def __init__(self, column_def, agg_func, name=None, fname=None, astype=None):
"""
Args:
column_def: List of (or a single) column definitions, as accepted by Column.
agg_func: List of (or a single) row-wise aggregation function (will be passed to
Panda's groupby.agg())
name: List of strings of names for the column_defintions. Must be of
same length as column_def, or be None, in which case the names
default to the string representation of the each column definition.
fname: List of strings of names for the aggregation functions. Must be of
same length as agg_func or be None, in which case the names
default to the string representation of the each aggregation function.
astype: List of pandas dtypes, which gets passed to Column together with the
column definitions. Must be of same length as column_def or be None,
in which case no casting will be performed.
Names for the resulting column reductions are of the format `columnname_functionname`.
"""
# make list of column reductions from arguments --> each ColumnReduction needs a Column(definition) and an agg_func
column_def = util.make_list(column_def)
agg_func = util.make_list(agg_func)
astype = util.make_list(astype)
if len(astype) == 1:
astype = astype*len(column_def)
if len(astype) != len(column_def):
raise ValueError("astype must be a datatype, or a list of datatypes of same length as column_def")
column_reductions = [ColumnReduction(Column(defn,at), func)
for (defn, at), func in product(zip(column_def, astype), agg_func)]
# make list of names from the arguments --> same length as the list of column reductions above!
name = [str(c) for c in column_def] if name is None else util.make_list(name)
if len(name) != len(column_def):
raise ValueError("name and column_def must be same length, or name must be None.")
fname = [str(a) for a in agg_func] if fname is None else util.make_list(fname)
if len(fname) != len(agg_func):
raise ValueError("fname and agg_func must be same length, or fname must be None.")
# if there's only on agg_func, you can override the function-naming-schema,
# by passing fname=False; then we only use the name
if len(agg_func) == 1 and fname == [False]:
column_names = name
else:
column_names = ['%s_%s'%(cn, fn) for cn, fn in product(name, fname)]
ColumnFunction.__init__(self, column_reductions, column_names)
def map_inputs(self):
kwargs = {}
args = []
if hasattr(self, 'inputs_mapping'):
inputs_mapping = util.make_list(self.inputs_mapping)
if len(self.inputs) < len(inputs_mapping):
raise ValueError('Too many inputs_mappings')
for input, mapping in zip(self.inputs, inputs_mapping):
result = input.get_result()
if isinstance(mapping, dict):
# pass through any missing keys, so {} is the identity
# do it first so that inputs_mapping overrides keys
for k in set(result.keys()).\
difference(set(mapping.keys())):
kwargs[k] = result[k]
for k in mapping:
if mapping[k] is not None:
kwargs[mapping[k]] = result[k]
elif isinstance(mapping, list):
if len(mapping) > len(result):
raise ValueError("More keywords than results")
for kw, r in zip(mapping, result):
if kw is not None:
kwargs[kw] = r
elif isinstance(mapping, string_types):
kwargs[mapping] = input.get_result()
elif mapping is None: # drop Nones
pass
else:
raise ValueError(
'Input mapping is neither dict nor str: %s' % mapping)
mapped_inputs = len(inputs_mapping)
else:
mapped_inputs = 0
for i in range(mapped_inputs, len(self.inputs)):
result = self.inputs[i].get_result()
# without a mapping we handle two cases
# when the result is a dict merge it with a global dict
if isinstance(result, dict):
# but do not override
kwargs.update(
{k: v
for k, v in result.items() if k not in kwargs})
# otherwise use it as a positional argument
else:
args.append(result)
return args, kwargs
def parse_workflows(workflows):
steps = []
for s in workflows:
if s.endswith('.yaml'):
steps += serialize.load(s)
else:
modulename = s[:s.split('(')[0].rfind('.')]
exec('import %s' % modulename)
if s.find('(') < 0:
s += '()'
ss = util.make_list(eval('%s' % s))
logging.info('Loaded %s with %s leaf steps' % (s, len(ss)))
steps += ss
return steps
def map_inputs(self):
kwargs = {}
args = []
if hasattr(self, 'inputs_mapping'):
inputs_mapping = util.make_list(self.inputs_mapping)
if len(self.inputs) < len(inputs_mapping):
raise ValueError('Too many inputs_mappings')
for input, mapping in zip(self.inputs, inputs_mapping):
result = input.get_result()
if isinstance(mapping, dict):
# pass through any missing keys, so {} is the identity
# do it first so that inputs_mapping overrides keys
for k in set(result.keys()).difference(set(mapping.keys())):
kwargs[k] = result[k]
for k in mapping:
if mapping[k] is not None:
kwargs[mapping[k]] = result[k]
elif isinstance(mapping, basestring):
kwargs[mapping] = input.get_result()
elif mapping is None: # drop Nones
pass
else:
raise ValueError('Input mapping is neither dict nor str: %s' % mapping)
mapped_inputs = len(inputs_mapping)
else:
mapped_inputs = 0
for i in range(mapped_inputs, len(self.inputs)):
result = self.inputs[i].get_result()
# without a mapping we handle two cases
# when the result is a dict merge it with a global dict
if isinstance(result, dict):
# but do not override
kwargs.update({k:v for k,v in result.iteritems() if k not in kwargs})
# otherwise use it as a positional argument
else:
args.append(result)
return args, kwargs
drain.PATH = os.path.abspath(args.path)
elif drain.PATH is None:
raise ValueError(
'Must pass path argument or set DRAINPATH environment variable')
steps = []
for s in args.steps.split(';'):
if s.endswith('.yaml'):
steps += serialize.load(s)
else:
modulename, fname = s.split('::')
mod = importlib.import_module(modulename)
s_attr = getattr(mod, fname)
# if s is callable, it should return a collection of Steps
# otherwise assume it is a collection of Steps
ss = util.make_list(
s_attr() if hasattr(s_attr, '__call__') else s_attr)
logging.info('Loaded %s with %s leaf steps' % (s, len(ss)))
steps += ss
if args.Drakeinput is None and os.path.exists('Drakefile'):
args.Drakeinput = 'Drakefile'
drakeinput = os.path.abspath(args.Drakeinput) if args.Drakeinput else None
workflow = drake.to_drakefile(steps,
preview=args.preview,
debug=args.debug,
input_drakefile=drakeinput,
bindir=os.path.dirname(__file__))
if not args.preview:
with open(args.drakeoutput, 'w') as drakefile:
step.OUTPUTDIR = os.path.abspath(args.outputdir)
drain.yaml.configure()
steps = []
for s in args.steps.split(';'):
if s.endswith('.yaml'):
steps += drain.yaml.load(s)
else:
print s
modulename, fname = s.split('::')
mod = importlib.import_module(modulename)
s = getattr(mod, fname)
# if s is callable, it should return a collection of Steps
# otherwise assume it is a collection of Steps
s = util.make_list(s() if hasattr(s, '__call__') else s)
steps += s
if args.Drakeinput is None and os.path.exists('Drakefile'):
args.Drakeinput = 'Drakefile'
drakeinput = os.path.abspath(args.Drakeinput) if args.Drakeinput else None
workflow = drake.to_drakefile(steps, preview=args.preview, debug=args.debug, input_drakefile=drakeinput)
if not args.preview:
with open(args.drakeoutput, 'w') as drakefile:
drakefile.write(workflow)
else:
sys.stdout.write(workflow)
drake_args = list(drake_args) if drake_args is not None else []
def dapply(self,
fn,
pairwise=False,
symmetric=True,
diagonal=False,
block=None,
**kwargs):
"""
Apply function to each step object in the index
Args:
fn: function to apply. If a list then each function is applied
pairwise: whether to apply the function to pairs of steps
symmetric, diagonal, block: passed to apply_pairwise when pairwise=True
kwargs: a keyword arguments to pass to each function. Arguments
with list value are grid searched using util.dict_product.
Returns: a StepFrame or StepSeries
"""
search_keys = [
k for k, v in kwargs.items() if isinstance(v, list) and len(v) > 1
]
functions = util.make_list(fn)
search = list(product(functions, util.dict_product(kwargs)))
results = []
for fn, kw in search:
if not pairwise:
r = self.index.to_series().apply(lambda step: fn(step, **kw))
else:
r = apply_pairwise(self,
fn,
symmetric=symmetric,
diagonal=diagonal,
block=block,
**kw)
name = [] if len(functions) == 1 else [fn.__name__]
name += util.dict_subset(kw, search_keys).values()
if isinstance(r, pd.DataFrame):
columns = pd.MultiIndex.from_tuples(
[tuple(name + util.make_list(c)) for c in r.columns])
r.columns = columns
else:
r.name = tuple(name)
results.append(r)
if len(results) > 1:
result = pd.concat(results, axis=1)
# get subset of parameters that were searched over
column_names = [] if len(functions) == 1 else [None]
column_names += search_keys
column_names += [None
] * (len(result.columns.names) - len(column_names))
result.columns.names = column_names
return StepFrame(result)
else:
result = results[0]
if isinstance(result, pd.DataFrame):
return StepFrame(result)
else:
result.name = functions[0].__name__
return StepSeries(result)
args.preview = True
step.OUTPUTDIR = os.path.abspath(args.outputdir)
drain.yaml.configure()
steps = []
for s in args.steps.split(';'):
if s.endswith('.yaml'):
steps += drain.yaml.load(s)
else:
modulename, fname = s.split('::')
mod = importlib.import_module(modulename)
s_attr = getattr(mod, fname)
# if s is callable, it should return a collection of Steps
# otherwise assume it is a collection of Steps
ss = util.make_list(s_attr() if hasattr(s_attr, '__call__') else s_attr)
logging.info('Loaded %s with %s leaf steps' % (s, len(ss)))
steps += ss
if args.Drakeinput is None and os.path.exists('Drakefile'):
args.Drakeinput = 'Drakefile'
drakeinput = os.path.abspath(args.Drakeinput) if args.Drakeinput else None
workflow = drake.to_drakefile(steps, preview=args.preview, debug=args.debug, input_drakefile=drakeinput)
if not args.preview:
with open(args.drakeoutput, 'w') as drakefile:
logging.info('Writing drake workflow %s' % args.drakeoutput)
drakefile.write(workflow)
else:
sys.stdout.write(workflow)
def __init__(self,
column_def,
agg_func,
name=None,
fname=None,
astype=None):
"""
Args:
column_def: List of (or a single) column definitions, as accepted by Column.
agg_func: List of (or a single) row-wise aggregation function (will be passed to
Panda's groupby.agg())
name: List of strings of names for the column_defintions. Must be of
same length as column_def, or be None, in which case the names
default to the string representation of the each column definition.
fname: List of strings of names for the aggregation functions. Must be of
same length as agg_func or be None, in which case the names
default to the string representation of the each aggregation function.
astype: List of pandas dtypes, which gets passed to Column together with the
column definitions. Must be of same length as column_def or be None,
in which case no casting will be performed.
Names for the resulting column reductions are of the format `columnname_functionname`.
"""
# make list of column reductions from arguments --> each ColumnReduction needs a Column(definition) and an agg_func
column_def = util.make_list(column_def)
agg_func = util.make_list(agg_func)
astype = util.make_list(astype)
if len(astype) == 1:
astype = astype * len(column_def)
if len(astype) != len(column_def):
raise ValueError(
"astype must be a datatype, or a list of datatypes of same length as column_def"
)
column_reductions = [
ColumnReduction(Column(defn, at), func)
for (defn, at), func in product(zip(column_def, astype), agg_func)
]
# make list of names from the arguments --> same length as the list of column reductions above!
name = [str(c)
for c in column_def] if name is None else util.make_list(name)
if len(name) != len(column_def):
raise ValueError(
"name and column_def must be same length, or name must be None."
)
fname = [str(a)
for a in agg_func] if fname is None else util.make_list(fname)
if len(fname) != len(agg_func):
raise ValueError(
"fname and agg_func must be same length, or fname must be None."
)
# if there's only on agg_func, you can override the function-naming-schema,
# by passing fname=False; then we only use the name
if len(agg_func) == 1 and fname == [False]:
column_names = name
else:
column_names = [
'%s_%s' % (cn, fn) for cn, fn in product(name, fname)
]
ColumnFunction.__init__(self, column_reductions, column_names)
