def test():
from hypergraphs.semirings import LazySort
from hypergraphs.semirings.sampling.lazy2 import Sample
from arsenal.iterextras import take
# Check that we do not have duplicate alignments
dups = set()
for x in subsequence('abcabc', 'abcb', default_cost, LazySort):
x = str(x.data)
assert x not in dups
dups.add(x)
print('number of alignments', len(dups))
K = 5
print()
print(f'Top K={K}')
print('========')
for x in take(K, subsequence('abcabc', 'abcb', default_cost, LazySort)):
pretty(x.score, x.data)
print()
print(f'Samples K={K}')
print('============')
for score, data in take(
K, subsequence('abcabc', 'abcb', default_cost, Sample)):
pretty(score, data)
print()
def test_kleene():
print('[kleene]')
from arsenal.iterextras import take
assert (list(take(10, map(''.join, kleene('01'))))
== ['', '0', '1', '00', '01', '10', '11', '000', '001', '010'])
assert (list(map(''.join, kleene('01', n=3)))
== ['', '0', '1', '00', '01', '10', '11', '000', '001', '010',
'011', '100', '101', '110', '111'])
assert (list(map(''.join, kleene('a', n=4)))
== ['', 'a', 'aa', 'aaa', 'aaaa'])
assert (list(map(''.join, kleene('', n=4)))
== [''])
assert length(powerset(range(5))) == 2**5
assert length(permute(range(5))) == factorial(5)
def job_summary(jobs):
def cnt(name, column):
n = len(jobs[jobs[column]])
pct(name, n, len(jobs))
def pct(name, n, m):
print '%-7s %.2f (%s/%s)' % (name, n / m, n, m)
print
print yellow % 'Job status'
print yellow % '=========='
cnt('running', 'running')
cnt('done', 'done')
cnt('log', 'log_exists')
qstat = load_qstat('tmp/jobs')
print
print yellow % 'SGE status'
print yellow % '=========='
if qstat.empty:
print red % 'Failed to find anything in qstat.'
return
for k, d in qstat.groupby('status'):
pct(k, len(d), len(qstat))
if 0:
# use kl-divergence to show what are relevant features of jobs still running.
from arsenal.math.featureselection import kl_filter
from arsenal.iterextras import take
list(
take(
50,
kl_filter([(x.done, [
'%s=%s' % (k, v)
for (k, v) in x.iteritems() if k.startswith('args_')
]) for _, x in jobs.iterrows()],
feature_count_cuttoff=2)))
return qstat
def main(f):
os.system('mkdir -p tmp/bibhacks')
for entry in take(10, find_entries(f)):
entry.pprint()
print
def main():
import numpy as np
import itertools
from arsenal.iterextras import take
from itertools import count
# weighted tuples are the idea as a path weight with backpointers; our weighted
# tuple copies the tuple, so it is inefficient compared to the lazier
# backpointer variant.
class WeightedTuple:
def __init__(self, w, *key):
self.key = key
self.w = w
def __lt__(self, other):
return (self.w, self.key) < (other.w, other.key)
def __eq__(self, other):
return (self.w, self.key) == (other.w, other.key)
def __mul__(self, other):
return LWeightedTuple(self.w * other.w, self, other)
def __add__(self, other):
return LWeightedTuple(self.w + other.w, self, other)
def __iter__(self):
return iter((self.w, self.key))
def __repr__(self):
return repr((self.w, self.key))
class LWeightedTuple(WeightedTuple):
"WeightedTuple with lazy concatenation of keys."
def __init__(self, w, a, b):
self.w = w
self.a = a
self.b = b
@property
def key(self):
return self.a.key + self.b.key
def wprod(xs):
return np.product([WeightedTuple(x, x) for x in xs])
def wsum(xs):
return np.sum([WeightedTuple(x, x) for x in xs])
def check(iters):
for p in [np.product, np.sum, tuple, wprod]:
# enumerate and sort; not lazy
want = list(sorted(p(x) for x in itertools.product(*iters)))
got = list(sorted_product(p, *iters))
print()
print('product operator:', p.__name__)
print('GOT:', got)
#if got != want:
print('WANT:', want)
assert got == want
print('pass.')
print('===========')
check([
(.1, .4, 0.5),
(0.09, 0.11, 0.8),
(0.111, .3, 0.6),
])
print('===========')
check([
(1, 2, 3),
(4, 7, 11),
])
print('===========')
check([
(0.01, .4, 0.5),
(0.11, 0.8),
(0.6, ),
])
print('===========')
check([
(1, 2, 3, 100),
(4, 7, 9),
(14, 17, 19),
(24, 27, 29),
])
print('===========')
a = (3**i for i in count(1))
b = (4**i for i in count(1))
c = (5**i for i in count(1))
for s, x in take(20, sorted_product(wsum, a, b, c)):
print(s, x)
def main(f):
os.system('mkdir -p tmp/bibhacks')
for entry in take(10, find_entries(f)):
entry.pprint()
print()