Homec4science
Diffusion invenio-infoscience (master) Browse

R3600/modules/bibmerge/lib/bibmerge_differ.py
d30e56fa001amaster

Last Change
Raw File

bibmerge_differ.py
View Options

## This file is part of Invenio.
## Copyright (C) 2002, 2003, 2004, 2005, 2006, 2007, 2008 CERN.
##
## Invenio is free software; you can redistribute it and/or
## modify it under the terms of the GNU General Public License as
## published by the Free Software Foundation; either version 2 of the
## License, or (at your option) any later version.
##
## Invenio is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
## General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with Invenio; if not, write to the Free Software Foundation, Inc.,
## 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA.
# pylint: disable=C0103
is_editdist_loaded = False
try:
import editdist
is_editdist_loaded = True
except ImportError:
# Okay, diffing will not be possible, but continue anyway,
# since this package is only recommended, not mandatory.
pass
def record_diff(rec1, rec2, compare_subfields, ind1='', ind2=''):
"""Compares two given records
Considers the change of order of fields as a change
@param rec1: First record
@param rec2: Second record
@return: dictionary of differences. Each difference is of a form:
field_id: None - if field is the same for both records
field_id: ('r',) - if field field_id exists in rec1 but not in
rec2
field_id: ('a',) - if field field_id exists in rec2 but not in rec1
field_id: ('c', new_value) - if field field_id exists in both
records, but it's value has changed
new_value describes the new value of a given field (which
allows to reconstruct new record from the old one)"""
# Very simple test to save computing power.
#if rec1 == rec2:
# return {}
result = {}
for tag in rec1:
result[tag] = record_field_diff_generic(rec1, rec2, tag, compare_subfields, ind1, ind2)
for tag in rec2:
if tag not in rec1:
result[tag] = record_field_diff_generic(rec1, rec2, tag, compare_subfields, ind1, ind2)
return result
def record_field_diff_generic(rec1, rec2, tag, compare_subfields, ind1='', ind2=''):
if tag not in rec2:
return ('r',)
if tag not in rec1:
return ('a',)
return record_field_diff(rec1[tag], rec2[tag], tag, compare_subfields, ind1, ind2)
def record_field_diff(fields1, fields2, tag, compare_subfields, ind1='', ind2=''):
"""Compares given field in two records.
returns a list containing at most one element
If the fields are identical (that means have the same order, the
same subfields), empty list is returned.
If the field is removed in second record, [(field, 'r')] is
returned
If the field is added in second record, [(field, 'a')]
is returned
If the field is changed [(field, 'c', comparison_table)] is returned.
The comparison table is a table containing pairs of indexes showing
the relations between fields."""
# Extract the fields.
fields1 = [field for field in enumerate(fields1)
if _has_indicators(field[1], ind1, ind2)]
fields2 = [field for field in enumerate(fields2)
if _has_indicators(field[1], ind1, ind2)]
if fields1 == fields2:
return None
fields_comparison = {}
idx1, idx2 = 0, 0
len1, len2 = len(fields1), len(fields2)
while idx1 < len1 and idx2 < len2:
list_index1, field1 = fields1[idx1]
list_index2, field2 = fields2[idx2]
subfields_are_similar, value = compare_subfields(field1[0], field2[0])
if subfields_are_similar:
ind_pair = (field1[1], field1[2])
fields_comparison.setdefault(ind_pair, []).append((list_index1, list_index2, value))
idx1 += 1
idx2 += 1
elif _field_in_fields(field1, fields2[idx2+1:], compare_subfields) \
is None:
ind_pair = (field1[1], field1[2])
fields_comparison.setdefault(ind_pair, []).append((list_index1, None, None))
idx1 += 1
else:
ind_pair = (field2[1], field2[2])
fields_comparison.setdefault(ind_pair, []).append((None, list_index2, None))
idx2 += 1
# Add the remaining elements.
for index in range(idx1, len(fields1)):
field1 = fields1[index][1]
list_index1 = fields1[index][0]
ind_pair = (field1[1], field1[2])
fields_comparison.setdefault(ind_pair, []).append((list_index1,None,None))
for index in range(idx2, len(fields2)):
field2 = fields2[index][1]
list_index2 = fields2[index][0]
ind_pair = (field2[1], field2[2])
fields_comparison.setdefault(ind_pair, []).append((None,list_index2,None))
comparisons_list = []
indicators = fields_comparison.keys()
indicators.sort()
for indicator in indicators:
comparisons_list.append((indicator, fields_comparison[indicator]))
return ('c', comparisons_list)
def _has_indicators(field, ind1, ind2):
"""Checks if the field has the indicators. Consider an empty
indicator as a wildcard."""
if not ind1 and not ind2:
return True
else:
return field[1:3] == (ind1, ind2)
def _same_indicators(field1, field2):
"""Checks if fields have the same indicators."""
return field1[1:3] == field2[1:3]
def _field_in_fields(field, fields, compare_subfields):
"""Checks if a field 'field' has an equivalent in the list of fields
'fields'. Uses the 'compare_subfields' method to achieve this."""
if compare_subfields is None:
compare_subfields = lambda a, b: a == b
for index, field2 in fields:
if compare_subfields(field[0], field2[0])[0]:
return index
return None
def compare_strings(str1, str2):
"""Compares 2 strings with the Levenshtein distance and returns a normalized
value between 0.0 and 1.0 (meaning totally different and exactly the same
respectively."""
if is_editdist_loaded:
if str1 == str2:
return 1.0
max_len = max(len(str1), len(str2))
if max_len == 0:
return 0.0
distance = editdist.distance(str1, str2)
return (max_len - distance) / float(max_len)
else:
# the edit distance module is not loadable, we have to fail the comparison
# all the strings will be treated as completely different
return 0.0
def compare_subfields(subfield1, subfield2):
"""Compare two subfields taking into account the subfield code and the
subfield string value."""
#compare subfield code
if subfield1[0] != subfield2[0]:
return 0.0
#compare subfield values
return compare_strings(subfield1[1], subfield2[1])
def diff_subfields(subfields1, subfields2):
"""Return a list of diffs for the subfields. A diff is a tuple of:
(index-of-subfield1, index-of-subfield2, distance-value)."""
# The result of the comparisons.
subfields_comparison = []
# Two indexes used to retain the position of the subfields to compare.
idx1, idx2 = 0, 0
len1, len2 = len(subfields1), len(subfields2)
while idx1 < len1 and idx2 < len2:
subfield1 = subfields1[idx1]
for i in range(idx2, len2):
subfield2 = subfields2[i]
# Compare the two current subfields.
value = compare_subfields(subfield1, subfield2)
if value >= 0.5: #we have a match
for j in range(idx2, i): #no match for subfields2 in between, if any
subfields_comparison.append((None, j, 0.0))
subfields_comparison.append((idx1, i, value))
idx2 = i+1
break
else: #no match for subfield1
subfields_comparison.append((idx1, None, 0.0))
idx1 += 1
# Add the remaining elements.
for index in range(idx1, len1):
subfields_comparison.append((index, None, 0.0))
for index in range(idx2, len2):
subfields_comparison.append((None, index, 0.0))
return subfields_comparison
def match_subfields(subfields1, subfields2):
"""False if subfields dont match, True if they do and also their diff is
returned."""
subfield_diffs = diff_subfields(subfields1, subfields2)
listofscores = [x[2] for x in subfield_diffs]
if len(listofscores)==0: #in case of a controlfield like '001', '003',...
return (False, None)
#length of listofscores should be normally != 0
average_score = sum(listofscores) / len(listofscores)
if average_score >= 0.5:
return (True, subfield_diffs)
else:
return (False, None)
def Levenshtein_diffs(str1, str2):
"""Actions (insert, delete, substitute, none) needed to perform on the two
strings to make them identical."""
actions = []
matrix = _Levenshtein_matrix(str1, str2)
i, j = len(str2), len(str1)
counter = 0
lastaction = None
while i!=0 and j!=0:
lastvalue = matrix[i][j]
values = [matrix[i-1][j-1], matrix[i][j-1], matrix[i-1][j]]
indexofmin = _min_index(values)
if indexofmin == 0:
if lastvalue == values[indexofmin]:
action = 'n'
else:
action = 's'
i, j = i-1, j-1
elif indexofmin == 1:
action = 'i'
j = j-1
else: # if indexofmin == 2
action = 'd'
i = i-1
if action != lastaction:
actions.append( (lastaction, counter) )
counter = 0
lastaction = action
counter = counter + 1
actions.append( (lastaction, counter) )
if i>0:
actions.append( ('d', i) )
if j>0:
actions.append( ('i', j) )
actions.pop(0)
actions.reverse()
return actions
def _Levenshtein_matrix(str1, str2):
len1, len2 = len(str1), len(str2)
# two-dimensional array of distances
dist = []
# initial values
for i in range( len2 + 1 ):
dist.append( [i] )
for j in range(1, len1 + 1 ):
dist[0].append( j )
# calculation of minimum distance
for i in range(1, len2 + 1 ):
for j in range(1, len1 + 1 ):
if str1[j-1] == str2[i-1]:
cost = 0
else:
cost = 1
# choose between deletion, insertion, substitution
dist[i].append( min( dist[i-1][j] + 1, \
dist[i][j-1] + 1, \
dist[i-1][j-1] + cost ) )
return dist
def _min_index(alist):
min_i = 0 #index of item with minimum value
for i in range(1, len(alist)):
if alist[i] < alist[min_i]:
min_i = i
return min_i
c4science · Help