oleander-kedro 0.0.1__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,7 @@
1
+ .venv/
2
+ .pytest_cache/
3
+ .ruff_cache/
4
+ dist/
5
+ *.egg-info/
6
+ manifest.json
7
+ logs/
@@ -0,0 +1 @@
1
+ 3.10
@@ -0,0 +1,150 @@
1
+ Metadata-Version: 2.5
2
+ Name: oleander-kedro
3
+ Version: 0.0.1
4
+ Summary: Declare Oleander metadata on ordinary Kedro pipelines
5
+ Requires-Python: >=3.10
6
+ Requires-Dist: croniter<7,>=6
7
+ Requires-Dist: email-validator<3,>=2.3
8
+ Requires-Dist: kedro<1.8,>=1.7
9
+ Requires-Dist: tzdata>=2025.2
10
+ Description-Content-Type: text/markdown
11
+
12
+ # oleander-kedro
13
+
14
+ Public Python package for declaring Oleander metadata on standard Kedro pipelines.
15
+ Requires **Python >=3.10** and **Kedro >=1.7,<1.8**. Intended for distribution on
16
+ PyPI; it has no Oleander SDK or cloud dependencies.
17
+
18
+ ## Author pipelines
19
+
20
+ Add this package to your Kedro project's dependencies, then decorate pipeline
21
+ factories:
22
+
23
+ ```python
24
+ from kedro.pipeline import node, pipeline
25
+ from oleander_kedro import run_as, schedule
26
+
27
+
28
+ @run_as("sales-pipeline@iam.application.oleander.dev")
29
+ @schedule(cron="0 * * * *", timezone="UTC")
30
+ def create_pipeline(**kwargs):
31
+ return pipeline([node(print, "params:message", None, name="print_message")])
32
+ ```
33
+
34
+ Register pipelines using the usual `register_pipelines()` function, optionally
35
+ calling `find_pipelines(raise_errors=True)`. Authors use conventional Kedro:
36
+
37
+ ```sh
38
+ uv run kedro run --pipelines orders
39
+ ```
40
+
41
+ This package installs **no command or Kedro hooks**. Normal `kedro run` executes
42
+ ordinary Kedro pipelines without applying schedules, selecting identities, or
43
+ injecting intervals. Kedro's catalog and parameter behavior remains unchanged.
44
+ Decorators still validate their arguments and factory return types; invalid
45
+ declarations fail early.
46
+
47
+ ## Declarations
48
+
49
+ - `@schedule(cron=..., timezone=...)` requires basic five-field cron syntax:
50
+ numeric fields with `*`, `,`, `-`, and `/`, plus an explicit IANA timezone.
51
+ For example, `0 9 * * 1-5` runs on weekdays at 09:00. Named months/weekdays
52
+ and extensions such as `L5`, `L`, `W`, `#`, `H`, `R`, and `?` are rejected
53
+ at declaration time, as are macros, seconds, and year expressions. Whitespace
54
+ is normalized. Croniter validates field values and expression structure;
55
+ its day-of-month/day-of-week convention is OR. Oleander's scheduler must
56
+ use the same interpretation.
57
+ Ranges must be ascending: use `22-23,0-2` instead of `22-2` for hours that
58
+ cross midnight. List entries must not repeat or overlap values, including
59
+ overlaps introduced by steps. Sunday may be `0` or `7`, but `0,7` is a
60
+ duplicate. A single full-week range such as `0-7` remains valid.
61
+ - `@run_as(email)` requires one email address. UUIDs, principal names, display
62
+ names, and address lists are rejected. Surrounding whitespace is stripped and
63
+ the address is lowercased and normalized to match Oleander's email convention.
64
+ Validation checks syntax locally, without DNS queries or delivery checks.
65
+ It declares the desired principal for **scheduled** execution only. It grants
66
+ no permissions and never changes the identity of a manual run. The example
67
+ application-principal email is a placeholder; it need not exist for local tests.
68
+ - Either decorator order works. Duplicate declarations are errors. Factories
69
+ preserve their signatures, names, and docstrings and return ordinary Kedro
70
+ `Pipeline` objects.
71
+ - Metadata belongs to the exact returned object. Copying, composing, filtering,
72
+ or namespacing a pipeline does not carry declarations forward. Decorate the
73
+ factory returning the final scheduled pipeline.
74
+
75
+ Oleander's private discovery/runtime validates the complete registry: pipelines
76
+ must declare both decorators or neither, and the same scheduled object cannot
77
+ appear under multiple registry names. Undecorated pipelines are discoverable
78
+ and can be run manually. Conventional Kedro does not enforce these Oleander
79
+ registry rules.
80
+
81
+ The deployment service must resolve the declared email to an existing principal
82
+ within the organization linked to the GitHub installation, authorize the
83
+ repository-to-principal binding, and store the resolved UUID with the deployment.
84
+ Missing, ambiguous, or unauthorized principals must fail deployment. Scheduled
85
+ runs use that stored UUID so email changes cannot silently change the execution
86
+ identity. This resolution and authorization belongs to Oleander's service; neither
87
+ Python package performs it. Principal membership and access must be rechecked at
88
+ execution time.
89
+
90
+ ## Metadata API
91
+
92
+ Private platform tools read metadata through the public, side-effect-free API:
93
+
94
+ ```python
95
+ from oleander_kedro import get_pipeline_metadata
96
+
97
+ metadata = get_pipeline_metadata(create_pipeline())
98
+ print(metadata.schedule.cron)
99
+ print(metadata.run_as)
100
+ ```
101
+
102
+ The returned `Declaration` and nested `Schedule` are frozen dataclasses.
103
+ `Declaration.run_as` contains the normalized email string.
104
+ Undecorated pipelines return a `Declaration` with both fields set to `None`.
105
+ `ConfigurationError` subclasses `ValueError`.
106
+
107
+ ## Development
108
+
109
+ From this directory:
110
+
111
+ ```sh
112
+ uv sync --locked
113
+ uv run pytest
114
+ uv run ruff check .
115
+ uv run ruff format --check .
116
+ uv build
117
+ ```
118
+
119
+ `.python-version` selects Python 3.10. Tests cover declaration validation and
120
+ real `kedro run` invocations with plain Python nodes and no SDK credentials.
121
+
122
+ From the repository root, also run the cross-language cron compatibility tests:
123
+
124
+ ```sh
125
+ uv sync --locked --project kedro/oleander-kedro-cli
126
+ yarn exec vitest run workflows/kedro/manifest-compatibility.test.ts
127
+ ```
128
+
129
+ These tests require the repository's Node dependencies and `uv`. They use the
130
+ same cron fixtures as the Python tests, generate manifests with the actual
131
+ Python decorators and CLI, and validate them with the TypeScript worker's
132
+ `parseManifest()`. No cloud services are involved.
133
+
134
+ The separate [private CLI project](../oleander-kedro-cli/README.md) depends on this
135
+ package and owns discovery, runtime interval injection, and the `oleander-kedro`
136
+ command.
137
+ It is not part of this distribution.
138
+
139
+ ## Publishing
140
+
141
+ The [publishing workflow](../../.github/workflows/publish-oleander-kedro.yml)
142
+ builds and publishes this extension when changes under `kedro/oleander-kedro-ext/`
143
+ reach `main`. It uses the existing GitHub Actions `PYPI_TOKEN` secret and skips
144
+ already-published artifacts, following the Python SDK's release pattern.
145
+
146
+ To release a new version, update `version` in `pyproject.toml` and refresh this
147
+ project's `uv.lock`. Update the private CLI's exact extension dependency and its
148
+ lockfile to the same version, then merge to `main`. The builder release workflow
149
+ also rebuilds its bundled packages when these directories change. The private
150
+ CLI is not published to PyPI.
@@ -0,0 +1,139 @@
1
+ # oleander-kedro
2
+
3
+ Public Python package for declaring Oleander metadata on standard Kedro pipelines.
4
+ Requires **Python >=3.10** and **Kedro >=1.7,<1.8**. Intended for distribution on
5
+ PyPI; it has no Oleander SDK or cloud dependencies.
6
+
7
+ ## Author pipelines
8
+
9
+ Add this package to your Kedro project's dependencies, then decorate pipeline
10
+ factories:
11
+
12
+ ```python
13
+ from kedro.pipeline import node, pipeline
14
+ from oleander_kedro import run_as, schedule
15
+
16
+
17
+ @run_as("sales-pipeline@iam.application.oleander.dev")
18
+ @schedule(cron="0 * * * *", timezone="UTC")
19
+ def create_pipeline(**kwargs):
20
+ return pipeline([node(print, "params:message", None, name="print_message")])
21
+ ```
22
+
23
+ Register pipelines using the usual `register_pipelines()` function, optionally
24
+ calling `find_pipelines(raise_errors=True)`. Authors use conventional Kedro:
25
+
26
+ ```sh
27
+ uv run kedro run --pipelines orders
28
+ ```
29
+
30
+ This package installs **no command or Kedro hooks**. Normal `kedro run` executes
31
+ ordinary Kedro pipelines without applying schedules, selecting identities, or
32
+ injecting intervals. Kedro's catalog and parameter behavior remains unchanged.
33
+ Decorators still validate their arguments and factory return types; invalid
34
+ declarations fail early.
35
+
36
+ ## Declarations
37
+
38
+ - `@schedule(cron=..., timezone=...)` requires basic five-field cron syntax:
39
+ numeric fields with `*`, `,`, `-`, and `/`, plus an explicit IANA timezone.
40
+ For example, `0 9 * * 1-5` runs on weekdays at 09:00. Named months/weekdays
41
+ and extensions such as `L5`, `L`, `W`, `#`, `H`, `R`, and `?` are rejected
42
+ at declaration time, as are macros, seconds, and year expressions. Whitespace
43
+ is normalized. Croniter validates field values and expression structure;
44
+ its day-of-month/day-of-week convention is OR. Oleander's scheduler must
45
+ use the same interpretation.
46
+ Ranges must be ascending: use `22-23,0-2` instead of `22-2` for hours that
47
+ cross midnight. List entries must not repeat or overlap values, including
48
+ overlaps introduced by steps. Sunday may be `0` or `7`, but `0,7` is a
49
+ duplicate. A single full-week range such as `0-7` remains valid.
50
+ - `@run_as(email)` requires one email address. UUIDs, principal names, display
51
+ names, and address lists are rejected. Surrounding whitespace is stripped and
52
+ the address is lowercased and normalized to match Oleander's email convention.
53
+ Validation checks syntax locally, without DNS queries or delivery checks.
54
+ It declares the desired principal for **scheduled** execution only. It grants
55
+ no permissions and never changes the identity of a manual run. The example
56
+ application-principal email is a placeholder; it need not exist for local tests.
57
+ - Either decorator order works. Duplicate declarations are errors. Factories
58
+ preserve their signatures, names, and docstrings and return ordinary Kedro
59
+ `Pipeline` objects.
60
+ - Metadata belongs to the exact returned object. Copying, composing, filtering,
61
+ or namespacing a pipeline does not carry declarations forward. Decorate the
62
+ factory returning the final scheduled pipeline.
63
+
64
+ Oleander's private discovery/runtime validates the complete registry: pipelines
65
+ must declare both decorators or neither, and the same scheduled object cannot
66
+ appear under multiple registry names. Undecorated pipelines are discoverable
67
+ and can be run manually. Conventional Kedro does not enforce these Oleander
68
+ registry rules.
69
+
70
+ The deployment service must resolve the declared email to an existing principal
71
+ within the organization linked to the GitHub installation, authorize the
72
+ repository-to-principal binding, and store the resolved UUID with the deployment.
73
+ Missing, ambiguous, or unauthorized principals must fail deployment. Scheduled
74
+ runs use that stored UUID so email changes cannot silently change the execution
75
+ identity. This resolution and authorization belongs to Oleander's service; neither
76
+ Python package performs it. Principal membership and access must be rechecked at
77
+ execution time.
78
+
79
+ ## Metadata API
80
+
81
+ Private platform tools read metadata through the public, side-effect-free API:
82
+
83
+ ```python
84
+ from oleander_kedro import get_pipeline_metadata
85
+
86
+ metadata = get_pipeline_metadata(create_pipeline())
87
+ print(metadata.schedule.cron)
88
+ print(metadata.run_as)
89
+ ```
90
+
91
+ The returned `Declaration` and nested `Schedule` are frozen dataclasses.
92
+ `Declaration.run_as` contains the normalized email string.
93
+ Undecorated pipelines return a `Declaration` with both fields set to `None`.
94
+ `ConfigurationError` subclasses `ValueError`.
95
+
96
+ ## Development
97
+
98
+ From this directory:
99
+
100
+ ```sh
101
+ uv sync --locked
102
+ uv run pytest
103
+ uv run ruff check .
104
+ uv run ruff format --check .
105
+ uv build
106
+ ```
107
+
108
+ `.python-version` selects Python 3.10. Tests cover declaration validation and
109
+ real `kedro run` invocations with plain Python nodes and no SDK credentials.
110
+
111
+ From the repository root, also run the cross-language cron compatibility tests:
112
+
113
+ ```sh
114
+ uv sync --locked --project kedro/oleander-kedro-cli
115
+ yarn exec vitest run workflows/kedro/manifest-compatibility.test.ts
116
+ ```
117
+
118
+ These tests require the repository's Node dependencies and `uv`. They use the
119
+ same cron fixtures as the Python tests, generate manifests with the actual
120
+ Python decorators and CLI, and validate them with the TypeScript worker's
121
+ `parseManifest()`. No cloud services are involved.
122
+
123
+ The separate [private CLI project](../oleander-kedro-cli/README.md) depends on this
124
+ package and owns discovery, runtime interval injection, and the `oleander-kedro`
125
+ command.
126
+ It is not part of this distribution.
127
+
128
+ ## Publishing
129
+
130
+ The [publishing workflow](../../.github/workflows/publish-oleander-kedro.yml)
131
+ builds and publishes this extension when changes under `kedro/oleander-kedro-ext/`
132
+ reach `main`. It uses the existing GitHub Actions `PYPI_TOKEN` secret and skips
133
+ already-published artifacts, following the Python SDK's release pattern.
134
+
135
+ To release a new version, update `version` in `pyproject.toml` and refresh this
136
+ project's `uv.lock`. Update the private CLI's exact extension dependency and its
137
+ lockfile to the same version, then merge to `main`. The builder release workflow
138
+ also rebuilds its bundled packages when these directories change. The private
139
+ CLI is not published to PyPI.
@@ -0,0 +1,32 @@
1
+ [build-system]
2
+ requires = ["hatchling"]
3
+ build-backend = "hatchling.build"
4
+
5
+ [project]
6
+ name = "oleander-kedro"
7
+ version = "0.0.1"
8
+ description = "Declare Oleander metadata on ordinary Kedro pipelines"
9
+ readme = "README.md"
10
+ requires-python = ">=3.10"
11
+ dependencies = [
12
+ "kedro>=1.7,<1.8",
13
+ "croniter>=6,<7",
14
+ "email-validator>=2.3,<3",
15
+ "tzdata>=2025.2",
16
+ ]
17
+
18
+ [dependency-groups]
19
+ dev = ["pytest>=8,<9", "pytest-socket>=0.7,<1", "ruff>=0.12,<1"]
20
+
21
+ [tool.hatch.build.targets.wheel]
22
+ packages = ["src/oleander_kedro"]
23
+
24
+ [tool.pytest.ini_options]
25
+ testpaths = ["tests"]
26
+ addopts = "--disable-socket"
27
+
28
+ [tool.ruff]
29
+ target-version = "py310"
30
+
31
+ [tool.ruff.lint]
32
+ select = ["E", "F", "I", "UP", "B"]
@@ -0,0 +1,13 @@
1
+ """Oleander metadata for ordinary Kedro pipelines."""
2
+
3
+ from .decorators import Declaration, Schedule, get_pipeline_metadata, run_as, schedule
4
+ from .errors import ConfigurationError
5
+
6
+ __all__ = [
7
+ "ConfigurationError",
8
+ "Declaration",
9
+ "Schedule",
10
+ "get_pipeline_metadata",
11
+ "run_as",
12
+ "schedule",
13
+ ]
@@ -0,0 +1,157 @@
1
+ """Metadata on ordinary Kedro pipelines, without an alternate DAG model."""
2
+
3
+ import re
4
+ from collections.abc import Callable
5
+ from dataclasses import dataclass, replace
6
+ from functools import wraps
7
+ from typing import ParamSpec
8
+ from weakref import WeakKeyDictionary
9
+ from zoneinfo import ZoneInfo, ZoneInfoNotFoundError
10
+
11
+ from croniter import croniter
12
+ from email_validator import EmailNotValidError, validate_email
13
+ from kedro.pipeline import Pipeline
14
+
15
+ from .errors import ConfigurationError
16
+
17
+ P = ParamSpec("P")
18
+
19
+
20
+ @dataclass(frozen=True)
21
+ class Schedule:
22
+ cron: str
23
+ timezone: str
24
+
25
+
26
+ @dataclass(frozen=True)
27
+ class Declaration:
28
+ schedule: Schedule | None = None
29
+ run_as: str | None = None
30
+
31
+
32
+ _declarations: WeakKeyDictionary[Pipeline, Declaration] = WeakKeyDictionary()
33
+
34
+
35
+ def get_pipeline_metadata(pipeline: Pipeline) -> Declaration:
36
+ """Return immutable declarations attached to this exact pipeline object.
37
+
38
+ Undecorated and transformed pipelines have empty declarations. Reading
39
+ metadata never validates a registry, schedules work, or selects credentials.
40
+ """
41
+ # Object association avoids accidental inheritance through copy/deepcopy,
42
+ # leaves Pipeline itself untouched, and does not keep pipelines alive.
43
+ return _declarations.get(pipeline, Declaration())
44
+
45
+
46
+ def _decorate(
47
+ field: str, value: object
48
+ ) -> Callable[[Callable[P, Pipeline]], Callable[P, Pipeline]]:
49
+ def decorate(factory: Callable[P, Pipeline]) -> Callable[P, Pipeline]:
50
+ @wraps(factory)
51
+ def wrapped(*args: P.args, **kwargs: P.kwargs) -> Pipeline:
52
+ result = factory(*args, **kwargs)
53
+ if not isinstance(result, Pipeline):
54
+ raise ConfigurationError(
55
+ f"{factory.__qualname__} must return a Kedro Pipeline"
56
+ )
57
+ metadata = get_pipeline_metadata(result)
58
+ if getattr(metadata, field) is not None:
59
+ raise ConfigurationError(
60
+ f"{factory.__qualname__} declares {field} more than once"
61
+ )
62
+ _declarations[result] = replace(metadata, **{field: value})
63
+ return result
64
+
65
+ return wrapped
66
+
67
+ return decorate
68
+
69
+
70
+ def _validate_cron_fields(expression: str) -> None:
71
+ """Reject numeric forms that the worker's cron-parser cannot accept.
72
+
73
+ Check each list entry before croniter deduplicates values or wraps ranges.
74
+ """
75
+ bounds = ((0, 59), (0, 23), (1, 31), (1, 12), (0, 7))
76
+ for index, (field, (minimum, maximum)) in enumerate(
77
+ zip(expression.split(), bounds, strict=True)
78
+ ):
79
+ seen: set[int] = set()
80
+ for item in field.split(","):
81
+ if re.fullmatch(r"(?:\*|[0-9]+(?:-[0-9]+)?)(?:/[0-9]+)?", item) is None:
82
+ raise ConfigurationError("Invalid cron expression")
83
+ base, _, interval = item.partition("/")
84
+ step = int(interval) if interval else 1
85
+ if base == "*":
86
+ start, end = minimum, maximum
87
+ else:
88
+ first, _, last = base.partition("-")
89
+ start = int(first)
90
+ end = maximum if interval else start
91
+ if last:
92
+ end = int(last)
93
+ if start > end:
94
+ raise ConfigurationError(
95
+ "cron ranges must be ascending; split wrapping ranges "
96
+ "into a comma-separated list"
97
+ )
98
+ if start < minimum or end > maximum or step < 1:
99
+ raise ConfigurationError("Invalid cron expression")
100
+ # Sunday may be 0 or 7. Deduplicate within one range (e.g. 0-7),
101
+ # but reject any overlap between separate comma-separated entries.
102
+ values = {
103
+ value % 7 if index == 4 else value
104
+ for value in range(start, end + 1, step)
105
+ }
106
+ if seen.intersection(values):
107
+ raise ConfigurationError(
108
+ "cron fields must not contain duplicate values "
109
+ "or overlapping list entries"
110
+ )
111
+ seen.update(values)
112
+
113
+
114
+ def schedule(
115
+ *, cron: str, timezone: str
116
+ ) -> Callable[[Callable[P, Pipeline]], Callable[P, Pipeline]]:
117
+ """Declare basic numeric five-field cron in an explicit IANA timezone."""
118
+ if not isinstance(cron, str) or len(cron.split()) != 5:
119
+ raise ConfigurationError("cron must be a five-field expression")
120
+ expression = " ".join(cron.split())
121
+ if re.fullmatch(r"[0-9*/,\- ]+", expression) is None:
122
+ raise ConfigurationError(
123
+ "cron supports only numeric fields and the operators *, /, , and -"
124
+ )
125
+ _validate_cron_fields(expression)
126
+ if not croniter.is_valid(expression):
127
+ raise ConfigurationError("Invalid cron expression")
128
+ if not isinstance(timezone, str):
129
+ raise ConfigurationError("timezone must be an IANA timezone name")
130
+ try:
131
+ ZoneInfo(timezone)
132
+ except (ValueError, ZoneInfoNotFoundError) as exc:
133
+ raise ConfigurationError("timezone must be an IANA timezone name") from exc
134
+ return _decorate("schedule", Schedule(expression, timezone))
135
+
136
+
137
+ def run_as(
138
+ email: str,
139
+ ) -> Callable[[Callable[P, Pipeline]], Callable[P, Pipeline]]:
140
+ """Declare a principal email for scheduled runs, with offline validation.
141
+
142
+ Oleander resolves and authorizes the email within the deployment organization.
143
+ This declaration never selects credentials or changes manual-run identity.
144
+ """
145
+ if not isinstance(email, str):
146
+ raise ConfigurationError("run_as requires a valid principal email address")
147
+ try:
148
+ # Match Oleander's case-insensitive email identity convention. Application
149
+ # principals need not have mailboxes; never perform DNS or delivery checks.
150
+ normalized = validate_email(
151
+ email.strip().lower(), check_deliverability=False
152
+ ).normalized
153
+ except EmailNotValidError as exc:
154
+ raise ConfigurationError(
155
+ "run_as requires a valid principal email address"
156
+ ) from exc
157
+ return _decorate("run_as", normalized)
@@ -0,0 +1,5 @@
1
+ """Invalid Oleander pipeline declarations and invocation arguments."""
2
+
3
+
4
+ class ConfigurationError(ValueError):
5
+ """The project or invocation is invalid; no nodes should be executed."""
File without changes
@@ -0,0 +1,51 @@
1
+ {
2
+ "accepted": [
3
+ "* * * * *",
4
+ "0 * * * *",
5
+ "0 9 * * 1-5",
6
+ "*/15 * * * *",
7
+ "0,30 9-17/2 1,15 1-12 0,6",
8
+ " \t0 9\n* * 1-5 ",
9
+ "0 22-23,0-2 * * *",
10
+ "55-59,0-5 * * * *",
11
+ "0 0 28-31,1-3 * *",
12
+ "0 0 * 11-12,1-2 *",
13
+ "0 0 * * 5-6,0-1",
14
+ "0 0 * * 0",
15
+ "0 0 * * 7",
16
+ "0 0 * * 0-7",
17
+ "0 0 * * 1-7",
18
+ "0 0 * * 6-7",
19
+ "0 0 * * 0-7/2",
20
+ "0 0 * * 0-7/7",
21
+ "0-10/2,1-9/2 * * * *",
22
+ "0-10/2,3 * * * *",
23
+ "5/15 * * * *",
24
+ "*/2,1-59/2 * * * *"
25
+ ],
26
+ "duplicates": [
27
+ "0,30 9-17/2 1,15 1-12 0,7",
28
+ "0 0 * * 7,0",
29
+ "0,0 * * * *",
30
+ "0 9,9 * * *",
31
+ "0 0 1,1 * *",
32
+ "0 0 * 1,1 *",
33
+ "0 0 * * 1,1",
34
+ "0-10,5 * * * *",
35
+ "0-10,5-15 * * * *",
36
+ "*/15,30 * * * *",
37
+ "*,0 * * * *",
38
+ "0 0 * * 0,6-7",
39
+ "0 0 * * 0-6,7",
40
+ "0 0 * * *,7"
41
+ ],
42
+ "wrapping": [
43
+ "55-5 * * * *",
44
+ "0 22-2 * * *",
45
+ "0 22-2/2 * * *",
46
+ "0 0 28-3 * *",
47
+ "0 0 * 11-2 *",
48
+ "0 0 * * 5-1",
49
+ "0 0 * * 7-0"
50
+ ]
51
+ }