Skip to content

Commit

Permalink
Change status Literals to Enums
Browse files Browse the repository at this point in the history
  • Loading branch information
@eigerx
eigerx committed Sep 18, 2024
1 parent 1e0d6d1 commit 9ab5584
Showing 1 changed file with 90 additions and 64 deletions.
154 changes: 90 additions & 64 deletions python/lsst/pipe/base/quantum_provenance_graph.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -43,7 +43,8 @@
import logging
import uuid
from collections.abc import Iterator, Sequence
from typing import TYPE_CHECKING, ClassVar, Literal, NamedTuple, TypeAlias, TypedDict, cast
from enum import Enum
from typing import TYPE_CHECKING, ClassVar, Literal, NamedTuple, TypedDict, cast

import networkx
import pydantic
Expand Down Expand Up @@ -132,7 +133,28 @@ class PrerequisiteDatasetKey(NamedTuple):
"""


QuantumRunStatus: TypeAlias = Literal["failed", "successful", "logs_missing", "blocked", "metadata_missing"]
class QuantumRunStatus(Enum):
"""Enum describing the status of a quantum-run collection combination.
Possible Statuses
-----------------
metadata_missing = -3: Metadata is missing for this quantum in this run.
It is impossible to tell whether execution of this quantum was
attempted due to missing metadata.
logs_missing = -2: Logs are missing for this quantum in this run. It was
attempted, but it is impossible to tell if it succeeded or failed due
to missing logs.
failed = -1: Attempts to execute the quantum failed in this run.
blocked = 0: This run does not include an executed version of this
quantum because an upstream task failed.
successful = 1: This quantum was executed successfully in this run.
"""

metadata_missing = -3
logs_missing = -2
failed = -1
blocked = 0
successful = 1


class QuantumRun(pydantic.BaseModel):
Expand All @@ -144,23 +166,48 @@ class QuantumRun(pydantic.BaseModel):

status: QuantumRunStatus = "metadata_missing"
"""The status of the quantum in that run.
"""


class QuantumInfoStatus(Enum):
"""The status of a quantum (a particular task run on a particular dataID)
across all runs.
Possible Statuses
-----------------
`failed`: Attempts to execute the quantum failed in this run.
`successful`: This quantum was executed successfully in this run.
`logs_missing`: Logs are missing for this quantum in this run. It was
attempted, but it is impossible to tell if it succeeded or failed due
to missing logs.
`blocked`: This run does not include an executed version of this quantum
because an upstream task failed.
`metadata_missing`: Metadata is missing for this quantum in this run. It is
impossible to tell whether execution of this quantum was attempted due
to missing metadata.
wonky = -3: The overall state of this quantum reflects inconsistencies or
is difficult to discern. There are a few specific ways to enter a wonky
state; it is impossible to exit and requires human intervention to
proceed with processing.
Currently, a quantum enters a wonky state for one of three reasons:
- Its `QuantumInfoStatus` exits a successful state. Something that
initially succeeded fails on
- A `QuantumRun` is missing logs.
- There are multiple runs associated with a dataset which comes up in a
findFirst search. This means that a dataset which will be used as an
input data product for further processing has heterogeneous inputs,
which may have had different inputs or a different data-query.
failed = -2: These quanta were attempted and failed. Failed quanta have
logs and no metadata.
unknown = -1: These are quanta which do not have any metadata associated
with processing, but for which it is impossible to tell the status due
to an additional absence of logs. Quanta which had not been processed
at all would reflect this state, as would quanta which were
conceptualized in the construction of the quantum graph but later
identified to be unneccesary or erroneous (deemed NoWorkFound by the
Science Pipelines).
blocked = 0: The quantum is not able to execute because its inputs are
missing due to an upstream failure. Blocked quanta are distinguished
from failed quanta by being successors of failed quanta in the graph.
All the successors of blocked quanta are also marked as blocked.
successful = 1: Attempts at executing this quantum were successful.
"""


QuantumInfoStatus: TypeAlias = Literal["successful", "wonky", "blocked", "unknown", "failed"]
wonky = -3
failed = -2
unknown = -1
blocked = 0
successful = 1


class QuantumInfo(TypedDict):
Expand All @@ -181,35 +228,6 @@ class QuantumInfo(TypedDict):
status: QuantumInfoStatus
"""The overall status of the quantum. Note that it is impossible to exit a
wonky state.
Possible Statuses
-----------------
`successful`: Attempts at executing this quantum were successful.
`wonky`: The overall state of this quantum reflects inconsistencies or is
difficult to discern. There are a few specific ways to enter a wonky
state; it is impossible to exit and requires human intervention to
proceed with processing.
Currently, a quantum enters a wonky state for one of three reasons:
- Its `QuantumInfoStatus` exits a successful state. Something that
initially succeeded fails on
- A `QuantumRun` is missing logs.
- There are multiple runs associated with a dataset which comes up in a
findFirst search. This means that a dataset which will be used as an
input data product for further processing has heterogeneous inputs,
which may have had different inputs or a different data-query.
`blocked`: The quantum is not able to execute because its inputs are
missing due to an upstream failure. Blocked quanta are distinguished
from failed quanta by being successors of failed quanta in the graph.
All the successors of blocked quanta are also marked as blocked.
`unknown`: These are quanta which do not have any metadata associated
with processing, but for which it is impossible to tell the status due
to an additional absence of logs. Quanta which had not been processed
at all would reflect this state, as would quanta which were
conceptualized in the construction of the quantum graph but later
identified to be unneccesary or erroneous (deemed `NoWorkFound` by the
Science Pipelines).
`failed`: These quanta were attempted and failed. Failed quanta have logs
and no metadata.
"""

recovered: bool
Expand Down Expand Up @@ -257,7 +275,35 @@ def _validate(self) -> DatasetRun:
return self


DatasetInfoStatus: TypeAlias = Literal["visible", "shadowed", "predicted_only", "unsuccessful", "cursed"]
class DatasetInfoStatus(Enum):
"""Status of the the DatasetType-dataID pair over all runs.
Possible Statuses
-----------------
cursed: The dataset was the result of an unsuccessful quantum and was
visible in the output collection anyway. These are flagged as
cursed so that they may be caught before they become inputs to
further processing.
unsuccessful: The dataset was not produced. These are the results of
failed or blocked quanta.
predicted_only: The dataset was predicted, and was not visible in any
run, but was the successor of a successful quantum. These datasets are
the result of pipelines NoWorkFound cases, in which a dataset is
predicted in the graph but found to not be necessary in processing.
shadowed: The dataset exists but is not queryable in a find_first
search. This could mean that the version of this dataset which is
passed as an input to further processing is not in the collections
given. A shadowed dataset will not be used as an input to further
processing.
visible: The dataset is queryable in a find_first search. This means
that it can be used as an input by subsequent tasks and processing.
"""

cursed = -2
unsuccessful = -1
predicted_only = 0
shadowed = 1
visible = 2


class DatasetInfo(TypedDict):
Expand All @@ -276,26 +322,6 @@ class DatasetInfo(TypedDict):

status: DatasetInfoStatus
"""Overall status of the dataset.
Possible Statuses
-----------------
`visible`: The dataset is queryable in a find_first search. This means
that it can be used as an input by subsequent tasks and processing.
`shadowed`: The dataset exists but is not queryable in a find_first
search. This could mean that the version of this dataset which is
passed as an input to further processing is not in the collections
given. A `shadowed` dataset will not be used as an input to further
processing.
`predicted_only`: The dataset was predicted, and was not visible in any
run, but was the successor of a successful quantum. These datasets are
the result of pipelines `NoWorkFound` cases, in which a dataset is
predicted in the graph but found to not be necessary in processing.
`unsuccessful`: The dataset was not produced. These are the results of
failed or blocked quanta.
`cursed`: The dataset was the result of an unsuccessful quantum and was
visible in the output collection anyway. These are flagged as
`cursed` so that they may be caught before they become inputs to
further processing.
"""

messages: list[str]
Expand Down

0 comments on commit 9ab5584

Please sign in to comment.