def test_Pred_cmp_self(self):
"""
All Pred instances should be equal.
"""
p1 = Pred()
p2 = Pred()
self.assertEqual(p1, p2)
self.assertLessEqual(p1, p2)
self.assertGreaterEqual(p1, p2)
self.assertFalse(p1 < p2)
self.assertFalse(p1 > p2)
self.assertFalse(p1 != p2)
def test_Pred_from_normalised_string(self):
"""
Pred.from_normalised_string should instantiate RealPreds or GPreds
depending on whether there is a leading underscore
"""
# Check the preds are of the right type
cat_pred = Pred.from_normalised_string('_cat_n_1')
the_pred = Pred.from_normalised_string('the_q')
self.assertIsInstance(cat_pred, RealPred)
self.assertIsInstance(the_pred, GPred)
# Check the preds are the equivalent to initialising directly
cat_realpred = RealPred.from_normalised_string('_cat_n_1')
the_gpred = GPred.from_normalised_string('the_q')
self.assertEqual(cat_pred, cat_realpred)
self.assertEqual(the_pred, the_gpred)
def test_Pred_new(self):
"""
Pred() instances denote underpecified preds,
and should not take any arguments
"""
with self.assertRaises(TypeError):
Pred('name')
def test_Pred_cmp_subclasses(self):
"""
Any Pred instance should be less than instances of subclasses.
"""
p = Pred()
cat = RealPred('cat', 'n', '1')
pron = GPred('pron')
self.assertEqual(p, p)
self.assertEqual(pron, pron)
self.assertEqual(cat, cat)
self.assertNotEqual(p, pron)
self.assertNotEqual(p, cat)
self.assertNotEqual(pron, cat)
self.assertLess(p, pron)
self.assertLess(p, cat)
self.assertLess(pron, cat)
self.assertLessEqual(p, pron)
self.assertLessEqual(p, cat)
self.assertLessEqual(pron, cat)
self.assertGreater(pron, p)
self.assertGreater(cat, p)
self.assertGreater(cat, pron)
self.assertGreaterEqual(pron, p)
self.assertGreaterEqual(cat, p)
self.assertGreaterEqual(cat, pron)
def test_Pred_from_normalised_string(self):
"""
Pred.from_normalised_string should instantiate RealPreds or GPreds
depending on whether there is a leading underscore
"""
# Check the preds are of the right type
cat_pred = Pred.from_normalised_string('_cat_n_1')
the_pred = Pred.from_normalised_string('the_q')
self.assertIsInstance(cat_pred, RealPred)
self.assertIsInstance(the_pred, GPred)
# Check the preds are the equivalent to initialising directly
cat_realpred = RealPred.from_normalised_string('_cat_n_1')
the_gpred = GPred.from_normalised_string('the_q')
self.assertEqual(cat_pred, cat_realpred)
self.assertEqual(the_pred, the_gpred)
def _parse_pred(string, nodeid, queries, equalities):
assert string.islower(), 'Predicates must be lower-case.'
assert ' ' not in string, 'Predicates must not contain spaces.'
if string[0] == '"' and string[-1] == '"':
string = string[1:-1]
assert '"' not in string, 'Predicates must not contain quotes.'
assert string[
0] != '\'', 'Predicates with opening single-quote have been deprecated.'
if (string[:4] == 'pred' and
(len(string) == 4 or string[4] in special_values)) or (
string[:8] == 'predsort' and
(len(string) == 8 or string[8] in special_values)):
i = 8 if string[:8] == 'predsort' else 4
value = _parse_value(
string[i:], None, queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]].pred))
assert not value
return Pred(), string[:i]
rel_suffix = ''
if string[-4:] == '_rel':
string = string[:-4]
rel_suffix = '_rel'
if string[0] != '_':
name = _parse_value(
string, '?', queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]].pred.name))
return GPred(name), name + rel_suffix
values = string[1:].rsplit('_', 2)
count = len(values)
assert count > 0, 'Invalid number of arguments for RealPred.'
if count == 1:
values.insert(0, '?')
values.append('unknown')
elif count == 2:
values.append(None)
lemma = _parse_value(
values[0], '?', queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]].pred.lemma))
pos = _parse_value(
values[1], 'u', queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]].pred.pos)) # u ???
sense = _parse_value(
values[2], 'unknown', queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]].pred.sense
)) # unknown ???
if count == 1:
ref_name = '_{}{}'.format(pos, rel_suffix)
elif count == 2:
ref_name = '_{}_{}{}'.format(lemma, pos, rel_suffix)
else:
ref_name = '_{}_{}_{}{}'.format(lemma, pos, sense, rel_suffix)
return RealPred(lemma, pos, sense), ref_name
def test_Pred_from_string(self):
"""
Pred.from_string should normalise the string as necessary
"""
cat_pred = RealPred.from_normalised_string('_cat_n_1')
self.assertEqual(Pred.from_string('_cat_n_1_rel'), cat_pred)
self.assertEqual(Pred.from_string('"_cat_n_1_rel"'), cat_pred)
self.assertEqual(Pred.from_string('_CAT_N_1_REL'), cat_pred)
the_pred = GPred.from_normalised_string('the')
self.assertEqual(Pred.from_string('the_rel'), the_pred)
self.assertEqual(Pred.from_string('"the_rel"'), the_pred)
self.assertEqual(Pred.from_string('THE_REL'), the_pred)
def test_Pred_normalise_string(self):
"""
Pred strings should be normalised - see comments below
"""
# Remove quotes from quoted preds
self.assertEqual(Pred.normalise_string('"pron"'), 'pron')
with self.assertRaises(DeprecationWarning):
warnings.simplefilter('error')
self.assertEqual(Pred.normalise_string("'pron"), "pron")
warnings.resetwarnings()
# No internal spaces or quotes
with self.assertRaises(ValueError):
Pred.normalise_string('pred name')
with self.assertRaises(ValueError):
Pred.normalise_string('pred"name')
# Force lower case
self.assertEqual(Pred.normalise_string('PRON'), 'pron')
# Strip trailing _rel
self.assertEqual(Pred.normalise_string('pron_rel'), 'pron')
def test_Pred_normalise_string(self):
"""
Pred strings should be normalised - see comments below
"""
# Remove quotes from quoted preds
self.assertEqual(Pred.normalise_string('"pron"'), 'pron')
with self.assertRaises(DeprecationWarning):
warnings.simplefilter('error')
self.assertEqual(Pred.normalise_string("'pron"), "pron")
warnings.resetwarnings()
# No internal spaces or quotes
with self.assertRaises(ValueError):
Pred.normalise_string('pred name')
with self.assertRaises(ValueError):
Pred.normalise_string('pred"name')
# Force lower case
with self.assertRaises(Warning):
warnings.simplefilter('error')
self.assertEqual(Pred.normalise_string('PRON'), 'pron')
warnings.resetwarnings()
# Strip trailing _rel
self.assertEqual(Pred.normalise_string('pron_rel'), 'pron')
def test_Pred_from_string(self):
"""
Pred.from_string should normalise the string as necessary
"""
cat_pred = RealPred.from_normalised_string('_cat_n_1')
self.assertEqual(Pred.from_string('_cat_n_1_rel'), cat_pred)
self.assertEqual(Pred.from_string('"_cat_n_1_rel"'), cat_pred)
with self.assertRaises(Warning):
warnings.simplefilter('error')
self.assertEqual(Pred.from_string('_CAT_N_1_REL'), cat_pred)
warnings.resetwarnings()
the_pred = GPred.from_normalised_string('the')
self.assertEqual(Pred.from_string('the_rel'), the_pred)
self.assertEqual(Pred.from_string('"the_rel"'), the_pred)
with self.assertRaises(Warning):
warnings.simplefilter('error')
self.assertEqual(Pred.from_string('THE_REL'), the_pred)
warnings.resetwarnings()
def _parse_node(string, nodeid, queries, equalities, anchors, sortinfo_classes,
sortinfo_shortforms):
m = string.find('(')
if m < 0:
m = string.find(' ')
if m < 0:
l = string.find(':')
else:
l = string.find(':', 0, m)
if l < 0:
ref_ids = None
l = 0
else:
ref_ids = string[:l]
l += 1
while string[l] == ' ':
l += 1
if string[l:l + 4] == 'node' and (len(string) - l == 4
or string[l + 4] in special_values):
value = _parse_value(string[l + 4:], None, queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]]))
assert not value
pred = Pred()
carg = '?'
sortinfo = Sortinfo()
ref_name = 'node'
elif m < 0:
pred, ref_name = _parse_pred(string[l:], nodeid, queries, equalities)
carg = None
sortinfo = None
else:
pred, ref_name = _parse_pred(string[l:m], nodeid, queries, equalities)
if string[m] == '(':
r = string.index(')', m)
if string[m + 1] == '"' and string[r - 1] == '"':
carg = string[m + 2:r - 1]
else:
carg = string[m + 1:r]
assert '"' not in carg
carg = _parse_value(
carg, '?', queries, equalities,
(lambda matching, dmrs: dmrs[matching[nodeid]].carg))
m = r + 1
else:
carg = None
if m < len(string) and string[m] == ' ':
while string[m] == ' ':
m += 1
sortinfo = _parse_sortinfo(string[m:], nodeid, queries, equalities,
sortinfo_classes, sortinfo_shortforms)
else:
sortinfo = None
if not ref_ids:
ref_ids = None
node = Node(nodeid, pred, sortinfo=sortinfo, carg=carg)
else:
if ref_ids[0] == '[' and ref_ids[-1] == ']':
ref_ids = ref_ids[1:-1].split(',')
node = AnchorNode(anchors=ref_ids,
nodeid=nodeid,
pred=pred,
sortinfo=sortinfo,
carg=carg)
elif ref_ids[0] == '(' and ref_ids[-1] == ')':
ref_ids = ref_ids[1:-1].split(',')
node = OptionalNode(anchors=ref_ids,
nodeid=nodeid,
pred=pred,
sortinfo=sortinfo,
carg=carg)
elif ref_ids[0] == '{' and ref_ids[-1] == '}':
ref_ids = ref_ids[1:-1].split(',')
node = SubgraphNode(anchors=ref_ids,
nodeid=nodeid,
pred=pred,
sortinfo=sortinfo,
carg=carg)
else:
ref_ids = ref_ids.split(',')
node = Node(nodeid=nodeid, pred=pred, sortinfo=sortinfo, carg=carg)
for ref_id in ref_ids:
assert ref_id not in anchors, 'Reference ids have to be unique.'
anchors[ref_id] = node
return node, ref_ids, ref_name
def _read_elempred(string, vs, rs):
assert string[0] == '[' and string[-1] == ']'
l = 1
while string[l] == ' ':
l += 1
m = string.find('<', l)
if m >= 0:
r = string.index('>', m)
c = string.find(':', m, r)
if c >= 0:
cfrom = int(string[m + 1:c])
cto = int(string[c + 1:r])
else:
assert m + 1 == r
cfrom = None
cto = None
else:
m = r = string.index(' ', l)
cfrom = None
cto = None
if string[m - 15:m] == '_u_unknown_rel"':
assert string[l:l + 2] == '"_'
slash = string.index('/', l + 2, m - 15)
pos = {
'FW': 'u',
'JJ': 'a',
'NN': 'n',
'NNS': 'n',
'RB': 'a',
'VB': 'v',
'VBP': 'v',
'VBG': 'v',
'VBN': 'v'
} # ?????????????????????????????????????
assert string[slash + 1:m -
15] in pos, 'Invalid unknown word POS: {}'.format(
string[slash + 1:m - 15])
pred = Pred.from_string(string[l + 1:slash] + '_' +
pos[string[slash + 1:m - 15]] +
string[m - 13:m - 1])
else:
assert string[l:m].islower()
if string[l] == '"':
assert string[m - 1] == '"'
pred = Pred.from_string(string[l + 1:m - 1])
else:
pred = Pred.from_string(string[l:m])
l = find_next(string, start=r + 1)
r = find_previous(string, start=l, end=len(string) - 1) + 1
attributes = _read_attributes(string[l:r], not_lower=('carg', ))
label = _read_reference(attributes.pop('lbl'), vs, rs)
carg = attributes.pop('carg', None)
assert (carg is not None) == (
isinstance(pred, GPred)
and pred.name in ('basic_card', 'basic_numbered_hour', 'card', 'dofm',
'dofw', 'mofy', 'named', 'named_n', 'numbered_hour',
'ord', 'season', 'year_range', 'yofc')), (
carg, pred) # gpreds with CARG
args = {}
quantifier = False
if isinstance(pred, GPred):
if pred.name in (
'def_explicit_q', 'def_implicit_q', 'def_poss_q', 'every_q',
'free_relative_q', 'free_relative_ever_q', 'idiom_q_i',
'number_q', 'pronoun_q', 'proper_q', 'some_q', 'udef_q',
'which_q'): # all quantifier gpreds according to core.smi
quantifier = True
handle = _read_reference(attributes.pop('rstr'), vs, rs)
args['rstr'] = handle
if 'body' in attributes:
handle = _read_reference(attributes.pop('body'), vs, rs)
args['body'] = handle
elif pred.name in (
'discourse', 'implicit_conj'
): # all conjunction gpreds with potential L/R-HNDL according to core.smi
if 'arg1' in attributes:
ref = _read_reference(attributes.pop('arg1'), vs, rs)
args['arg1'] = ref
ref = _read_reference(attributes.pop('arg2'), vs, rs)
args['arg2'] = ref
else:
ref = _read_reference(attributes.pop('l-index'), vs, rs)
args['l-index'] = ref
ref = _read_reference(attributes.pop('r-index'), vs, rs)
args['r-index'] = ref
if 'l-hndl' in attributes:
assert pred.name == 'implicit_conj'
handle = _read_reference(attributes.pop('l-hndl'), vs, rs)
args['l-hndl'] = handle
handle = _read_reference(attributes.pop('r-hndl'), vs, rs)
args['r-hndl'] = handle
elif pred.name in (
'fw_seq', 'num_seq'
): # all conjunction gpreds without potential L/R-HNDL according to core.smi
ref = _read_reference(attributes.pop('l-index'), vs, rs)
args['l-index'] = ref
ref = _read_reference(attributes.pop('r-index'), vs, rs)
args['r-index'] = ref
elif pred.name == 'unknown': # only predicate with ARG
ref = _read_reference(attributes.pop('arg'), vs, rs)
args['arg'] = ref
else: # regular gpreds
for n in range(1, 4):
role = 'arg' + str(n)
if role not in attributes:
break
ref = _read_reference(attributes.pop(role), vs, rs)
args[role] = ref
elif isinstance(pred, RealPred):
assert pred.pos in 'acnpquvx'
if pred.pos == 'q': # quantifier
quantifier = True
handle = _read_reference(attributes.pop('rstr'), vs, rs)
args['rstr'] = handle
if 'body' in attributes:
handle = _read_reference(attributes.pop('body'), vs, rs)
args['body'] = handle
elif pred.pos == 'c' and pred.lemma not in (
'vice+versa', ): # conjunction
if 'arg1' in attributes:
ref = _read_reference(attributes.pop('arg1'), vs, rs)
args['arg1'] = ref
if 'arg2' in attributes:
ref = _read_reference(attributes.pop('arg2'), vs, rs)
args['arg2'] = ref
else:
ref = _read_reference(attributes.pop('l-index'), vs, rs)
args['l-index'] = ref
ref = _read_reference(attributes.pop('r-index'), vs, rs)
args['r-index'] = ref
if 'l-hndl' in attributes: # optional L/R-HNDL
handle = _read_reference(attributes.pop('l-hndl'), vs, rs)
args['l-hndl'] = handle
handle = _read_reference(attributes.pop('r-hndl'), vs, rs)
args['r-hndl'] = handle
else: # regular realpreds
for n in range(1, 5):
role = 'arg' + str(n)
if role not in attributes:
break
ref = _read_reference(attributes.pop(role), vs, rs)
args[role] = ref
else:
assert False
if quantifier:
intrinsic = _read_reference(attributes.pop('arg0'), vs, rs)
var = None
else:
intrinsic, var = _read_variable(attributes.pop('arg0'), vs, rs)
assert not attributes, 'Invalid attributes for predicate {}: {}'.format(
pred, attributes)
return ElemPred(label=label,
pred=pred,
intrinsic=intrinsic,
carg=carg,
args=args,
cfrom=cfrom,
cto=cto), var
def test_Pred_repr(self):
"""
The 'official' string representation of Pred()
should evaluate to Pred()
"""
self.assertEqual(eval(repr(Pred())), Pred())
def test_Pred_str(self):
"""
The 'informal' string representation of a Pred should be
'predsort', the type name for predicates in Delph-in grammars
"""
self.assertEqual(str(Pred()), 'predsort')
def predsort():
dmrs = DictDmrs()
dmrs.add_node(
Node(pred=Pred(),
sortinfo=Sortinfo())) # underspecified predicate and sortinfo
return dmrs
# Also remove pronouns
extended_filter = DEFAULT_FILTER | {GPred('pron')}
# Replace the first pred with the second:
rename = [(RealPred('forwards', 'p'), RealPred('forward', 'p', 'dir'))]
# Replace a pair of nodes with a single node
# (the first pred linked to the second pred, is replaced by the third pred)
shrink = [('_left_a_1', 'ARG1/EQ', 'place_n', '_left_n_1'),
('_right_a_1', 'ARG1/EQ', 'place_n', '_right_n_1'),
('loc_nonsp', 'ARG2/NEQ', '_left_n_1', '_left_p_dir'),
('loc_nonsp', 'ARG2/NEQ', '_right_n_1', '_right_p_dir'),
('_to_p', 'ARG2/NEQ', '_left_n_1', '_left_p_dir'),
('_to_p', 'ARG2/NEQ', '_right_n_1', '_right_p_dir')]
shrink = [(Pred.from_string(a), LinkLabel.from_string(b), Pred.from_string(c),
Pred.from_string(d)) for a, b, c, d in shrink]
def simplify(dmrs):
"""
Simplify an input DMRS to a form that can be converted to robot commands
"""
# Remove unnecessary GPreds (defaults, plus pronouns)
gpred_filtering(dmrs, extended_filter)
# Remove quantifiers
for node in copy(dmrs.nodes):
if dmrs.is_quantifier(node.nodeid):
dmrs.remove_node(node.nodeid)
def test_Node_underspecification(self):
with self.assertRaises(TypeError):
Node(pred='_the_q').is_more_specific(4)
# complete underspecification
self.assertFalse(Node().is_more_specific(Node()))
self.assertFalse(Node().is_less_specific(Node()))
# pred underspecification
self.assertFalse(Node(pred=Pred()).is_more_specific(Node()))
self.assertTrue(Node(pred=Pred()).is_less_specific(Node()))
self.assertTrue(Node().is_more_specific(Node(pred=Pred())))
self.assertFalse(Node().is_less_specific(Node(pred=Pred())))
self.assertFalse(Node(pred=Pred()).is_more_specific(Node(pred=Pred())))
self.assertFalse(Node(pred=Pred()).is_less_specific(Node(pred=Pred())))
self.assertFalse(
Node(pred=Pred()).is_more_specific(Node(pred=GPred(name='abc'))))
self.assertTrue(
Node(pred=Pred()).is_less_specific(Node(pred=GPred(name='abc'))))
self.assertTrue(
Node(pred=GPred(name='abc')).is_more_specific(Node(pred=Pred())))
self.assertFalse(
Node(pred=GPred(name='abc')).is_less_specific(Node(pred=Pred())))
# carg underspecification
self.assertFalse(Node(carg='?').is_more_specific(Node()))
self.assertTrue(Node(carg='?').is_less_specific(Node()))
self.assertTrue(Node().is_more_specific(Node(carg='?')))
self.assertFalse(Node().is_less_specific(Node(carg='?')))
self.assertFalse(Node(carg='?').is_more_specific(Node(carg='?')))
self.assertFalse(Node(carg='?').is_less_specific(Node(carg='?')))
self.assertFalse(Node(carg='?').is_more_specific(Node(carg='abc')))
self.assertTrue(Node(carg='?').is_less_specific(Node(carg='abc')))
self.assertTrue(Node(carg='abc').is_more_specific(Node(carg='?')))
self.assertFalse(Node(carg='abc').is_less_specific(Node(carg='?')))
# sortinfo underspecification
self.assertFalse(Node(sortinfo=Sortinfo()).is_more_specific(Node()))
self.assertTrue(Node(sortinfo=Sortinfo()).is_less_specific(Node()))
self.assertTrue(Node().is_more_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(Node().is_less_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=Sortinfo()).is_more_specific(
Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=Sortinfo()).is_less_specific(
Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=Sortinfo()).is_more_specific(
Node(sortinfo=EventSortinfo(sf='abc'))))
self.assertTrue(
Node(sortinfo=Sortinfo()).is_less_specific(
Node(sortinfo=EventSortinfo(sf='abc'))))
self.assertTrue(
Node(sortinfo=EventSortinfo(sf='abc')).is_more_specific(
Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(sortinfo=EventSortinfo(sf='abc')).is_less_specific(
Node(sortinfo=Sortinfo())))
# mixed specification
self.assertFalse(Node(pred=Pred()).is_more_specific(Node(carg='?')))
self.assertFalse(Node(pred=Pred()).is_less_specific(Node(carg='?')))
self.assertFalse(
Node(pred=Pred()).is_more_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(pred=Pred()).is_less_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(carg='?').is_more_specific(Node(sortinfo=Sortinfo())))
self.assertFalse(
Node(carg='?').is_less_specific(Node(sortinfo=Sortinfo())))
# Also remove pronouns
extended_filter = DEFAULT_FILTER | {GPred('pron')}
# Replace the first pred with the second:
rename = [(RealPred('forwards','p'), RealPred('forward','p','dir'))]
# Replace a pair of nodes with a single node
# (the first pred linked to the second pred, is replaced by the third pred)
shrink = [('_left_a_1', 'ARG1/EQ', 'place_n', '_left_n_1'),
('_right_a_1', 'ARG1/EQ', 'place_n', '_right_n_1'),
('loc_nonsp', 'ARG2/NEQ', '_left_n_1', '_left_p_dir'),
('loc_nonsp', 'ARG2/NEQ', '_right_n_1', '_right_p_dir'),
('_to_p', 'ARG2/NEQ', '_left_n_1', '_left_p_dir'),
('_to_p', 'ARG2/NEQ', '_right_n_1', '_right_p_dir')]
shrink = [(Pred.from_string(a),
LinkLabel.from_string(b),
Pred.from_string(c),
Pred.from_string(d)) for a,b,c,d in shrink]
def simplify(dmrs):
"""
Simplify an input DMRS to a form that can be converted to robot commands
"""
# Remove unnecessary GPreds (defaults, plus pronouns)
gpred_filtering(dmrs, extended_filter)
# Remove quantifiers
for node in copy(dmrs.nodes):
if dmrs.is_quantifier(node.nodeid):
dmrs.remove_node(node.nodeid)