Page MenuHomec4science

bibrecord_tests.py
No OneTemporary

File Metadata

Created
Tue, Nov 5, 06:30

bibrecord_tests.py

## $Id$
## This file is part of the CERN Document Server Software (CDSware).
## Copyright (C) 2002 CERN.
##
## The CDSware 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.
##
## The CDSware 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 CDSware; if not, write to the Free Software Foundation, Inc.,
## 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA.
<protect># -*- coding: utf-8 -*-</protect>
from config import tmpdir, etcdir
import bibrecord
import unittest
from string import expandtabs, replace
class SanityTest(unittest.TestCase):
### check for sanity -- xml -> create_record -> xml
def test_for_sanity(self):
""" bibrecord - checking for sanity """
f=open(tmpdir + '/demobibdata.xml','r')
xmltext = f.read()
f.close()
rs = bibrecord.create_records(xmltext)
recs = map((lambda x:x[0]),rs)
xmlT = bibrecord.records_xml_output(recs)
x = xmlT.replace('\n','')
y = xmltext.replace('\n','')
xx=expandtabs(x)
yy=expandtabs(y)
xxx = xx.replace(' ','')
yyy = yy.replace(' ','')
xxx = xxx.replace('<!DOCTYPEcollectionSYSTEM"file://%s/bibedit/MARC21slim.dtd"><collection>' % etcdir,
'<collectionxmlns="http://www.loc.gov/MARC21/slim">')
self.assertEqual(xxx,yyy)
### testing for success
class SuccessTest(unittest.TestCase):
""" bibrecord - testing for success """
def setUp(self):
f=open(tmpdir + '/demobibdata.xml','r')
xmltext = f.read()
f.close()
self.recs = map((lambda x:x[0]),bibrecord.create_records(xmltext))
def test_records_created(self):
""" bibrecord - number of records created """
## check if it creates every records (the file demobibdata.xml has 75 records)
self.assertEqual(75,len(self.recs))
def test_tags_created(self):
""" bibrecord - tags created """
## check if the tags are correct
tags= ['020', '037', '041', '080', '088', '100', '245', '246', '250', '260', '270', '300', '340', '490', '500', '502', '520', '590', '595', '650', '653', '690', '700', '710', '856','909','980','999']
t=[]
for rec in self.recs:
t.extend(rec.keys())
t.sort()
#eliminate the elements repeated
tt = []
for x in t:
if not x in tt:
tt.append(x)
self.assertEqual(tags,tt)
def test_fields_created(self):
"""bibrecord - fields created"""
## check if the number of fields for each record is correct
fields=[13,13, 8, 11, 10,12, 10, 14, 10, 17, 13, 15, 10, 9, 14, 10, 11, 11, 11, 9, 10, 10, 10, 8, 8, 8, 9, 9, 9, 10, 8, 8, 8,8, 14, 13, 14, 14, 15, 12,12, 12,14, 13, 11, 15, 15, 14, 14, 13, 15, 14, 14, 14, 15, 14, 15, 14, 14, 15, 14, 13, 13, 14, 11, 13, 11, 14, 8, 10, 13, 12, 11, 12,6]
cr=[]
ret=[]
for rec in self.recs:
cr.append(len(rec.values()))
ret.append(rec)
self.assertEqual(fields,cr)
class BadInputTreatmentTest(unittest.TestCase):
""" bibrecord - testing for bad input treatment """
### check bad input treatment ###
def test_wrong_attribute(self):
"""bibrecord - bad input : Has \'cde\' instead \'code\' in a subfield attribute"""
ws = bibrecord.cfg_bibrecord_warning_msgs
xml_error1 = """
<record>
<controlfield tag="001">33</controlfield>
<datafield tag="041" ind1="" ind2="">
<subfield code="a">eng</subfield>
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Doe, John</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="">
<subfield cde="a">On the foo and bar</subfield>
</datafield>
</record>
"""
(rec,st,e) = bibrecord.create_record(xml_error1,1,1)
ee=''
for i in e:
if type(i).__name__ == 'str':
if i.count(ws[3])>0:
ee = i
self.assertEqual(bibrecord.warning((3,'(field number: 4)')),ee)
def test_missing_attribute(self):
""" bibrecord - bad input : Missing attribute \"tag\" """
ws = bibrecord.cfg_bibrecord_warning_msgs
xml_error2 = """
<record>
<controlfield tag="001">33</controlfield>
<datafield ind1="" ind2="">
<subfield code="a">eng</subfield>
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Doe, John</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="">
<subfield code="a">On the foo and bar</subfield>
</datafield>
</record>
"""
(rec,st,e) = bibrecord.create_record(xml_error2,1,1)
ee=''
for i in e:
if type(i).__name__ == 'str':
if i.count(ws[1])>0:
ee = i
self.assertEqual(bibrecord.warning((1,'(field number(s): [2])')),ee)
def test_empty_datafield(self):
""" bibrecord - bad input : Datafield without any subfield """
ws = bibrecord.cfg_bibrecord_warning_msgs
xml_error3 = """
<record>
<controlfield tag="001">33</controlfield>
<datafield tag="041" ind1="" ind2="">
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Doe, John</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="">
<subfield code="a">On the foo and bar</subfield>
</datafield>
</record>
"""
(rec,st,e) = bibrecord.create_record(xml_error3,1,1)
ee=''
for i in e:
if type(i).__name__ == 'str':
if i.count(ws[8])>0:
ee = i
self.assertEqual(bibrecord.warning((8,'(field number: 2)')),ee)
def test_missing_tag(self):
"""bibrecord - bad input : Missing end \"tag\""""
ws = bibrecord.cfg_bibrecord_warning_msgs
xml_error4 = """
<record>
<controlfield tag="001">33</controlfield>
<datafield tag="041" ind1="" ind2="">
<subfield code="a">eng</subfield>
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Doe, John</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="">
<subfield code="a">On the foo and bar</subfield>
</record>
"""
(rec,st,e) = bibrecord.create_record(xml_error4,1,1)
ee = ''
for i in e:
if type(i).__name__ == 'str':
if i.count(ws[99])>0:
ee = i
self.assertEqual(bibrecord.warning((99,'(Tagname : datafield)')),ee)
class AccentedUnicodeLettersTest(unittest.TestCase):
""" bibrecord - testing accented UTF-8 letters """
def setUp(self):
xml_example_record = """
<record>
<controlfield tag="001">33</controlfield>
<datafield tag="041" ind1="" ind2="">
<subfield code="a">eng</subfield>
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Döè1, John</subfield>
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Doe2, J>ohn</subfield>
<subfield code="b">editor</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="1">
<subfield code="a">Пушкин</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="2">
<subfield code="a">On the foo and bar2</subfield>
</datafield>
</record>
"""
(self.rec, st, e) = bibrecord.create_record(xml_example_record,1,1)
def test_accented_unicode_characters(self):
"""bibrecord - accented Unicode letters"""
self.assertEqual(bibrecord.record_get_field_instances(self.rec, "100", "", ""),
[([('a', 'Döè1, John')], '', '', '', 3), ([('a', 'Doe2, J>ohn'), ('b', 'editor')], '', '', '', 4)])
self.assertEqual(bibrecord.record_get_field_instances(self.rec, "245", "", "1"),
[([('a', 'Пушкин')], '', '1', '', 5)])
class GettingFieldValuesTest(unittest.TestCase):
""" bibrecord - testing for getting field/subfield values """
def setUp(self):
xml_example_record = """
<record>
<controlfield tag="001">33</controlfield>
<datafield tag="041" ind1="" ind2="">
<subfield code="a">eng</subfield>
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Doe1, John</subfield>
</datafield>
<datafield tag="100" ind1="" ind2="">
<subfield code="a">Doe2, John</subfield>
<subfield code="b">editor</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="1">
<subfield code="a">On the foo and bar1</subfield>
</datafield>
<datafield tag="245" ind1="" ind2="2">
<subfield code="a">On the foo and bar2</subfield>
</datafield>
</record>
"""
(self.rec, st, e) = bibrecord.create_record(xml_example_record,1,1)
def test_get_field_instances(self):
"""bibrecord - getting field instances"""
self.assertEqual(bibrecord.record_get_field_instances(self.rec, "100", "", ""),
[([('a', 'Doe1, John')], '', '', '', 3), ([('a', 'Doe2, John'), ('b', 'editor')], '', '', '', 4)])
self.assertEqual(bibrecord.record_get_field_instances(self.rec, "", "", ""),
[('245', [([('a', 'On the foo and bar1')], '', '1', '', 5), ([('a', 'On the foo and bar2')], '', '2', '', 6)]), ('001', [([], '', '', '33', 1)]), ('100', [([('a', 'Doe1, John')], '', '', '', 3), ([('a', 'Doe2, John'), ('b', 'editor')], '', '', '', 4)]), ('041', [([('a', 'eng')], '', '', '', 2)])])
def test_get_field_values(self):
"""bibrecord - getting field values"""
self.assertEqual(bibrecord.record_get_field_values(self.rec, "100", "", "", "a"),
['Doe1, John', 'Doe2, John'])
self.assertEqual(bibrecord.record_get_field_values(self.rec, "100", "", "", "b"),
['editor'])
def test_get_subfield_values(self):
"""bibrecord - getting subfield values"""
fi1, fi2 = bibrecord.record_get_field_instances(self.rec, "100", "", "")
self.assertEqual(bibrecord.field_get_subfield_values(fi1, "b"), [])
self.assertEqual(bibrecord.field_get_subfield_values(fi2, "b"), ["editor"])
def create_test_suite():
"""Return test suite for the bibrecord module"""
return unittest.TestSuite((unittest.makeSuite(SanityTest,'test'),
unittest.makeSuite(SuccessTest,'test'),
unittest.makeSuite(BadInputTreatmentTest,'test'),
unittest.makeSuite(GettingFieldValuesTest,'test'),
unittest.makeSuite(AccentedUnicodeLettersTest,'test')))
if __name__ == '__main__':
unittest.TextTestRunner(verbosity=2).run(create_test_suite())

Event Timeline