pipelines/sdk/python/kfp/deprecated/dsl/_for_loop.py

258 lines
9.8 KiB
Python

# Copyright 2021 The Kubeflow Authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import re
from typing import Any, Dict, List, Optional, Tuple, Union
from kfp.deprecated import dsl
from kfp.deprecated.dsl import _pipeline_param
ItemList = List[Union[int, float, str, Dict[str, Any]]]
class LoopArguments(dsl.PipelineParam):
"""Class representing the arguments that are looped over in a ParallelFor
loop in the KFP DSL.
This doesn't need to be instantiated by the end user, rather it will
be automatically created by a ParallelFor ops group.
"""
LOOP_ITEM_NAME_BASE = 'loop-item'
LOOP_ITEM_PARAM_NAME_BASE = 'loop-item-param'
# number of characters in the code which is passed to the constructor
NUM_CODE_CHARS = 8
LEGAL_SUBVAR_NAME_REGEX = re.compile(r'^[a-zA-Z_][0-9a-zA-Z_]*$')
@classmethod
def _subvar_name_is_legal(cls, proposed_variable_name: str):
return re.match(cls.LEGAL_SUBVAR_NAME_REGEX,
proposed_variable_name) is not None
def __init__(self,
items: Union[ItemList, dsl.PipelineParam],
code: str,
name_override: Optional[str] = None,
op_name: Optional[str] = None,
*args,
**kwargs):
"""LoopArguments represent the set of items to loop over in a
ParallelFor loop.
This class shouldn't be instantiated by the user but rather is created by
_ops_group.ParallelFor.
Args:
items: List of items to loop over. If a list of dicts then, all
dicts must have the same keys and every key must be a legal Python
variable name.
code: A unique code used to identify these loop arguments. Should
match the code for the ParallelFor ops_group which created these
LoopArguments. This prevents parameter name collisions.
"""
if name_override is None:
super().__init__(name=self._make_name(code), *args, **kwargs)
else:
super().__init__(
name=name_override, op_name=op_name, *args, **kwargs)
if not isinstance(items, (list, tuple, dsl.PipelineParam)):
raise TypeError(
'Expected list, tuple, or PipelineParam, got {}.'.format(
type(items)))
if isinstance(items, tuple):
items = list(items)
if isinstance(items, list) and isinstance(items[0], dict):
subvar_names = set(items[0].keys())
for item in items:
if not set(item.keys()) == subvar_names:
raise ValueError(
'If you input a list of dicts then all dicts should have the same keys. '
'Got: {}.'.format(items))
# then this block creates loop_args.variable_a and loop_args.variable_b
for subvar_name in subvar_names:
if not self._subvar_name_is_legal(subvar_name):
raise ValueError(
"Tried to create subvariable named {} but that's not a legal Python variable "
'name.'.format(subvar_name))
setattr(
self, subvar_name,
LoopArgumentVariable(
loop_args_name=self.name,
this_variable_name=subvar_name,
loop_args_op_name=self.op_name,
loop_args=self,
))
self.items_or_pipeline_param = items
self.referenced_subvar_names = []
@classmethod
def from_pipeline_param(cls, param: dsl.PipelineParam) -> 'LoopArguments':
return LoopArguments(
items=param,
code=None,
name_override=param.name + '-' + cls.LOOP_ITEM_NAME_BASE,
op_name=param.op_name,
value=param.value,
)
def __getattr__(self, item):
# this is being overridden so that we can access subvariables of the
# LoopArguments (i.e.: item.a) without knowing the subvariable names ahead
# of time
self.referenced_subvar_names.append(item)
return LoopArgumentVariable(
loop_args_name=self.name,
this_variable_name=item,
loop_args_op_name=self.op_name,
loop_args=self,
)
def to_list_for_task_yaml(self):
if isinstance(self.items_or_pipeline_param, (list, tuple)):
return self.items_or_pipeline_param
else:
raise ValueError(
'You should only call this method on loop args which have list items, '
'not pipeline param items.')
@classmethod
def _make_name(cls, code: str):
"""Make a name for this parameter.
Code is a
"""
return '{}-{}'.format(cls.LOOP_ITEM_PARAM_NAME_BASE, code)
@classmethod
def name_is_loop_argument(cls, param_name: str) -> bool:
"""Return True if the given parameter name looks like a loop argument.
Either it came from a withItems loop item or withParams loop
item.
"""
return cls.name_is_withitems_loop_argument(param_name) \
or cls.name_is_withparams_loop_argument(param_name)
@classmethod
def name_is_withitems_loop_argument(cls, param_name: str) -> bool:
"""Return True if the given parameter name looks like it came from a
loop arguments parameter."""
return (cls.LOOP_ITEM_PARAM_NAME_BASE + '-') in param_name
@classmethod
def name_is_withparams_loop_argument(cls, param_name: str) -> bool:
"""Return True if the given parameter name looks like it came from a
withParams loop item."""
return ('-' + cls.LOOP_ITEM_NAME_BASE) in param_name
@classmethod
def remove_loop_item_base_name(cls, param_name: str) -> str:
"""Removes the last LOOP_ITEM_NAME_BASE from the end of param name."""
if ('-' + cls.LOOP_ITEM_NAME_BASE) in param_name:
# Split from the right, so that it handles multi-level nested args.
return param_name.rsplit('-' + cls.LOOP_ITEM_NAME_BASE, 1)[0]
return param_name
class LoopArgumentVariable(dsl.PipelineParam):
"""Represents a subvariable for loop arguments.
This is used for cases where we're looping over maps, each of
which contains several variables.
"""
SUBVAR_NAME_DELIMITER = '-subvar-'
def __init__(
self,
loop_args_name: str,
this_variable_name: str,
loop_args_op_name: Optional[str],
# For backward compatible, add loop_args as an optional argument.
# Ideally, this should replace loop_args_name and loop_args_op_name.
loop_args: Optional[LoopArguments] = None,
):
"""
If the user ran:
with dsl.ParallelFor([{'a': 1, 'b': 2}, {'a': 3, 'b': 4}]) as item:
...
Then there's be one LoopArgumentsVariable for 'a' and another for 'b'.
Args:
loop_args_name: The name of the LoopArguments object that this is
a subvariable to.
this_variable_name: The name of this subvariable, which is the name
of the dict key that spawned this subvariable.
loop_args_op_name: The name of the op that produced the loop arguments.
loop_args: Optional; The LoopArguments object this subvariable is based on.
"""
super().__init__(
name=self.get_name(
loop_args_name=loop_args_name,
this_variable_name=this_variable_name),
op_name=loop_args_op_name,
)
self.loop_args = loop_args
@property
def items_or_pipeline_param(
self) -> Union[ItemList, _pipeline_param.PipelineParam]:
return self.loop_args.items_or_pipeline_param
@classmethod
def get_name(cls, loop_args_name: str, this_variable_name: str) -> str:
"""Get the name.
Args:
loop_args_name: the name of the loop args parameter that this
LoopArgsVariable is attached to.
this_variable_name: the name of this LoopArgumentsVariable subvar.
Returns:
The name of this loop args variable.
"""
return '{}{}{}'.format(loop_args_name, cls.SUBVAR_NAME_DELIMITER,
this_variable_name)
@classmethod
def name_is_loop_arguments_variable(cls, param_name: str) -> bool:
"""Return True if the given parameter name looks like it came from a
LoopArgumentsVariable."""
return re.match('.+%s.+' % cls.SUBVAR_NAME_DELIMITER,
param_name) is not None
@classmethod
def parse_loop_args_name_and_this_var_name(cls, t: str) -> Tuple[str, str]:
"""Get the loop arguments param name and this subvariable name from the
given parameter name."""
m = re.match(
'(?P<loop_args_name>.*){}(?P<this_var_name>.*)'.format(
cls.SUBVAR_NAME_DELIMITER), t)
if m is None:
return None
else:
return m.groupdict()['loop_args_name'], m.groupdict(
)['this_var_name']
@classmethod
def get_subvar_name(cls, t: str) -> str:
"""Get the subvariable name from a given LoopArgumentsVariable
parameter name."""
out = cls.parse_loop_args_name_and_this_var_name(t)
if out is None:
raise ValueError("Couldn't parse variable name: {}".format(t))
return out[1]