def test_simple():
if not os.path.exists('./noaa_data'):
p = params(clevel=5, storage='./noaa_data')
t = Table([], dshape='{f0: int, f1:int, f2:int, f3:float}', params=p)
# TODO: chunkwise copy
t.append(adapter[:])
t.commit()
else:
t = open('ctable://noaa_data')
print '--------------------------------------'
print 'mean', mean(t, 'f3')
print 'std', std(t, 'f2')
print '--------------------------------------'
qs1 = select(t, lambda x: x > 80000, 'f0')
qs2 = select2(t, lambda x, y: x > y, ['f0', 'f1'])
result = t[qs1]
def test_simple():
if not os.path.exists('./noaa_data'):
p = params(clevel=5, storage='./noaa_data')
t = Table([], dshape='{f0: int, f1:int, f2:int, f3:float}', params=p)
# TODO: chunkwise copy
t.append(adapter[:])
t.commit()
else:
t = open('ctable://noaa_data')
print '--------------------------------------'
print 'mean', mean(t, 'f3')
print 'std', std(t, 'f2')
print '--------------------------------------'
qs1 = select(t, lambda x: x > 80000, 'f0')
qs2 = select2(t, lambda x,y: x > y, ['f0', 'f1'])
result = t[qs1]
def test_mean(self, data):
from blaze import mean
assert compute(mean(t['amount']), data) == x['amount'].mean()
def test_mean(self, data):
from blaze import mean
assert compute(mean(t['amount']), data) == x['amount'].mean()
def groupmeans(data, groups, numbers, cutoff=0.01, quantile=0.95, minsize=None):
"""
Yields the significant differences in average between every pair of
groups and numbers.
Parameters
----------
data : blaze data object
groups : non-empty iterable containing category column names in data
numbers : non-empty iterable containing numeric column names in data
cutoff : ignore anything with prob > cutoff.
cutoff=None ignores significance checks, speeding it up a LOT.
quantile : number that represents target improvement. Defaults to .95.
The ``diff`` returned is the % impact of everyone moving to the 95th
percentile
minsize : each group should contain at least minsize values.
If minsize=None, automatically set the minimum size to
1% of the dataset, or 10, whichever is larger.
"""
if minsize is None:
minsize = max(data.nrows / 100, 10)
means = {col: data[col].mean() for col in numbers}
results = []
for group in groups:
agg = {number: bz.mean(data[number]) for number in numbers}
agg["#"] = bz.count(data)
ave = bz.by(data[group], **agg).sort("#", ascending=False)
ave = bz.into(pd.DataFrame, ave)
ave.index = ave[group]
sizes = ave["#"]
# Each group should contain at least minsize values
biggies = sizes[sizes >= minsize].index
# ... and at least 2 groups overall, to compare.
if len(biggies) < 2:
continue
for number in numbers:
if number == group:
continue
sorted_cats = ave[number][biggies].dropna().sort_values()
if len(sorted_cats) < 2:
continue
lo = bz.into(list, data[number][data[group] == sorted_cats.index[0]])
hi = bz.into(list, data[number][data[group] == sorted_cats.index[-1]])
_, prob = ttest_ind(np.ma.masked_array(lo, np.isnan(lo)), np.ma.masked_array(hi, np.isnan(hi)))
if prob > cutoff:
continue
results.append(
{
"group": group,
"number": number,
"prob": prob,
"gain": (sorted_cats.iloc[-1] / means[number] - 1)[0],
"biggies": ave.ix[biggies][number],
"means": ave[[number, "#"]].sort_values(by=number),
}
)
results = pd.DataFrame(results)
if len(results) > 0:
results = results.set_index(["group", "number"])
return results
def blaze_tutorial():
accounts = bl.Symbol('accounts',
'var * {id: int, name: string, amount: int}')
deadbeats = accounts[accounts.amount < 0].name
list_ = [[1, 'Alice', 100],
[2, 'Bob', -200],
[3, 'Charlie', 300],
[4, 'Denis', 400],
[5, 'Edith', -500]]
print(list(bl.compute(deadbeats, list_)))
df_ = bl.DataFrame(list_, columns=['id', 'name', 'amount'])
print(bl.compute(deadbeats, df_))
bl_df_dir = dir(df_)
df_ = pd.DataFrame(list_, columns=['id', 'name', 'amount'])
print(df_[df_.amount < 0].name)
pd_df_dir = dir(df_)
print(len(bl_df_dir), len(pd_df_dir))
print(len([d for d in bl_df_dir if d in pd_df_dir]))
print([d for d in bl_df_dir if d not in pd_df_dir])
print([d for d in pd_df_dir if d not in bl_df_dir])
df_ = bl.Data([(1, 'Alice', 100),
(2, 'Bob', -200),
(3, 'Charlie', 300),
(4, 'Denis', 400),
(5, 'Edith', -500)],
fields=['id', 'name', 'balance'])
print(repr(df_))
print(repr(df_[df_.balance < 0]))
print(repr(df_[df_.balance < 0].name))
print(list(df_[df_.balance < 0].name))
iris = bl.Data('examples/iris.csv')
print(repr(iris))
iris = bl.Data('sqlite:///examples/iris.db::iris')
print(repr(iris))
print(repr(bl.by(iris.species, min=iris.petal_width.min(),
max=iris.petal_width.max())))
result = bl.by(iris.species, min=iris.petal_width.min(),
max=iris.petal_width.max())
print(odo(result, bl.DataFrame))
print(odo(result, pd.DataFrame))
### odo has weird issue with unicode filenames, apparently...
name = 'output.csv'
print(odo(result, bl.CSV(name)))
print(repr(iris.sepal_length.mean()))
print(repr(bl.mean(iris.sepal_length)))
print(repr(bl.by(iris.species, shortest=iris.petal_length.min(),
longest=iris.petal_length.max(),
average=iris.petal_length.mean())))
print(repr(iris.head()))
iris = bl.transform(iris, sepal_ratio=iris.sepal_length / iris.sepal_width,
petal_ratio=iris.petal_length / iris.petal_width)
print(repr(iris.head()))
versicolor = iris[iris.species.like('%versicolor')]
print(repr(versicolor))
print((len(versicolor), len(versicolor.fields)))
print(repr(iris.relabel(petal_length='PETAL-LENGTH',
petal_width='PETAL-WIDTH')))
pd_df = pd.DataFrame({'name': ['Alice', 'Bob', 'Joe', 'Bob'],
'amount': [100, 200, 300, 400],
'id': [1, 2, 3, 4]})
# put the `df` DataFrame odo a Blaze Data object
bl_df = bl.DataFrame(pd_df)
bl_dt = bl.Data(pd_df)
print(repr(pd_df.amount * 2))
print(repr(bl_df.amount * 2))
print(repr(bl_dt.amount * 2))
print(repr(pd_df[['id', 'amount']]))
print(repr(bl_df[['id', 'amount']]))
print(repr(bl_dt[['id', 'amount']]))
print(repr(pd_df[pd_df.amount > 300]))
print(repr(bl_df[bl_df.amount > 300]))
print(repr(bl_dt[bl_dt.amount > 300]))
print(repr(pd_df.groupby('name').amount.mean()))
print(repr(pd_df.groupby(['name', 'id']).amount.mean()))
print(repr(bl_df.groupby('name').amount.mean()))
print(repr(bl_df.groupby(['name', 'id']).amount.mean()))
print(repr(bl.by(bl_dt.name, amount=bl_dt.amount.mean())))
print(repr(bl.by(bl.merge(bl_dt.name, bl_dt.id),
amount=bl_dt.amount.mean())))
#pd.merge(pd_df, pd_df2, on='name')
#bl.join(bl_dt, bl_dt2, 'name')
print(repr(pd_df.amount.map(lambda x: x + 1)))
print(repr(bl_df.amount.map(lambda x: x + 1)))
print(repr(bl_dt.amount.map(lambda x: x + 1, 'int64')))
print(repr(pd_df.rename(columns={'name': 'alias', 'amount': 'dollars'})))
print(repr(bl_df.rename(columns={'name': 'alias', 'amount': 'dollars'})))
print(repr(bl_dt.relabel(name='alias', amount='dollars')))
print(repr(pd_df.drop_duplicates()))
print(repr(bl_df.drop_duplicates()))
print(repr(bl_dt.distinct()))
print(repr(pd_df.name.drop_duplicates()))
print(repr(bl_df.name.drop_duplicates()))
print(repr(bl_dt.name.distinct()))
print(repr(pd_df.amount.mean()))
print(repr(bl_df.amount.mean()))
print(repr(bl_dt.amount.mean()))
print(repr(pd_df))
print(repr(bl_df))
print(repr(bl_dt))
print(repr(pd_df.amount.value_counts()), '\n')
print(repr(bl_df.amount.value_counts()), '\n')
print(repr(bl_dt.amount.count_values()), '\n')
print(repr(pd_df.dtypes), '\n')
print(repr(bl_df.dtypes), '\n')
print(repr(bl_df.columns), '\n')
print(repr(bl_dt.dshape), '\n')
print(repr(pd_df.amount.dtypes), '\n')
print(repr(bl_df.amount.dtypes), '\n')
print(repr(bl_dt.amount.dshape), '\n')
print(type(pd_df), type(bl_df), type(bl_dt), '\n')
os.remove('output.csv')
for fn_ in glob.glob('*.csv.gz'):
os.remove(fn_)
return
b = carray(np.arange(N, dtype='f8')+1)
t = ctable((a, b), ('f0', 'f1'), rootdir='example1', mode='w')
t.flush()
#------------------------------------------------------------------------
from time import time
print '-------------------'
t = blaze.open('ctable://example1')
# Using chunked blaze array we can optimize for IO doing the sum
# operations chunkwise from disk.
t0 = time()
print blaze.mean(t, 'f0')
print "Chunked mean", round(time()-t0, 6)
# Using NumPy is just going to through the iterator protocol on
# carray which isn't going to efficient.
t0 = time()
print np.mean(t.data.ca['f0'])
print "NumPy mean", round(time()-t0, 6)
print '==================='
t0 = time()
#assert blaze.std(t, 'f0') == 28867.513458037913
print blaze.std(t, 'f0')
print "Chunked std", round(time()-t0, 6)
def groupmeans(data, groups, numbers,
cutoff=.01,
quantile=.95,
min_size=None):
'''
Yields the significant differences in average between every pair of
groups and numbers.
Parameters
----------
data : blaze data object
groups : non-empty iterable containing category column names in data
numbers : non-empty iterable containing numeric column names in data
cutoff : ignore anything with prob > cutoff.
cutoff=None ignores significance checks, speeding it up a LOT.
quantile : number that represents target improvement. Defaults to .95.
The ``diff`` returned is the % impact of everyone moving to the 95th
percentile
min_size : each group should contain at least min_size values.
If min_size=None, automatically set the minimum size to
1% of the dataset, or 10, whichever is larger.
'''
if min_size is None:
# compute nrows, bz.compute(data.nrows) doesn't work for sqlite
min_size = max(bz.into(int, data.nrows) / 100, 10)
# compute mean of each number column
means = {col: bz.into(float, data[col].mean()) for col in numbers}
# pre-create aggregation expressions (mean, count)
agg = {number: bz.mean(data[number]) for number in numbers}
for group in groups:
agg['#'] = data[group].count()
ave = bz.by(data[group], **agg).sort('#', ascending=False)
ave = bz.into(pd.DataFrame, ave)
ave.index = ave[group]
sizes = ave['#']
# Each group should contain at least min_size values
biggies = sizes[sizes >= min_size].index
# ... and at least 2 groups overall, to compare.
if len(biggies) < 2:
continue
for number in numbers:
if number == group:
continue
sorted_cats = ave[number][biggies].dropna().sort_values()
if len(sorted_cats) < 2:
continue
sohi = sorted_cats.index[-1]
solo = sorted_cats.index[0]
# If sorted_cats.index items are of numpy type, then
# convert them to native type, skip conversion for unicode, str
# See https://github.com/blaze/blaze/issues/1461
if isinstance(solo, np.generic):
solo, sohi = solo.item(), sohi.item()
lo = bz.into(list, data[number][data[group] == solo])
hi = bz.into(list, data[number][data[group] == sohi])
_, prob = ttest_ind(
np.ma.masked_array(lo, np.isnan(lo)),
np.ma.masked_array(hi, np.isnan(hi))
)
# All results will be returned by default
# Up to the user to ignore or show insignificant results
# Uncomment below two lines to return only significant results
# if prob > cutoff:
# continue
yield ({
'group': group,
'number': number,
'prob': float(prob),
'gain': sorted_cats.iloc[-1] / means[number] - 1,
'biggies': ave.ix[biggies][number].to_dict(),
'means': ave[[number, '#']].sort_values(by=number).reset_index().to_dict(
orient='records'),
})
def groupmeans(data, groups, numbers,
cutoff=.01,
quantile=.95,
min_size=None):
'''
Yields the significant differences in average between every pair of
groups and numbers.
Parameters
----------
data : blaze data object
groups : non-empty iterable containing category column names in data
numbers : non-empty iterable containing numeric column names in data
cutoff : ignore anything with prob > cutoff.
cutoff=None ignores significance checks, speeding it up a LOT.
quantile : number that represents target improvement. Defaults to .95.
The ``diff`` returned is the % impact of everyone moving to the 95th
percentile
min_size : each group should contain at least min_size values.
If min_size=None, automatically set the minimum size to
1% of the dataset, or 10, whichever is larger.
'''
if min_size is None:
# compute nrows, bz.compute(data.nrows) doesn't work for sqlite
min_size = max(bz.into(int, data.nrows) / 100, 10)
# compute mean of each number column
means = {col: bz.into(float, data[col].mean()) for col in numbers}
# pre-create aggregation expressions (mean, count)
agg = {number: bz.mean(data[number]) for number in numbers}
for group in groups:
agg['#'] = data[group].count()
ave = bz.by(data[group], **agg).sort('#', ascending=False)
ave = bz.into(pd.DataFrame, ave)
ave.index = ave[group]
sizes = ave['#']
# Each group should contain at least min_size values
biggies = sizes[sizes >= min_size].index
# ... and at least 2 groups overall, to compare.
if len(biggies) < 2:
continue
for number in numbers:
if number == group:
continue
sorted_cats = ave[number][biggies].dropna().sort_values()
if len(sorted_cats) < 2:
continue
sohi = sorted_cats.index[-1]
solo = sorted_cats.index[0]
# If sorted_cats.index items are of numpy type, then
# convert them to native type, skip conversion for unicode, str
# See https://github.com/blaze/blaze/issues/1461
if isinstance(solo, np.generic):
solo, sohi = solo.item(), sohi.item()
lo = bz.into(list, data[number][data[group] == solo])
hi = bz.into(list, data[number][data[group] == sohi])
_, prob = ttest_ind(
np.ma.masked_array(lo, np.isnan(lo)),
np.ma.masked_array(hi, np.isnan(hi))
)
if prob > cutoff:
continue
yield ({
'group': group,
'number': number,
'prob': float(prob),
'gain': sorted_cats.iloc[-1] / means[number] - 1,
'biggies': ave.ix[biggies][number].to_dict(),
'means': ave[[number, '#']].sort_values(by=number).reset_index().to_dict(
orient='records'),
})
