def _parse_attributes(attributes):
if attributes is None:
_attributes = {}
else:
try:
_attributes = json_loads(attributes)
except ValueError:
# Failed to parse, warn the client
Messager.warning(
('Unable to parse attributes string "%s" for '
'"createSpan", ignoring attributes for request and '
'assuming no attributes set') %
(attributes, ))
_attributes = {}
# XXX: Hack since the client is sending back False and True as values...
# These are __not__ to be sent, they violate the protocol
for _del in [k for k, v in list(_attributes.items()) if v == False]:
del _attributes[_del]
# These are to be old-style modifiers without values
for _revalue in [k for k, v in list(_attributes.items()) if v]:
_attributes[_revalue] = True
###
return _attributes
def undo(collection, document, token):
try:
token = json_loads(token)
except ValueError:
raise CorruptUndoTokenError
try:
action = token['action']
except KeyError:
raise InvalidTokenError('action')
if action == 'add_tb':
# Undo an addition
return delete_span(collection, document, token['id'])
if action == 'mod_tb':
# Undo a modification
# TODO: We do not handle attributes and comments
return create_span(
collection,
document,
token['start'],
token['end'],
token['type'],
id=token['id'],
attributes=token['attributes'],
comment=token['comment'] if 'comment' in token else None)
else:
raise NonUndoableActionError
assert False, 'should have returned prior to this point'
def undo(collection, document, token):
try:
token = json_loads(token)
except ValueError:
raise CorruptUndoTokenError
try:
action = token['action']
except KeyError:
raise InvalidTokenError('action')
if action == 'add_tb':
# Undo an addition
return delete_span(collection, document, token['id'])
if action == 'mod_tb':
# Undo a modification
# TODO: We do not handle attributes and comments
return create_span(
collection,
document,
token['start'],
token['end'],
token['type'],
id=token['id'],
attributes=token['attributes'],
comment=token['comment'] if 'comment' in token else None)
else:
raise NonUndoableActionError
assert False, 'should have returned prior to this point'
def _json_offsets_to_list(offsets):
try:
offsets = json_loads(offsets)
except Exception:
Messager.error('create_span: protocol argument error: expected offsets as JSON, but failed to parse "%s"' % str(offsets))
raise ProtocolArgumentError
try:
offsets = [(int(s),int(e)) for s,e in offsets]
except Exception:
Messager.error('create_span: protocol argument error: expected offsets as list of int pairs, received "%s"' % str(offsets))
raise ProtocolArgumentError
return offsets
def _parse_span_normalizations(normalizations):
if normalizations is None:
_normalizations = {}
else:
try:
_normalizations = json_loads(normalizations)
except ValueError:
# Failed to parse, warn the client
Messager.warning((u'Unable to parse normalizations string "%s" for '
u'"createSpan", ignoring normalizations for request and '
u'assuming no normalizations set') % (normalizations, ))
_normalizations = {}
return _normalizations
def _parse_span_normalizations(normalizations):
if normalizations is None:
_normalizations = {}
else:
try:
_normalizations = json_loads(normalizations)
except ValueError:
# Failed to parse, warn the client
Messager.warning((u'Unable to parse normalizations string "%s" for '
u'"createSpan", ignoring normalizations for request and '
u'assuming no normalizations set') % (normalizations, ))
_normalizations = {}
return _normalizations
def _json_offsets_to_list(offsets):
try:
offsets = json_loads(offsets)
except Exception:
Messager.error(
'create_span: protocol argument error: expected offsets as JSON, but failed to parse "%s"' %
str(offsets))
raise ProtocolArgumentError
try:
offsets = [(int(s), int(e)) for s, e in offsets]
except Exception:
Messager.error(
'create_span: protocol argument error: expected offsets as list of int pairs, received "%s"' %
str(offsets))
raise ProtocolArgumentError
return offsets
def split_span(collection, document, args, id):
directory = collection
real_dir = real_directory(directory)
document = path_join(real_dir, document)
# TODO don't know how to pass an array directly, so doing extra catenate and split
tosplit_args = json_loads(args)
with TextAnnotations(document) as ann_obj:
# bail as quick as possible if read-only
if ann_obj._read_only:
raise AnnotationsIsReadOnlyError(ann_obj.get_document())
mods = ModificationTracker()
ann = ann_obj.get_ann_by_id(id)
# currently only allowing splits for events
if not isinstance(ann, EventAnnotation):
raise AnnotationSplitError("Cannot split an annotation of type %s" % ann.type)
# group event arguments into ones that will be split on and
# ones that will not, placing the former into a dict keyed by
# the argument without trailing numbers (e.g. "Theme1" ->
# "Theme") and the latter in a straight list.
split_args = {}
nonsplit_args = []
import re
for arg, aid in ann.args:
m = re.match(r'^(.*?)\d*$', arg)
if m:
arg = m.group(1)
if arg in tosplit_args:
if arg not in split_args:
split_args[arg] = []
split_args[arg].append(aid)
else:
nonsplit_args.append((arg, aid))
# verify that split is possible
for a in tosplit_args:
acount = len(split_args.get(a,[]))
if acount < 2:
raise AnnotationSplitError("Cannot split %s on %s: only %d %s arguments (need two or more)" % (ann.id, a, acount, a))
# create all combinations of the args on which to split
argument_combos = [[]]
for a in tosplit_args:
new_combos = []
for aid in split_args[a]:
for c in argument_combos:
new_combos.append(c + [(a, aid)])
argument_combos = new_combos
# create the new events (first combo will use the existing event)
from copy import deepcopy
new_events = []
for i, arg_combo in enumerate(argument_combos):
# tweak args
if i == 0:
ann.args = nonsplit_args[:] + arg_combo
else:
newann = deepcopy(ann)
newann.id = ann_obj.get_new_id("E") # TODO: avoid hard-coding ID prefix
newann.args = nonsplit_args[:] + arg_combo
ann_obj.add_annotation(newann)
new_events.append(newann)
mods.addition(newann)
# then, go through all the annotations referencing the original
# event, and create appropriate copies
for a in ann_obj:
soft_deps, hard_deps = a.get_deps()
refs = soft_deps | hard_deps
if ann.id in refs:
# Referenced; make duplicates appropriately
if isinstance(a, EventAnnotation):
# go through args and make copies for referencing
new_args = []
for arg, aid in a.args:
if aid == ann.id:
for newe in new_events:
new_args.append((arg, newe.id))
a.args.extend(new_args)
elif isinstance(a, AttributeAnnotation):
for newe in new_events:
newmod = deepcopy(a)
newmod.target = newe.id
newmod.id = ann_obj.get_new_id("A") # TODO: avoid hard-coding ID prefix
ann_obj.add_annotation(newmod)
mods.addition(newmod)
elif isinstance(a, BinaryRelationAnnotation):
# TODO
raise AnnotationSplitError("Cannot adjust annotation referencing split: not implemented for relations! (WARNING: annotations may be in inconsistent state, please reload!) (Please complain to the developers to fix this!)")
elif isinstance(a, OnelineCommentAnnotation):
for newe in new_events:
newcomm = deepcopy(a)
newcomm.target = newe.id
newcomm.id = ann_obj.get_new_id("#") # TODO: avoid hard-coding ID prefix
ann_obj.add_annotation(newcomm)
mods.addition(newcomm)
elif isinstance(a, NormalizationAnnotation):
for newe in new_events:
newnorm = deepcopy(a)
newnorm.target = newe.id
newnorm.id = ann_obj.get_new_id("N") # TODO: avoid hard-coding ID prefix
ann_obj.add_annotation(newnorm)
mods.addition(newnorm)
else:
raise AnnotationSplitError("Cannot adjust annotation referencing split: not implemented for %s! (Please complain to the lazy developers to fix this!)" % a.__class__)
mods_json = mods.json_response()
mods_json['annotations'] = _json_from_ann(ann_obj)
return mods_json
def split_span(collection, document, args, id):
directory = collection
real_dir = real_directory(directory)
document = path_join(real_dir, document)
# TODO don't know how to pass an array directly, so doing extra catenate
# and split
tosplit_args = json_loads(args)
with TextAnnotations(document) as ann_obj:
# bail as quick as possible if read-only
if ann_obj._read_only:
raise AnnotationsIsReadOnlyError(ann_obj.get_document())
mods = ModificationTracker()
ann = ann_obj.get_ann_by_id(id)
# currently only allowing splits for events
if not isinstance(ann, EventAnnotation):
raise AnnotationSplitError(
"Cannot split an annotation of type %s" %
ann.type)
# group event arguments into ones that will be split on and
# ones that will not, placing the former into a dict keyed by
# the argument without trailing numbers (e.g. "Theme1" ->
# "Theme") and the latter in a straight list.
split_args = {}
nonsplit_args = []
import re
for arg, aid in ann.args:
m = re.match(r'^(.*?)\d*$', arg)
if m:
arg = m.group(1)
if arg in tosplit_args:
if arg not in split_args:
split_args[arg] = []
split_args[arg].append(aid)
else:
nonsplit_args.append((arg, aid))
# verify that split is possible
for a in tosplit_args:
acount = len(split_args.get(a, []))
if acount < 2:
raise AnnotationSplitError(
"Cannot split %s on %s: only %d %s arguments (need two or more)" %
(ann.id, a, acount, a))
# create all combinations of the args on which to split
argument_combos = [[]]
for a in tosplit_args:
new_combos = []
for aid in split_args[a]:
for c in argument_combos:
new_combos.append(c + [(a, aid)])
argument_combos = new_combos
# create the new events (first combo will use the existing event)
from copy import deepcopy
new_events = []
for i, arg_combo in enumerate(argument_combos):
# tweak args
if i == 0:
ann.args = nonsplit_args[:] + arg_combo
else:
newann = deepcopy(ann)
# TODO: avoid hard-coding ID prefix
newann.id = ann_obj.get_new_id("E")
newann.args = nonsplit_args[:] + arg_combo
ann_obj.add_annotation(newann)
new_events.append(newann)
mods.addition(newann)
# then, go through all the annotations referencing the original
# event, and create appropriate copies
for a in ann_obj:
soft_deps, hard_deps = a.get_deps()
refs = soft_deps | hard_deps
if ann.id in refs:
# Referenced; make duplicates appropriately
if isinstance(a, EventAnnotation):
# go through args and make copies for referencing
new_args = []
for arg, aid in a.args:
if aid == ann.id:
for newe in new_events:
new_args.append((arg, newe.id))
a.args.extend(new_args)
elif isinstance(a, AttributeAnnotation):
for newe in new_events:
newmod = deepcopy(a)
newmod.target = newe.id
# TODO: avoid hard-coding ID prefix
newmod.id = ann_obj.get_new_id("A")
ann_obj.add_annotation(newmod)
mods.addition(newmod)
elif isinstance(a, BinaryRelationAnnotation):
# TODO
raise AnnotationSplitError(
"Cannot adjust annotation referencing split: not implemented for relations! (WARNING: annotations may be in inconsistent state, please reload!) (Please complain to the developers to fix this!)")
elif isinstance(a, OnelineCommentAnnotation):
for newe in new_events:
newcomm = deepcopy(a)
newcomm.target = newe.id
# TODO: avoid hard-coding ID prefix
newcomm.id = ann_obj.get_new_id("#")
ann_obj.add_annotation(newcomm)
mods.addition(newcomm)
elif isinstance(a, NormalizationAnnotation):
for newe in new_events:
newnorm = deepcopy(a)
newnorm.target = newe.id
# TODO: avoid hard-coding ID prefix
newnorm.id = ann_obj.get_new_id("N")
ann_obj.add_annotation(newnorm)
mods.addition(newnorm)
else:
raise AnnotationSplitError(
"Cannot adjust annotation referencing split: not implemented for %s! (Please complain to the lazy developers to fix this!)" %
a.__class__)
mods_json = mods.json_response()
mods_json['annotations'] = _json_from_ann(ann_obj)
return mods_json
def create_span(collection, document, start, end, type, attributes=None, id=None, comment=None):
directory = collection
if attributes is None:
_attributes = {}
else:
# TODO: Catch parse failures here
_attributes = json_loads(attributes)
### XXX: Hack since the client is sending back False and True as values...
# These are __not__ to be sent, they violate the protocol
for _del in [k for k, v in _attributes.items() if v == False]:
del _attributes[_del]
# These are to be old-style modifiers without values
for _revalue in [k for k, v in _attributes.items() if v == True]:
_attributes[_revalue] = True
###
# log_info('ATTR: %s' %(_attributes, ))
real_dir = real_directory(directory)
document = path_join(real_dir, document)
projectconf = ProjectConfiguration(real_dir)
txt_file_path = document + "." + TEXT_FILE_SUFFIX
working_directory = path_split(document)[0]
with TextAnnotations(document) as ann_obj:
mods = ModificationTracker()
if id is not None:
# We are to edit an existing annotation
tb_ann, e_ann = _edit_span(ann_obj, mods, id, start, end, projectconf, _attributes, type)
else:
# We are to create a new annotation
tb_ann, e_ann = _create_span(ann_obj, mods, type, start, end, txt_file_path, projectconf, _attributes)
# Set annotation attributes
if e_ann is not None:
# Assign attributes to the event, not the trigger
_set_attributes(ann_obj, e_ann, _attributes, mods)
else:
_set_attributes(ann_obj, tb_ann, _attributes, mods)
# Handle annotation comments
if tb_ann is not None:
# If this is an event, we want to attach the comment to it
if e_ann is not None:
comment_on = e_ann
else:
comment_on = tb_ann
# We are only interested in id;ed comments
try:
comment_on.id
has_id = True
except AttributeError:
has_id = False
if has_id:
# Check if there is already an annotation comment
for com_ann in ann_obj.get_oneline_comments():
if com_ann.type == "AnnotatorNotes" and com_ann.target == comment_on.id:
found = com_ann
break
else:
found = None
if comment:
if found is not None:
# Change the comment
# XXX: Note the ugly tab, it is for parsing the tail
before = unicode(found)
found.tail = "\t" + comment
mods.change(before, found)
else:
# Create a new comment
new_comment = OnelineCommentAnnotation(
comment_on.id,
ann_obj.get_new_id("#"),
# XXX: Note the ugly tab
"AnnotatorNotes",
"\t" + comment,
)
ann_obj.add_annotation(new_comment)
mods.addition(new_comment)
else:
# We are to erase the annotation
if found is not None:
ann_obj.del_annotation(found)
mods.deletion(found)
if tb_ann is not None:
if DEBUG:
mods_json = mods.json_response()
else:
mods_json = {}
else:
# Hack, we had a new-line in the span
mods_json = {}
Messager.error("Text span contained new-line, rejected", duration=3)
# save a roundtrip and send the annotations also
txt_file_path = document + "." + TEXT_FILE_SUFFIX
j_dic = _json_from_ann_and_txt(ann_obj, txt_file_path)
mods_json["annotations"] = j_dic
return mods_json
def _create_span(collection, document, start, end, _type, attributes=None,
_id=None, comment=None):
directory = collection
undo_resp = {}
if attributes is None:
_attributes = {}
else:
try:
_attributes = json_loads(attributes)
except ValueError:
# Failed to parse attributes, warn the client
Messager.warning((u'Unable to parse attributes string "%s" for '
u'"createSpan", ignoring attributes for request and '
u'assuming no attributes set') % (attributes, ))
_attributes = {}
### XXX: Hack since the client is sending back False and True as values...
# These are __not__ to be sent, they violate the protocol
for _del in [k for k, v in _attributes.items() if v == False]:
del _attributes[_del]
# These are to be old-style modifiers without values
for _revalue in [k for k, v in _attributes.items() if v == True]:
_attributes[_revalue] = True
###
#log_info('ATTR: %s' %(_attributes, ))
real_dir = real_directory(directory)
document = path_join(real_dir, document)
projectconf = ProjectConfiguration(real_dir)
txt_file_path = document + '.' + TEXT_FILE_SUFFIX
working_directory = path_split(document)[0]
with TextAnnotations(document) as ann_obj:
# bail as quick as possible if read-only
if ann_obj._read_only:
raise AnnotationsIsReadOnlyError(ann_obj.get_document())
mods = ModificationTracker()
if _id is not None:
# We are to edit an existing annotation
tb_ann, e_ann = _edit_span(ann_obj, mods, _id, start, end, projectconf,
_attributes, _type, undo_resp=undo_resp)
else:
# We are to create a new annotation
tb_ann, e_ann = __create_span(ann_obj, mods, _type, start, end, txt_file_path,
projectconf, _attributes)
undo_resp['action'] = 'add_tb'
if e_ann is not None:
undo_resp['id'] = e_ann.id
else:
undo_resp['id'] = tb_ann.id
# Set annotation attributes
if e_ann is not None:
# Assign attributes to the event, not the trigger
_set_attributes(ann_obj, e_ann, _attributes, mods,
undo_resp=undo_resp)
else:
_set_attributes(ann_obj, tb_ann, _attributes, mods,
undo_resp=undo_resp)
# Handle annotation comments
if tb_ann is not None:
# If this is an event, we want to attach the comment to it
if e_ann is not None:
comment_on = e_ann
else:
comment_on = tb_ann
# We are only interested in id;ed comments
try:
comment_on.id
has_id = True
except AttributeError:
has_id = False
if has_id:
# Check if there is already an annotation comment
for com_ann in ann_obj.get_oneline_comments():
if (com_ann.type == 'AnnotatorNotes'
and com_ann.target == comment_on.id):
found = com_ann
# Note the comment in the undo
undo_resp['comment'] = found.tail[1:]
break
else:
found = None
if comment:
if found is not None:
# Change the comment
# XXX: Note the ugly tab, it is for parsing the tail
before = unicode(found)
found.tail = u'\t' + comment
mods.change(before, found)
else:
# Create a new comment
new_comment = OnelineCommentAnnotation(
comment_on.id, ann_obj.get_new_id('#'),
# XXX: Note the ugly tab
u'AnnotatorNotes', u'\t' + comment)
ann_obj.add_annotation(new_comment)
mods.addition(new_comment)
else:
# We are to erase the annotation
if found is not None:
ann_obj.del_annotation(found)
mods.deletion(found)
if tb_ann is not None:
mods_json = mods.json_response()
else:
# Hack, we had a new-line in the span
mods_json = {}
Messager.error('Text span contained new-line, rejected', duration=3)
if undo_resp:
mods_json['undo'] = json_dumps(undo_resp)
mods_json['annotations'] = _json_from_ann(ann_obj)
return mods_json
