def test_match(self):
t = PrefTree()
urls = [
('aab/ced',True),
('aab/ru',False),
('bcadr/ste',False),
('aab/t/est',True),
('aab/t/le',True),
('aab/stts',False),
]
for u, s in urls:
t.add_url(u, s)
t, pruned = t.prune(min_rate=0.9, min_children=2,min_urls=1)
c, s, b = t.match_with_branch('aab/t/lu')
self.assertEqual((c,s),(2,2))
self.assertEqual(''.join(b), 'aab/t/*')
c, s, b = t.match_with_branch('aab/xx')
self.assertEqual((c,s),(0,0))
self.assertEqual(''.join(b), 'aab/<unk>')
c, s, b = t.match_with_branch('aab/ced')
self.assertEqual((c,s),(1,1))
self.assertEqual(''.join(b), 'aab/ced')
def test_prune(self):
t = PrefTree()
with self.assertRaises(ValueError):
t.prune(min_urls=0)
for url, success in [
('arxiv.org/pdf/1410.1234', True),
('arxiv.org/pdf/1409.1094', True),
('arxiv.org/pdf/1201.5480', True),
('arxiv.org/pdf/1601.01234', True),
('arxiv.org/pdf/1602.01i34', False), # oops
]:
t.add_url(url, success)
t, pruned = t.prune(min_rate=0.75,min_children=2,min_urls=1)
self.assertEqual(len(t.urls()), 1)
self.assertTrue(t.has_wildcard())
self.assertEqual(t.match('arxiv.org/pdf/1784.1920'), (5,4))
self.assertEqual(t.match('arxiv.org/pdf/2340.0124'), (0,0))
self.assertTrue(t.predict_success('arxiv.org/pdf/1784.1920', threshold=0.6, min_urls=3))
t.print_as_tree()
def test_prune_failures(self):
t = PrefTree()
for url, success in [
('sciencedirect.com/paper1.pdf', False),
('sciencedirect.com/paper2.pdf', False),
('sciencedirect.com/paper3.pdf', False),
('mysciencework.com/paper9.pdf', True),
('mysciencework.com/paper8.pdf', False)]:
t.add_url(url, success)
t, pruned = t.prune(min_rate=0.95, min_children=2, min_urls=2)
self.assertEqual(t.match('sciencedirect.com/paper4.pdf'), (3,0))
self.assertFalse(t.predict_success('sciencedirect.com/paper4.pdf'))
def test_prune_with_reverse(self):
t = PrefTree()
for url, success in [
('researchgate.net/publication/233865122_uriset', False),
('researchgate.net/publication/143874230_albtedru', False),
('researchgate.net/publication/320748374_kelbcad', False),
('researchgate.net/publication/233865122_uriset.pdf', True),
('researchgate.net/publication/143874230_albtedru.pdf', True),
('researchgate.net/publication/320748374_kelbcad.pdf', True),
('onlinelibrary.wiley.com/wol1/doi/10.1002/anie.200800037.abstract', False),
('onlinelibrary.wiley.com/wol1/doi/10.1002/anie.200800037.pdf', False)]:
t.add_url(url, success)
t.print_as_tree()
t, pruned = t.prune(reverse=True)
t.print_as_tree()
self.assertTrue(t.check_sanity())
for u, c, s in t.urls():
print flatten(u), c, s
def test_prune_empty_tree(self):
t = PrefTree()
t.prune() # should not raise anything