def test_group_first(self):
from pyglottolog.util import group_first
for key, items in group_first([(1, 2), (1, 3)]):
self.assertEqual(key, 1)
self.assertEqual(len(list(items)), 2)
break
def test_group_first(self):
from pyglottolog.util import group_first
for key, items in group_first([(1, 2), (1, 3)]):
self.assertEqual(key, 1)
self.assertEqual(len(list(items)), 2)
break
def show_merges(self):
with self.connect() as conn:
cursor = conn.execute('SELECT hash, refid, filename, bibkey '
'FROM entry AS e WHERE EXISTS (SELECT 1 FROM entry '
'WHERE hash = e.hash AND refid != e.refid) '
'ORDER BY hash, refid DESC, filename, bibkey')
for hash, group in group_first(cursor):
self.print_group(conn, group)
new = self._merged_entry(self._entrygrp(conn, hash), raw=True)
cand = [(ri, self._merged_entry(self._entrygrp(conn, ri), raw=True))
for ri in unique(ri for hs, ri, fn, bk in group)]
old = min(cand, key=lambda p: distance(new, p[1]))[0]
print('-> %s\n' % old)
def show_splits(self):
with self.connect() as conn:
cursor = conn.execute('SELECT refid, hash, filename, bibkey '
'FROM entry AS e WHERE EXISTS (SELECT 1 FROM entry '
'WHERE refid = e.refid AND hash != e.hash) '
'ORDER BY refid, hash, filename, bibkey')
for refid, group in group_first(cursor):
self.print_group(conn, group)
old = self._merged_entry(self._entrygrp(conn, refid), raw=True)
cand = [(hs, self._merged_entry(self._entrygrp(conn, hs), raw=True))
for hs in unique(hs for ri, hs, fn, bk in group)]
new = min(cand, key=lambda p: distance(old, p[1]))[0]
print('-> %s\n' % new)
def show_merges(self):
with self.connect() as conn:
cursor = conn.execute(
'SELECT hash, refid, filename, bibkey '
'FROM entry AS e WHERE EXISTS (SELECT 1 FROM entry '
'WHERE hash = e.hash AND refid != e.refid) '
'ORDER BY hash, refid DESC, filename, bibkey')
for hash, group in group_first(cursor):
self.print_group(conn, group)
new = self._merged_entry(self._entrygrp(conn, hash), raw=True)
cand = [(ri,
self._merged_entry(self._entrygrp(conn, ri),
raw=True))
for ri in unique(ri for hs, ri, fn, bk in group)]
old = min(cand, key=lambda p: distance(new, p[1]))[0]
print('-> %s\n' % old)
def show_splits(self):
with self.connect() as conn:
cursor = conn.execute(
'SELECT refid, hash, filename, bibkey '
'FROM entry AS e WHERE EXISTS (SELECT 1 FROM entry '
'WHERE refid = e.refid AND hash != e.hash) '
'ORDER BY refid, hash, filename, bibkey')
for refid, group in group_first(cursor):
self.print_group(conn, group)
old = self._merged_entry(self._entrygrp(conn, refid), raw=True)
cand = [(hs,
self._merged_entry(self._entrygrp(conn, hs),
raw=True))
for hs in unique(hs for ri, hs, fn, bk in group)]
new = min(cand, key=lambda p: distance(old, p[1]))[0]
print('-> %s\n' % new)
def assign_ids(conn, verbose=False):
merged_entry, entrygrp = Database._merged_entry, Database._entrygrp
allhash, = conn.execute('SELECT NOT EXISTS (SELECT 1 FROM entry '
'WHERE hash IS NULL)').fetchone()
assert allhash
print('%d entries' % conn.execute('UPDATE entry SET id = NULL, srefid = refid').rowcount)
# resolve splits: srefid = refid only for entries from the most similar hash group
nsplit = 0
cursor = conn.execute('SELECT refid, hash, filename, bibkey FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE refid = e.refid AND hash != e.hash) '
'ORDER BY refid, hash, filename, bibkey')
for refid, group in group_first(cursor):
old = merged_entry(entrygrp(conn, refid), raw=True)
nsplit += len(group)
cand = [(hs, merged_entry(entrygrp(conn, hs), raw=True))
for hs in unique(hs for ri, hs, fn, bk in group)]
new = min(cand, key=lambda p: distance(old, p[1]))[0]
separated = conn.execute('UPDATE entry SET srefid = NULL WHERE refid = ? AND hash != ?',
(refid, new)).rowcount
if verbose:
for row in group:
print(row)
for ri, hs, fn, bk in group:
print('\t%r, %r, %r, %r' % hashfields(conn, fn, bk))
print('-> %s' % new)
print('%d: %d separated from %s\n' % (refid, separated, new))
print('%d splitted' % nsplit)
nosplits, = conn.execute('SELECT NOT EXISTS (SELECT 1 FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE srefid = e.srefid AND hash != e.hash))').fetchone()
assert nosplits
# resolve merges: id = srefid of the most similar srefid group
nmerge = 0
cursor = conn.execute('SELECT hash, srefid, filename, bibkey FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE hash = e.hash AND srefid != e.srefid) '
'ORDER BY hash, srefid DESC, filename, bibkey')
for hash, group in group_first(cursor):
new = merged_entry(entrygrp(conn, hash), raw=True)
nmerge += len(group)
cand = [(ri, merged_entry(entrygrp(conn, ri), raw=True))
for ri in unique(ri for hs, ri, fn, bk in group)]
old = min(cand, key=lambda p: distance(new, p[1]))[0]
merged = conn.execute('UPDATE entry SET id = ? WHERE hash = ? AND srefid != ?',
(old, hash, old)).rowcount
if verbose:
for row in group:
print(row)
for hs, ri, fn, bk in group:
print('\t%r, %r, %r, %r' % hashfields(conn, fn, bk))
print('-> %s' % old)
print('%s: %d merged into %d\n' % (hash, merged, old))
print('%d merged' % nmerge)
# unchanged entries
print('%d unchanged' % conn.execute('UPDATE entry SET id = srefid '
'WHERE id IS NULL AND srefid IS NOT NULL').rowcount)
nomerges, = conn.execute('SELECT NOT EXISTS (SELECT 1 FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE hash = e.hash AND id != e.id))').fetchone()
assert nomerges
# identified
print('%d identified (new/separated)' % conn.execute('UPDATE entry '
'SET id = (SELECT id FROM entry AS e WHERE e.hash = entry.hash AND e.id IS NOT NULL) '
'WHERE refid IS NULL AND id IS NULL AND EXISTS '
'(SELECT 1 FROM entry AS e WHERE e.hash = entry.hash AND e.id IS NOT NULL)').rowcount)
# assign new ids to hash groups of separated/new entries
nextid, = conn.execute('SELECT coalesce(max(refid), 0) + 1 FROM entry').fetchone()
cursor = conn.execute('SELECT hash FROM entry WHERE id IS NULL GROUP BY hash ORDER BY hash')
print('%d new ids (new/separated)' % conn.executemany('UPDATE entry SET id = ? WHERE hash = ?',
((id, hash) for id, (hash,) in enumerate(cursor, nextid))).rowcount)
assert allid(conn)
assert onetoone(conn)
# supersede relation
superseded, = conn.execute('SELECT count(*) FROM entry WHERE id != srefid').fetchone()
print('%d supersede pairs' % superseded)
def _show(self, sql):
with self.connect() as conn:
cursor = conn.execute(sql)
for hash, group in group_first(cursor):
self.print_group(conn, group)
print()
def assign_ids(conn, verbose=False):
merged_entry, entrygrp = Database._merged_entry, Database._entrygrp
allhash, = conn.execute('SELECT NOT EXISTS (SELECT 1 FROM entry '
'WHERE hash IS NULL)').fetchone()
assert allhash
print('%d entries' %
conn.execute('UPDATE entry SET id = NULL, srefid = refid').rowcount)
# resolve splits: srefid = refid only for entries from the most similar hash group
nsplit = 0
cursor = conn.execute(
'SELECT refid, hash, filename, bibkey FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE refid = e.refid AND hash != e.hash) '
'ORDER BY refid, hash, filename, bibkey')
for refid, group in group_first(cursor):
old = merged_entry(entrygrp(conn, refid), raw=True)
nsplit += len(group)
cand = [(hs, merged_entry(entrygrp(conn, hs), raw=True))
for hs in unique(hs for ri, hs, fn, bk in group)]
new = min(cand, key=lambda (hs, fields): distance(old, fields))[0]
separated = conn.execute(
'UPDATE entry SET srefid = NULL WHERE refid = ? AND hash != ?',
(refid, new)).rowcount
if verbose:
for row in group:
print(row)
for ri, hs, fn, bk in group:
print('\t%r, %r, %r, %r' % hashfields(conn, fn, bk))
print('-> %s' % new)
print('%d: %d separated from %s\n' % (refid, separated, new))
print('%d splitted' % nsplit)
nosplits, = conn.execute(
'SELECT NOT EXISTS (SELECT 1 FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE srefid = e.srefid AND hash != e.hash))'
).fetchone()
assert nosplits
# resolve merges: id = srefid of the most similar srefid group
nmerge = 0
cursor = conn.execute(
'SELECT hash, srefid, filename, bibkey FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE hash = e.hash AND srefid != e.srefid) '
'ORDER BY hash, srefid DESC, filename, bibkey')
for hash, group in group_first(cursor):
new = merged_entry(entrygrp(conn, hash), raw=True)
nmerge += len(group)
cand = [(ri, merged_entry(entrygrp(conn, ri), raw=True))
for ri in unique(ri for hs, ri, fn, bk in group)]
old = min(cand, key=lambda (ri, fields): distance(new, fields))[0]
merged = conn.execute(
'UPDATE entry SET id = ? WHERE hash = ? AND srefid != ?',
(old, hash, old)).rowcount
if verbose:
for row in group:
print(row)
for hs, ri, fn, bk in group:
print('\t%r, %r, %r, %r' % hashfields(conn, fn, bk))
print('-> %s' % old)
print('%s: %d merged into %d\n' % (hash, merged, old))
print('%d merged' % nmerge)
# unchanged entries
print('%d unchanged' %
conn.execute('UPDATE entry SET id = srefid '
'WHERE id IS NULL AND srefid IS NOT NULL').rowcount)
nomerges, = conn.execute(
'SELECT NOT EXISTS (SELECT 1 FROM entry AS e '
'WHERE EXISTS (SELECT 1 FROM entry WHERE hash = e.hash AND id != e.id))'
).fetchone()
assert nomerges
# identified
print('%d identified (new/separated)' % conn.execute(
'UPDATE entry '
'SET id = (SELECT id FROM entry AS e WHERE e.hash = entry.hash AND e.id IS NOT NULL) '
'WHERE refid IS NULL AND id IS NULL AND EXISTS '
'(SELECT 1 FROM entry AS e WHERE e.hash = entry.hash AND e.id IS NOT NULL)'
).rowcount)
# assign new ids to hash groups of separated/new entries
nextid, = conn.execute(
'SELECT coalesce(max(refid), 0) + 1 FROM entry').fetchone()
cursor = conn.execute(
'SELECT hash FROM entry WHERE id IS NULL GROUP BY hash ORDER BY hash')
print('%d new ids (new/separated)' %
conn.executemany('UPDATE entry SET id = ? WHERE hash = ?', (
(id, hash)
for id, (hash, ) in enumerate(cursor, nextid))).rowcount)
assert allid(conn)
assert onetoone(conn)
# supersede relation
superseded, = conn.execute(
'SELECT count(*) FROM entry WHERE id != srefid').fetchone()
print('%d supersede pairs' % superseded)
def _show(self, sql):
with self.connect() as conn:
cursor = conn.execute(sql)
for hash, group in group_first(cursor):
self.print_group(conn, group)
print()
def test_group_first():
key, items = next(util.group_first([(1, 2), (1, 3)]))
assert key, list(items) == (1, [(1, 2), (1, 3)])
