def ancestors(obj, refattrs=(ALIGNMENT, SEGMENTATION)):
"""
>>> for anc in query.ancestors(igt.get_item('g1'), refattrs=(ALIGNMENT, SEGMENTATION)):
... print(anc)
(<Tier object (id: g type: glosses) at ...>, 'alignment', <Tier object (id: m type: morphemes) at ...>, [<Item object (id: m1) at ...>])
(<Tier object (id: m type: morphemes) at ...>, 'segmentation', <Tier object (id: w type: words) at ...>, [<Item object (id: w1) at ...>])
(<Tier object (id: w type: words) at ...>, 'segmentation', <Tier object (id: p type: phrases) at ...>, [<Item object (id: p1) at ...>])
"""
if hasattr(obj, 'tier'):
tier = obj.tier
items = [obj]
else:
tier = obj
items = tier.items
# a tier may be visited twice (e.g. A > B > A), but then it stops;
# this is to avoid cycles
visited = set([tier.id])
while True:
# get the first specified attribute
refattr = next((ra for ra in refattrs if ra in tier.attributes), None)
if not refattr:
break
reftier = ref.dereference(tier, refattr)
ids = set(chain.from_iterable(
ref.ids(item.attributes.get(refattr, '')) for item in items
))
refitems = [item for item in reftier.items if item.id in ids]
yield (tier, refattr, reftier, refitems)
# cycle detection; break if we've now encountered something twice
if reftier.id in visited:
break
visited.update(reftier.id)
tier = reftier
items = refitems
def get_agenda(tier):
agenda = [(ref.ids(item.alignment or item.segmentation), item)
for item in tier.items]
# then group those with the same alignment (still a list [(ids, item)])
agenda = deque(tuple([k, [g[1] for g in gs]])
for k, gs in groupby(agenda, key=lambda x: x[0]))
return agenda
def ancestors(obj, refattrs=(ALIGNMENT, SEGMENTATION)):
"""
>>> for anc in query.ancestors(igt.get_item('g1'), refattrs=(ALIGNMENT, SEGMENTATION)):
... print(anc)
(<Tier object (id: g type: glosses) at ...>, 'alignment', <Tier object (id: m type: morphemes) at ...>, [<Item object (id: m1) at ...>])
(<Tier object (id: m type: morphemes) at ...>, 'segmentation', <Tier object (id: w type: words) at ...>, [<Item object (id: w1) at ...>])
(<Tier object (id: w type: words) at ...>, 'segmentation', <Tier object (id: p type: phrases) at ...>, [<Item object (id: p1) at ...>])
"""
if hasattr(obj, 'tier'):
tier = obj.tier
items = [obj]
else:
tier = obj
items = tier.items
while True:
# get the first specified attribute
refattr = next((ra for ra in refattrs if ra in tier.attributes), None)
if not refattr:
break
reftier = ref.dereference(tier, refattr)
ids = set(
chain.from_iterable(
ref.ids(item.attributes.get(refattr, '')) for item in items))
refitems = [item for item in reftier.items if item.id in ids]
yield (tier, refattr, reftier, refitems)
tier = reftier
items = refitems
def _update_referent_index(self, obj):
if obj.id is None:
warnings.warn('Cannot cache referents for an object with no id.',
XigtWarning)
return
rdict = self._referent_cache.setdefault(obj.id, {})
for refattr in obj.allowed_reference_attributes():
rdict[refattr] = ref.ids(obj.attributes.get(refattr, ''))
def get_agenda(tier):
agenda = [(ref.ids(item.alignment or item.segmentation), item)
for item in tier.items]
# then group those with the same alignment (still a list [(ids, item)])
agenda = deque(
tuple([k, [g[1] for g in gs]])
for k, gs in groupby(agenda, key=lambda x: x[0]))
return agenda
def odin_ancestor(obj):
# ODIN_LOG.debug("Looking up the odin ancestor for {}".format(str(obj)))
# If we are at an ODIN item, return.
if isinstance(obj, Item) and obj.tier.type == ODIN_TIER_TYPE:
return obj
# An Igt instance can't have a tier ancestor.
elif isinstance(obj, Igt):
return None
# Also, an ODIN tier can't get a specific item...
elif isinstance(obj, Tier) and obj.type == ODIN_TIER_TYPE:
return None
else:
if SEGMENTATION in obj.attributes:
ref_attr = SEGMENTATION
elif CONTENT in obj.attributes:
ref_attr = CONTENT
elif ALIGNMENT in obj.attributes:
ref_attr = ALIGNMENT
elif DS_DEP_ATTRIBUTE in obj.attributes:
ref_attr = DS_DEP_ATTRIBUTE
else:
return None
# If this item is a tier, we would like to follow a random object
if isinstance(obj, Tier):
if len(obj) == 0:
id = obj.attributes[ref_attr]
else:
id = [ids(i.attributes[ref_attr])[0] for i in obj if ref_attr in i.attributes][0]
elif isinstance(obj, Item):
id = ids(obj.attributes[ref_attr])[0]
else:
raise Exception
item = xigt_find(obj.igt, id=id)
if item is None:
return None
else:
return odin_ancestor(item)
def _update_referent_index(self, obj):
if obj.id is None:
warnings.warn(
'Cannot cache referents for an object with no id.',
XigtWarning
)
return
rdict = self._referent_cache.setdefault(obj.id, {})
for refattr in obj.allowed_reference_attributes():
rdict[refattr] = ref.ids(obj.attributes.get(refattr, ''))
def _update_referrer_index(self, obj):
o_id = obj.id
if o_id is None:
warnings.warn('Cannot cache referrers for an object with no id.',
XigtWarning)
return
rdict = self._referrer_cache
attrget = obj.attributes.get # just loop optimization
for refattr in obj.allowed_reference_attributes():
ids = ref.ids(attrget(refattr, ''))
for id in ids:
rdict.setdefault(id, {}).setdefault(refattr, []).append(o_id)
def test_ids(self):
assert ref.ids('') == []
assert ref.ids('a1') == ['a1']
assert ref.ids('a1[3:5]') == ['a1']
assert ref.ids('a1[3:5+6:7]+a2[1:4]') == ['a1', 'a2']
assert ref.ids('a1[3:5+6:7]+a1[1:4]+a1') == ['a1', 'a1', 'a1']
assert ref.ids('a1 a2 a3') == ['a1', 'a2', 'a3']
def _update_referrer_index(self, obj):
o_id = obj.id
if o_id is None:
warnings.warn(
'Cannot cache referrers for an object with no id.',
XigtWarning
)
return
rdict = self._referrer_cache
attrget = obj.attributes.get # just loop optimization
for refattr in obj.allowed_reference_attributes():
ids = ref.ids(attrget(refattr, ''))
for id in ids:
rdict.setdefault(id, {}).setdefault(refattr, []).append(o_id)
def descendants(obj, refattrs=(SEGMENTATION, ALIGNMENT), follow='first'):
"""
>>> for des in query.descendants(igt.get_item('p1'), refattrs=(SEGMENTATION, ALIGNMENT)):
... print(des)
(<Tier object (id: p type: phrases) at ...>, 'segmentation', <Tier object (id: w type: words) at ...>, [<Item object (id: w1) at ...>])
(<Tier object (id: p type: phrases) at ...>, 'alignment', <Tier object (id: t type: translations) at ...>, [<Item object (id: t1) at ...>])
(<Tier object (id: w type: words) at ...>, 'segmentation', <Tier object (id: m type: morphemes) at ...>, [<Item object (id: m1) at ...>])
(<Tier object (id: m type: morphemes) at ...>, 'alignment', <Tier object (id: g type: glosses) at ...>, [<Item object (id: g1) at ...>])
"""
if hasattr(obj, 'tier'):
tier = obj.tier
items = [obj]
else:
tier = obj
items = tier.items
igt = tier.igt
visited = set()
agenda = deque([(tier, items)])
while agenda:
tier, items = agenda.popleft()
tier_refs = tier.referrers(refattrs)
item_ids = set(item.id for item in items)
# get followable refattrs with something on the referrers list
ras = [ra for ra in refattrs if tier_refs[ra]]
if follow == 'first' and ras:
ras = [ras[0]]
if not ras:
continue
# unlike ancestors, descendants for a refattr may have 1+ tiers
for refattr in ras:
# try to avoid cycles
if (tier.id, refattr) in visited:
continue
else:
visited.add((tier.id, refattr))
for reftier_id in tier_refs[refattr]:
reftier = igt[reftier_id]
refitems = [
item for item in reftier.items
if set(ref.ids(item.attributes.get(
refattr, ''))).intersection(item_ids)
]
yield (tier, refattr, reftier, refitems)
agenda.append((reftier, refitems))
#def ingroup(obj, refattrs)
#def filter([objs], lambda x:
def descendants(obj, refattrs=(SEGMENTATION, ALIGNMENT), follow='first'):
"""
>>> for des in query.descendants(igt.get_item('p1'), refattrs=(SEGMENTATION, ALIGNMENT)):
... print(des)
(<Tier object (id: p type: phrases) at ...>, 'segmentation', <Tier object (id: w type: words) at ...>, [<Item object (id: w1) at ...>])
(<Tier object (id: p type: phrases) at ...>, 'alignment', <Tier object (id: t type: translations) at ...>, [<Item object (id: t1) at ...>])
(<Tier object (id: w type: words) at ...>, 'segmentation', <Tier object (id: m type: morphemes) at ...>, [<Item object (id: m1) at ...>])
(<Tier object (id: m type: morphemes) at ...>, 'alignment', <Tier object (id: g type: glosses) at ...>, [<Item object (id: g1) at ...>])
"""
if hasattr(obj, 'tier'):
tier = obj.tier
items = [obj]
else:
tier = obj
items = tier.items
igt = tier.igt
visited = set()
agenda = deque([(tier, items)])
while agenda:
tier, items = agenda.popleft()
tier_refs = tier.referrers(refattrs)
item_ids = set(item.id for item in items)
# get followable refattrs with something on the referrers list
ras = [ra for ra in refattrs if tier_refs[ra]]
if follow == 'first' and ras:
ras = [ras[0]]
if not ras:
continue
# unlike ancestors, descendants for a refattr may have 1+ tiers
for refattr in ras:
# try to avoid cycles
if (tier.id, refattr) in visited:
continue
else:
visited.add((tier.id, refattr))
for reftier_id in tier_refs[refattr]:
reftier = igt[reftier_id]
refitems = [
item for item in reftier.items
if set(ref.ids(item.attributes.get(refattr,'')))
.intersection(item_ids)
]
yield (tier, refattr, reftier, refitems)
agenda.append((reftier, refitems))
#def ingroup(obj, refattrs)
#def filter([objs], lambda x:
def algnexpr_ids_in_referred_tier(item, refattr):
itemref = item.attributes.get(refattr)
reftier = get_referred_tier(item, refattr)
if not itemref or not reftier:
return
missing = []
for ae_id in ids(itemref):
if reftier.get(ae_id) is None:
missing.append(ae_id)
if missing:
return (
'The "{}" alignment expression {{modal}} select available '
'<item> ids from the aligned <tier> ("{}"). The following are '
'unavailable: {}'
.format(refattr, str(reftier.id), ', '.join(missing))
)
def test_ids(self):
self.assertEqual(ref.ids(''), [])
self.assertEqual(ref.ids('a1'),
['a1'])
self.assertEqual(ref.ids('a1[3:5]'),
['a1'])
self.assertEqual(ref.ids('a1[3:5+6:7]+a2[1:4]'),
['a1', 'a2'])
self.assertEqual(ref.ids('a1[3:5+6:7]+a1[1:4]+a1'),
['a1', 'a1', 'a1'])
self.assertEqual(ref.ids('a1 a2 a3'),
['a1', 'a2', 'a3'])
def ref_match(o, target_ref, ref_type):
if ref_type in o.attributes:
my_ref = o.attributes.get(ref_type)
if my_ref and target_ref in ref.ids(my_ref):
return True
return False
