_all():
if namedentity["id"] == row[1]:
# if it matched on acronym, check the titles if they are the same
if row[TITLE] != namedentity[TITLE]:
# if not, check the rest and maybe add new row into table
continue
# add the missing acronym if needed
if row[ACRONYM] is None and namedentity[ACRONYM] is not None:
# update the row
q.cleanup()
q.update(name_tbl, {ACRONYM: namedentity[ACRONYM]})
q.where("id=", row["id"])
q()
return row["id"]
# if nothing matched, insert new name
name_obj = self._table_to_class_map[name_tbl]()
for attr in (TITLE, ACRONYM):
if attr in namedentity:
name_obj[attr] = namedentity[attr]
if name_obj.empty():
return False
name_obj[namedentity._table_name] = namedentity
self._rrsdb.insert(name_obj, source_module)
return name_obj["id"]
else:
raise RRSDatabaseEntityError("%s is not a named entity." % type(namedentity))
self._db.refresh()
class ResearchrPublicationFeeder:
def __init__(self, config, importer_kwargs):
#data ziskana z api
self.rPublication = None
#objekt typu RRSPublication, ktery po naplneni budeme importovat do db
self.publication = None
#nastaveni pro importer
self.importer_kwargs = importer_kwargs
#sleeper range
self.LimitMin = 0.1
self.LimitMax = 0.2
#objekt pro vytvareni sql dotazu
self.q = FluentSQLQuery()
#researchr API
self.researchrClass = ResearchrClass()
#nejvyssi vrstva, pro nacteni objektu podle id
self.rrsdb = RRSDatabase()
#normalizator
self.norm = Normalize()
#importer
self.importer = RRSXMLImporter(self.importer_kwargs)
def __FillType(self):
"""
Transform rPublication.type to publication.type
"""
_id = self.__GetId("publication_type", "type=", self.rPublication.publication_type)
if (_id != None):
self.publication["type"] = self.rrsdb.load("publication_type", _id)
def __FillSeries(self):
"""
Add rPublication.series to publication_series table
"""
if (self.rPublication.series != None and self.rPublication.series != ""):
_id = None
while (_id == None):
_id = self.__GetId("publication_series", "title=", self.rPublication.series)
if (_id == None):
series = RRSPublication_series(title=self.rPublication.series)
#importer = RRSXMLImporter(self.importer_kwargs)
self.importer.import_model(series)
continue
self.publication["series"] = self.rrsdb.load("publication_series", _id)
def __GetId(self, _from, where, _is):
"""
Try to find ID in table and return it
@type _from: string
@param _from: Name of table.
@type where: string
@param where: Name of column.
@type _is: string
@param _is: What it is equal.
@rtype: int
@return: Id of selected entry.
"""
self.q.select("id").from_table(_from)
self.q.where(where, _is)
self.q()
data = self.q.fetch_one()
#print(self.q.sql())
self.q.cleanup()
if data != None:
return data[0]
return None
def __FillPublisher(self):
"""
Add rPublication.publisher to organization table
"""
if (self.rPublication.publisher != None and self.rPublication.publisher != ""):
_id = None
normalized_title = self.norm.organization(self.rPublication.publisher)
while (_id == None):
_id = self.__GetId("organization", "title_normalized=", normalized_title)
if (_id == None):
organization = RRSOrganization(title=self.rPublication.publisher,
title_normalized=normalized_title)
#importer = RRSXMLImporter(self.importer_kwargs)
self.importer.import_model(organization)
continue
self.publication["publisher"] = self.rrsdb.load("organization", _id)
def __FillAuthors(self, authorData, isEditor):
"""
FillAuthor Add (if there are not) person to db and
contain them with actual publication. Foreach
rPublication.authors, take only person's url and fullname.
@type authorData: list
@param authorData: List of authors data (person, alias)
@type isEditor: bool
@param isEditor: True if authors are editors of this publication.
"""
if (len(authorData) != 0):
rank = 0
for author in authorData:
if 'author' in author:
rFullname = author["person"]["fullname"]
rUrl = author["person"]["url"]
else:
rFullname = author["alias"]["name"]
rUrl = author["alias"]["url"]
personUrl = RRSRelationshipPersonUrl()
rank += 1
self.__FillUrl(personUrl, rUrl)
self.__FillPerson(personUrl, rFullname, rank, isEditor)
def __FillUrl(self, personUrl, rUrl):
"""
This function add url to db bind url to person
@type personUrl: RRSRelationshipPersonUrl
@param personUrl: Relationship object to add url into it.
@type rUrl: string
@param isEditor: rPublication.(person/alias) url, url of author/editor.
"""
_id = None
while (_id == None):
_id = self.__GetId("url", "link=", rUrl)
if (_id == None):
url = RRSUrl(link=rUrl)
url["type"] = self.rrsdb.load("url_type", "1")
#importer = RRSXMLImporter(self.importer_kwargs)
self.importer.import_model(url)
continue
url = self.rrsdb.load("url", _id)
personUrl.set_entity(url)
#print( personUrl)
def __FillPerson(self, personUrl, rFullname, rank, isEditor):
"""
This function try fill first name, middle name, last name of person.
@type personUrl: RRSRelationshipPersonUrl
@param personUrl: Relationship object to bind to person["url"].
@type rFullname: string
@param rFullname: Fullname of author.
@type rank: int
@param rank: Rank of author, first author get 1, second 2 and so on.
@type isEditor: bool
@param isEditor: True if person is editor of this publication.
"""
_id = None
while (_id == None):
_id = self.__GetId("person", "full_name=", rFullname)
if (_id == None):
person = RRSPerson()
person["full_name"] = rFullname
person["url"] = personUrl
self.__SetPersonNames(person, rFullname)
person["full_name_ascii"] = unicodedata.normalize('NFKD', rFullname).encode('ascii', 'ignore')
#importer = RRSXMLImporter(self.importer_kwargs)
#print(person)
self.importer.import_model(person)
continue
publicationPerson = RRSRelationshipPersonPublication(author_rank=rank, editor=isEditor)
publicationPerson.set_entity(self.rrsdb.load("person", _id))
#print(publicationPerson)
self.publication['person'] = publicationPerson
def __SetPersonNames(self, person, rFullname):
"""
This function try fill first name, middle name, last name of person.
@type person: RRSPerson
@param person: Object of author of publication.
@type rFullname: string
@param rFullname: Fullname of author.
"""
splitName = rFullname.split()
if (len(splitName) == 3):
person["first_name"] = splitName[0]
person["middle_name"] = splitName[1]
person["last_name"] = splitName[2]
elif (len(splitName) == 2):
person["first_name"] = splitName[0]
person["last_name"] = splitName[1]
def FillPublication(self, key):
"""
This function call all private function with prefix Fill,
this function load data to rPublication structure and then
assign data from rPublication to publication(RRSPublication).
@type key: string
@param key: Key of the publication.
"""
self.__FillRPublication(key)
self.publication = RRSPublication()
self.__FillAuthors(self.rPublication.authors, False)
self.__FillAuthors(self.rPublication.editors, True)
self.__FillPublisher()
self.__FillType()
self.__FillSeries()
self.publication["title"] = self.rPublication.title
self.publication["title_normalized"] = self.norm.publication(self.rPublication.title)
if (self.rPublication.year != None and self.rPublication.year != ""):
self.publication["year"] = int(self.rPublication.year) # "2000" -> 2000
if (self.rPublication.month != None and self.rPublication.month != ""):
self.publication["month"] = int(strptime(self.rPublication.month[:3],'%b').tm_mon)
if (self.rPublication.volume != None and self.rPublication.volume != "" and self.rPublication.volume.isdigit()):
self.publication["volume"] = int(self.rPublication.volume)
if (self.rPublication.number != None and self.rPublication.number != "" and self.rPublication.volume.isdigit()):
self.publication["number"] = int(self.rPublication.number)
if (self.rPublication.abstract != None and self.rPublication.abstract != ""):
self.publication["abstract"] = self.rPublication.abstract
if (self.rPublication.doi != None and "http://dx.doi.org/" in self.rPublication.doi):
self.publication["doi"] = self.rPublication.doi.strip('http://dx.doi.org/')
if (self.rPublication.firstpage != None and self.rPublication.lastpage != None and
self.rPublication.firstpage != "" and self.rPublication.lastpage != ""):
self.publication["pages"] = str(self.rPublication.firstpage) + " - " + str(self.rPublication.lastpage)
self.publication["language"] = self.rrsdb.load('language', 1)
self.publication.set("researchr_key", self.rPublication.key, strict=False)
#print(self.publication)
#importer = RRSXMLImporter(self.importer_kwargs)
try:
self.importer.import_model(self.publication)
except RRSDatabaseEntityError as e:
print('RRSDatabaseEntityError - %s, %s' % (self.rPublication.key, str(e)))
logging.warning('RRSDatabaseEntityError - %s, %s' % (self.rPublication.key, str(e)))
except DatabaseError as e:
print('DatabaseError - %s, %s' % (self.rPublication.key, str(e)))
logging.warning('DatabaseError - %s, %s' % (self.rPublication.key, str(e)))
except TypeError as e:
print('TypeError - %s, %s' % (self.rPublication.key, str(e)))
logging.warning('TypeError - %s, %s' % (self.rPublication.key, str(e)))
except:
print('Unexpected error - %s, %s' % (self.rPublication.key, sys.exc_info()[0]))
logging.warning('Unexpected error - %s, %s' % (self.rPublication.key, sys.exc_info()[0]))
def __FillRPublication(self, key):
"""
Fill rPublication object.
@type key: string
@param key: Name od publication.
"""
self.rPublication = RPublication()
publicationData = self.researchrClass.getPublication(key)
time.sleep(random.uniform(self.LimitMin, self.LimitMax))
#print(publicationData)
for key, value in publicationData.items():
if key == 'abstract':
self.rPublication.abstract = value
elif key == 'address':
self.rPublication.address = value
elif key == 'authors':
self.rPublication.authors = value
elif key == 'booktitle':
self.rPublication.booktitle = value
elif key == 'conference':
self.rPublication.conference = value
elif key == 'conferenceYear':
self.rPublication.conferenceYear = value
elif key == 'doi':
self.rPublication.doi = value
elif key == 'editors':
self.rPublication.editors = value
elif key == 'firstpage':
self.rPublication.firstpage = value
elif key == 'key':
self.rPublication.key = value
elif key == 'issuenumber':
self.rPublication.issuenumber = value
elif key == 'journal':
self.rPublication.journal = value
elif key == 'key':
self.rPublication.key = value
elif key == 'lastpage':
self.rPublication.lastpage = value
elif key == 'month':
self.rPublication.month = value
elif key == 'note':
self.rPublication.note = value
elif key == 'number':
self.rPublication.number = value
elif key == 'organization':
self.rPublication.organization = value
elif key == 'publisher':
self.rPublication.publisher = value
elif key == 'series':
self.rPublication.series = value
elif key == 'title':
self.rPublication.title = value
elif key == 'type':
self.rPublication.publication_type = value
elif key == 'url':
self.rPublication.url = value
elif key == 'volume':
self.rPublication.volume = value
elif key == 'volumenumber':
self.rPublication.volumenumber = value
elif key == 'year':
self.rPublication.year = value
