def from_df(df, type_overrides={}):
"""Loads data store from pandas DataFrame.
- df: By default, we will load all numeric type as BucketizedFloatColumn
and all other type as string_cols unless instructed by type_overrides.
- type_overrides: A dict of col->type which allows us to override the types of
each column. Valid override types are string, bucketized_float and float (raw_float).
"""
try:
import numpy
import pandas
except ImportError:
raise ImportError('Please install numpy and pandas.')
d = _DataStore()
raw_float_cols = set([
k for k, v in type_overrides.iteritems()
if v == 'float' or v == 'raw_float'
])
string_cols = set(
[k for k, v in type_overrides.iteritems() if v == 'string'])
for col in df.select_dtypes(include=[numpy.number]).keys():
if col in raw_float_cols:
d.add_raw_float_col(col, list(df[col]))
elif col not in string_cols:
d.add_bucketized_float_col(col, list(df[col]))
for col in df.select_dtypes(exclude=[numpy.number]).keys():
d.add_string_col(col, list(df[col]))
return DataStore(d)
def from_df(df, type_overrides={}):
"""Loads data store from pandas DataFrame.
- df: By default, we will load all numeric type as BucketizedFloatColumn
and all other type as string_cols unless instructed by type_overrides.
- type_overrides: A dict of col->type which allows us to override the types of
each column. Valid override types are string, bucketized_float and float (raw_float).
"""
try:
import numpy
import pandas
except ImportError:
raise ImportError('Please install numpy and pandas.')
d = _DataStore()
raw_float_cols = set([k for k, v in type_overrides.iteritems()
if v == 'float' or v == 'raw_float'])
string_cols = set([k for k, v in type_overrides.iteritems()
if v == 'string'])
for col in df.select_dtypes(include=[numpy.number]).keys():
if col in raw_float_cols:
d.add_raw_float_col(col, list(df[col]))
elif col not in string_cols:
d.add_bucketized_float_col(col, list(df[col]))
for col in df.select_dtypes(exclude=[numpy.number]).keys():
d.add_string_col(col, list(df[col]))
return DataStore(d)
def slice(self, index):
d = _DataStore()
for key, value in self.iteritems():
if type(value) is StringColumn:
d.add_string_col(key, [value[i] for i in index])
elif type(value) is BucketizedFloatColumn:
d.add_bucketized_float_col(key, [value[i] for i in index])
elif type(value) is RawFloatColumn:
d.add_raw_float_col(key, [value[i] for i in index])
return DataStore(d)
def slice(self, index):
d = _DataStore()
for key, value in self.iteritems():
if type(value) is StringColumn:
d.add_string_col(key, [value[i] for i in index])
elif type(value) is BucketizedFloatColumn:
d.add_bucketized_float_col(key, [value[i] for i in index])
elif type(value) is RawFloatColumn:
d.add_raw_float_col(key, [value[i] for i in index])
return DataStore(d)
def from_tsvs(tsvs, bucketized_float_cols=[], string_cols=[], raw_float_cols=[]):
"""Loads data from tsvs.
Inputs:
tsvs: Blocks of tsvs, among which only the first contains header.
bucketized_float_cols: Float columns that will be bucketized. All features will be bucketized.
string_cols: String cols.
raw_float_cols: Float columns that are loaded raw. Target columns are usually not bucketized.
"""
d = _DataStore()
d.load_tsv(tsvs,
bucketized_float_cols=bucketized_float_cols,
string_cols=string_cols,
raw_float_cols=raw_float_cols)
return DataStore(d)
def from_dict(bucketized_float_cols={}, string_cols={}, raw_float_cols={}):
"""Loads data from dict of columns.
bucketized_float_cols: Float columns that will be bucketized. All features will be bucketized.
string_cols: String cols.
raw_float_cols: Float columns that are loaded raw. Target columns are usually not bucketized.
"""
d = _DataStore()
for key, value in bucketized_float_cols.iteritems():
d.add_bucketized_float_col(key, value)
for key, value in string_cols.iteritems():
d.add_string_col(key, value)
for key, value in raw_float_cols.iteritems():
d.add_raw_float_col(key, value)
return DataStore(d)
def from_dict(bucketized_float_cols={}, string_cols={}, raw_float_cols={}):
"""Loads data from dict of columns.
bucketized_float_cols: Float columns that will be bucketized. All features will be bucketized.
string_cols: String cols.
raw_float_cols: Float columns that are loaded raw. Target columns are usually not bucketized.
"""
d = _DataStore()
for key, value in bucketized_float_cols.iteritems():
d.add_bucketized_float_col(key, value)
for key, value in string_cols.iteritems():
d.add_string_col(key, value)
for key, value in raw_float_cols.iteritems():
d.add_raw_float_col(key, value)
return DataStore(d)
def from_tsvs(tsvs,
bucketized_float_cols=[],
string_cols=[],
raw_float_cols=[]):
"""Loads data from tsvs.
Inputs:
tsvs: Blocks of tsvs, among which only the first contains header.
bucketized_float_cols: Float columns that will be bucketized. All features will be bucketized.
string_cols: String cols.
raw_float_cols: Float columns that are loaded raw. Target columns are usually not bucketized.
"""
d = _DataStore()
d.load_tsv(tsvs,
bucketized_float_cols=bucketized_float_cols,
string_cols=string_cols,
raw_float_cols=raw_float_cols)
return DataStore(d)
