Coverage for python/lsst/pipe/base/graph.py : 35%

Hot-keys on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
1# This file is part of pipe_base.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (http://www.lsst.org).
6# See the COPYRIGHT file at the top-level directory of this distribution
7# for details of code ownership.
8#
9# This program is free software: you can redistribute it and/or modify
10# it under the terms of the GNU General Public License as published by
11# the Free Software Foundation, either version 3 of the License, or
12# (at your option) any later version.
13#
14# This program is distributed in the hope that it will be useful,
15# but WITHOUT ANY WARRANTY; without even the implied warranty of
16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17# GNU General Public License for more details.
18#
19# You should have received a copy of the GNU General Public License
20# along with this program. If not, see <http://www.gnu.org/licenses/>.
21from __future__ import annotations
23"""Module defining quantum graph classes and related methods.
25There could be different representations of the quantum graph depending
26on the client needs. Presently this module contains graph implementation
27which is based on requirements of command-line environment. In the future
28we could add other implementations and methods to convert between those
29representations.
30"""
32# "exported" names
33__all__ = ["QuantumGraph", "QuantumGraphTaskNodes", "QuantumIterData"]
35# -------------------------------
36# Imports of standard modules --
37# -------------------------------
38from itertools import chain
39from dataclasses import dataclass
40from typing import List, FrozenSet, Mapping
42# -----------------------------
43# Imports for other modules --
44# -----------------------------
45from .pipeline import TaskDef
46from .pipeTools import orderPipeline
47from lsst.daf.butler import DatasetRef, DatasetType, NamedKeyDict, Quantum
49# ----------------------------------
50# Local non-exported definitions --
51# ----------------------------------
53# ------------------------
54# Exported definitions --
55# ------------------------
58@dataclass
59class QuantumIterData:
60 """Helper class for iterating over quanta in a graph.
62 The `QuantumGraph.traverse` method needs to return topologically ordered
63 Quanta together with their dependencies. This class is used as a value
64 for the iterator, it contains enumerated Quantum and its dependencies.
65 """
67 __slots__ = ["index", "quantum", "taskDef", "dependencies"]
69 index: int
70 """Index of this Quantum, a unique but arbitrary integer."""
72 quantum: Quantum
73 """Quantum corresponding to a graph node."""
75 taskDef: TaskDef
76 """Task class to be run on this quantum, and corresponding label and
77 config.
78 """
80 dependencies: FrozenSet(int)
81 """Possibly empty set of indices of dependencies for this Quantum.
82 Dependencies include other nodes in the graph; they do not reflect data
83 already in butler (there are no graph nodes for those).
84 """
87@dataclass
88class QuantumGraphTaskNodes:
89 """QuantumGraphTaskNodes represents a bunch of nodes in an quantum graph
90 corresponding to a single task.
92 The node in quantum graph is represented by the `PipelineTask` and a
93 single `~lsst.daf.butler.Quantum` instance. One possible representation
94 of the graph is just a list of nodes without edges (edges can be deduced
95 from nodes' quantum inputs and outputs if needed). That representation can
96 be reduced to the list of PipelineTasks (or their corresponding TaskDefs)
97 and the corresponding list of Quanta. This class is used in this reduced
98 representation for a single task, and full `QuantumGraph` is a sequence of
99 tinstances of this class for one or more tasks.
101 Different frameworks may use different graph representation, this
102 representation was based mostly on requirements of command-line
103 executor which does not need explicit edges information.
104 """
106 taskDef: TaskDef
107 """Task defintion for this set of nodes."""
109 quanta: List[Quantum]
110 """List of quanta corresponding to the task."""
112 initInputs: Mapping[DatasetType, DatasetRef]
113 """Datasets that must be loaded or created to construct this task."""
115 initOutputs: Mapping[DatasetType, DatasetRef]
116 """Datasets that may be written after constructing this task."""
119class QuantumGraph(list):
120 """QuantumGraph is a sequence of `QuantumGraphTaskNodes` objects.
122 Typically the order of the tasks in the list will be the same as the
123 order of tasks in a pipeline (obviously depends on the code which
124 constructs graph).
126 Parameters
127 ----------
128 iterable : iterable of `QuantumGraphTaskNodes`, optional
129 Initial sequence of per-task nodes.
130 """
131 def __init__(self, iterable=None):
132 list.__init__(self, iterable or [])
133 self.initInputs = NamedKeyDict()
134 self.initIntermediates = NamedKeyDict()
135 self.initOutputs = NamedKeyDict()
137 initInputs: NamedKeyDict
138 """Datasets that must be provided to construct one or more Tasks in this
139 graph, and must be obtained from the data repository.
141 This is disjoint with both `initIntermediates` and `initOutputs`.
142 """
144 initIntermediates: NamedKeyDict
145 """Datasets that must be provided to construct one or more Tasks in this
146 graph, but are also produced after constructing a Task in this graph.
148 This is disjoint with both `initInputs` and `initOutputs`.
149 """
151 initOutputs: NamedKeyDict
152 """Datasets that are produced after constructing a Task in this graph,
153 and are not used to construct any other Task in this graph.
155 This is disjoint from both `initInputs` and `initIntermediates`.
156 """
158 def quanta(self):
159 """Iterator over quanta in a graph.
161 Quanta are returned in unspecified order.
163 Yields
164 ------
165 taskDef : `TaskDef`
166 Task definition for a Quantum.
167 quantum : `~lsst.daf.butler.Quantum`
168 Single quantum.
169 """
170 for taskNodes in self:
171 taskDef = taskNodes.taskDef
172 for quantum in taskNodes.quanta:
173 yield taskDef, quantum
175 def quantaAsQgraph(self):
176 """Iterator over quanta in a graph.
178 QuantumGraph containing individual quanta are returned.
180 Yields
181 ------
182 graph : `QuantumGraph`
183 """
184 for taskDef, quantum in self.quanta():
185 node = QuantumGraphTaskNodes(taskDef, [quantum],
186 quantum.initInputs, quantum.outputs)
187 graph = QuantumGraph([node])
188 yield graph
190 def countQuanta(self):
191 """Return total count of quanta in a graph.
193 Returns
194 -------
195 count : `int`
196 Number of quanta in a graph.
197 """
198 return sum(len(taskNodes.quanta) for taskNodes in self)
200 def traverse(self):
201 """Return topologically ordered Quanta and their dependencies.
203 This method iterates over all Quanta in topological order, enumerating
204 them during iteration. Returned `QuantumIterData` object contains
205 Quantum instance, its ``index`` and the ``index`` of all its
206 prerequsites (Quanta that produce inputs for this Quantum):
208 - the ``index`` values are generated by an iteration of a
209 QuantumGraph, and are not intrinsic to the QuantumGraph
210 - during iteration, each ID will appear in index before it ever
211 appears in dependencies.
213 Yields
214 ------
215 quantumData : `QuantumIterData`
216 """
218 def orderedTaskNodes(graph):
219 """Return topologically ordered task nodes.
221 Yields
222 ------
223 nodes : `QuantumGraphTaskNodes`
224 """
225 # Tasks in a graph are probably topologically sorted already but there
226 # is no guarantee for that. Just re-construct Pipeline and order tasks
227 # in a pipeline using existing method.
228 nodesMap = {id(item.taskDef): item for item in graph}
229 pipeline = orderPipeline([item.taskDef for item in graph])
230 for taskDef in pipeline:
231 yield nodesMap[id(taskDef)]
233 index = 0
234 outputs = {} # maps (DatasetType.name, dataId) to its producing quantum index
235 for nodes in orderedTaskNodes(self):
236 for quantum in nodes.quanta:
238 # Find quantum dependencies (must be in `outputs` already)
239 prereq = []
240 for dataRef in chain.from_iterable(quantum.predictedInputs.values()):
241 # if data exists in butler then `id` is not None
242 if dataRef.id is None:
243 # Get the base name if this is a component
244 name, component = dataRef.datasetType.nameAndComponent()
245 key = (name, dataRef.dataId)
246 try:
247 prereq.append(outputs[key])
248 except KeyError:
249 # The Quantum that makes our inputs is not in the graph,
250 # this could happen if we run on a "split graph" which is
251 # usually just one quantum. Check for number of Quanta
252 # in a graph and ignore error if it's just one.
253 # TODO: This code has to be removed or replaced with
254 # something more generic
255 if not (len(self) == 1 and len(self[0].quanta) == 1):
256 raise
258 # Update `outputs` with this quantum outputs
259 for dataRef in chain.from_iterable(quantum.outputs.values()):
260 key = (dataRef.datasetType.name, dataRef.dataId)
261 outputs[key] = index
263 yield QuantumIterData(index=index, quantum=quantum, taskDef=nodes.taskDef,
264 dependencies=frozenset(prereq))
265 index += 1