Skip to content

Add support for upstream/downstream model selectors with graph operators #2056

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
devin-ai-integration wants to merge 4 commits into
base: master
Choose a base branch
from
Open

Add support for upstream/downstream model selectors with graph operators #2056

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
d21f10f
Add support for upstream/downstream model selectors with graph operators
devin-ai-integration[bot] Nov 19, 2025
5a86050
Fix CI issues: remove unused import, fix selector consistency, handle…
devin-ai-integration[bot] Nov 19, 2025
be8d198
Merge branch 'master' into devin/1763544153-upstream-downstream-selec…
arbiv Jan 6, 2026
cf4dd43
Merge branch 'master' into devin/1763544153-upstream-downstream-selec…
arbiv Jan 27, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 9 additions & 5 deletions elementary/monitor/cli.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -144,10 +144,12 @@ def decorator(func):
"--select",
type=str,
default=None,
help="Filter the report by last_invocation / invocation_id:<INVOCATION_ID> / invocation_time:<INVOCATION_TIME>."
if cmd in (Command.REPORT, Command.SEND_REPORT)
else "DEPRECATED! Please use --filters instead! - Filter the alerts by tags:<TAGS> / owners:<OWNERS> / models:<MODELS> / "
"statuses:<warn/fail/error/skipped> / resource_types:<model/test>.",
help=(
"Filter the report by last_invocation / invocation_id:<INVOCATION_ID> / invocation_time:<INVOCATION_TIME>."
if cmd in (Command.REPORT, Command.SEND_REPORT)
else "DEPRECATED! Please use --filters instead! - Filter the alerts by tags:<TAGS> / owners:<OWNERS> / models:<MODELS> / "
"statuses:<warn/fail/error/skipped> / resource_types:<model/test>."
),
)(func)
return func

Expand Down Expand Up @@ -364,7 +366,9 @@ def monitor(

alert_filters = FiltersSchema()
if bool(filters) or bool(excludes):
alert_filters = FiltersSchema.from_cli_params(filters, excludes)
alert_filters = FiltersSchema.from_cli_params(
filters, excludes, config, anonymous_tracking
)
elif select is not None:
click.secho(
'\n"--select" is deprecated and won\'t be supported in the near future.\n'
Expand Down
26 changes: 24 additions & 2 deletions elementary/monitor/data_monitoring/schema.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -190,8 +190,13 @@ def validate_report_selector(self) -> None:

@staticmethod
def from_cli_params(
cli_filters: Tuple[str], cli_excludes: Tuple[str]
cli_filters: Tuple[str],
cli_excludes: Tuple[str],
config: Optional[Any] = None,
tracking: Optional[Any] = None,
) -> "FiltersSchema":
from elementary.monitor.data_monitoring.selector_filter import SelectorFilter

all_filters: list[tuple[str, FilterType]] = []
for cli_filter in cli_filters:
all_filters.append((cli_filter, FilterType.IS))
Expand All @@ -206,6 +211,7 @@ def from_cli_params(
models = []
statuses = []
resource_types = []
node_names = []

for cli_filter, filter_type in all_filters:
tags_match = FiltersSchema._match_filter_regex(
Expand All @@ -226,7 +232,22 @@ def from_cli_params(
filter_string=cli_filter, regex=re.compile(r"models:(.*)")
)
if models_match:
models.append(FilterSchema(values=models_match, type=filter_type))
model_value = (
models_match[0]
if len(models_match) == 1
else ",".join(models_match)
)
if (
config
and filter_type == FilterType.IS
and SelectorFilter._has_graph_operators(model_value)
):
selector_filter = SelectorFilter(config, tracking, model_value)
filter_result = selector_filter.get_filter()
if filter_result.node_names:
node_names.extend(filter_result.node_names)
Comment on lines +247 to +248
Copy link
Copy Markdown
Contributor Author

@devin-ai-integration devin-ai-integration Bot Mar 10, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔴 Graph operator model filter silently dropped when dbt ls returns no matching nodes

In from_cli_params, when a models: filter with graph operators (e.g., models:customers+) is resolved via SelectorFilter and dbt ls returns an empty list, the filter is silently dropped. At elementary/monitor/data_monitoring/schema.py:247, if filter_result.node_names: is falsy for an empty list, so nothing is added to either node_names or models. The continue on line 251 then skips to the next filter. This means a user who specifies --filters models:customers+ where no nodes match the graph selector will see all alerts instead of no alerts, because the filter was effectively removed rather than applied as an empty match set.

Prompt for agents 
In elementary/monitor/data_monitoring/schema.py, in the from_cli_params method around lines 245-248, after getting filter_result from SelectorFilter, the code currently only extends node_names if filter_result.node_names is truthy. When dbt ls returns no matching nodes (empty list), this silently drops the filter entirely, causing all alerts to pass through instead of none.

The fix should ensure that even when the resolved node_names list is empty, the filter is still applied. One approach: always extend node_names from filter_result.node_names (remove the if guard). When node_names is populated but empty of real values, the FilterSchema with IS type will correctly filter out everything. Alternatively, if node_names remains empty and no other node_names are set, you could add a sentinel value like a UUID that will never match any real node name.

Change lines 247-248 from:
    if filter_result.node_names:
        node_names.extend(filter_result.node_names)
To:
    node_names.extend(filter_result.node_names)

But also verify that downstream logic in FiltersSchema.apply() handles the case where node_names is an empty list correctly (it should filter out everything). Currently at line 353-358, `if self.node_names` is falsy for an empty list, so it returns True (passes filter). This means you also need to track a flag indicating that graph resolution was attempted, and ensure an empty resolved list results in filtering out all alerts. The simplest fix may be to add a sentinel impossible-to-match node name when resolution returns empty.
Open in Devin Review

Was this helpful? React with 👍 or 👎 to provide feedback.

else:
models.append(FilterSchema(values=models_match, type=filter_type))
continue

statuses_match = FiltersSchema._match_filter_regex(
Expand Down Expand Up @@ -269,6 +290,7 @@ def from_cli_params(
models=models,
statuses=statuses,
resource_types=resource_types,
node_names=node_names,
)

@staticmethod
Expand Down
29 changes: 23 additions & 6 deletions elementary/monitor/data_monitoring/selector_filter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,12 +99,25 @@ def _parse_selector(self, selector: Optional[str] = None) -> FiltersSchema:
selector=selector,
)
elif model_match:
if self.tracking:
self.tracking.set_env("select_method", "model")
data_monitoring_filter = FiltersSchema(
models=[FilterSchema(values=[model_match.group(1)])],
selector=selector,
)
model_value = model_match.group(1)
if self.selector_fetcher and self._has_graph_operators(model_value):
if self.tracking:
self.tracking.set_env(
"select_method", "model with graph operators"
)
node_names = self.selector_fetcher.get_selector_results(
selector=model_value
)
data_monitoring_filter = FiltersSchema(
node_names=node_names, selector=selector
)
else:
if self.tracking:
self.tracking.set_env("select_method", "model")
data_monitoring_filter = FiltersSchema(
models=[FilterSchema(values=[model_value])],
selector=selector,
)
elif statuses_match:
if self.tracking:
self.tracking.set_env("select_method", "statuses")
Expand Down Expand Up @@ -148,6 +161,10 @@ def _create_user_dbt_runner(self, config: Config) -> Optional[BaseDbtRunner]:
def get_filter(self) -> FiltersSchema:
return self.filter

@staticmethod
def _has_graph_operators(selector: str) -> bool:
return "+" in selector

@staticmethod
def _can_use_fetcher(selector):
non_dbt_selectors = [
Expand Down
103 changes: 103 additions & 0 deletions tests/unit/monitor/data_monitoring/test_filters_schema.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,10 @@
from unittest.mock import patch

import pytest

from elementary.monitor.data_monitoring.schema import FiltersSchema, FilterType
from tests.mocks.anonymous_tracking_mock import MockAnonymousTracking
from tests.mocks.config_mock import MockConfig


def test_empty_from_cli_params():
Expand Down Expand Up @@ -310,3 +314,102 @@ def test_exclude_statuses_filters():
["fail", "error", "runtime error", "warn"]
)
assert filter_schema.statuses[1].type == FilterType.IS


def test_models_with_graph_operators_from_cli_params():
with patch(
"elementary.clients.dbt.command_line_dbt_runner.CommandLineDbtRunner.ls"
) as mock_ls:
mock_ls.return_value = ["model.customers", "model.orders", "model.payments"]

config = MockConfig("mock_project_dir")
tracking = MockAnonymousTracking()

cli_filter = ("models:customers+",)
cli_excludes = ()
filter_schema = FiltersSchema.from_cli_params(
cli_filter, cli_excludes, config, tracking
)
assert len(filter_schema.tags) == 0
assert len(filter_schema.models) == 0
assert len(filter_schema.owners) == 0
assert len(filter_schema.node_names) == 3
assert sorted(filter_schema.node_names) == sorted(
["model.customers", "model.orders", "model.payments"]
)
assert len(filter_schema.statuses) == 1
assert sorted(filter_schema.statuses[0].values) == sorted(
["fail", "error", "runtime error", "warn"]
)
assert len(filter_schema.resource_types) == 0


def test_models_with_upstream_graph_operators_from_cli_params():
with patch(
"elementary.clients.dbt.command_line_dbt_runner.CommandLineDbtRunner.ls"
) as mock_ls:
mock_ls.return_value = [
"model.raw_customers",
"model.stg_customers",
"model.customers",
]

config = MockConfig("mock_project_dir")
tracking = MockAnonymousTracking()

cli_filter = ("models:+customers",)
cli_excludes = ()
filter_schema = FiltersSchema.from_cli_params(
cli_filter, cli_excludes, config, tracking
)
assert len(filter_schema.tags) == 0
assert len(filter_schema.models) == 0
assert len(filter_schema.owners) == 0
assert len(filter_schema.node_names) == 3
assert sorted(filter_schema.node_names) == sorted(
["model.raw_customers", "model.stg_customers", "model.customers"]
)
assert len(filter_schema.statuses) == 1
assert sorted(filter_schema.statuses[0].values) == sorted(
["fail", "error", "runtime error", "warn"]
)
assert len(filter_schema.resource_types) == 0


def test_models_without_graph_operators_from_cli_params_no_config():
cli_filter = ("models:customers+",)
cli_excludes = ()
filter_schema = FiltersSchema.from_cli_params(cli_filter, cli_excludes)
assert len(filter_schema.tags) == 0
assert len(filter_schema.models) == 1
assert filter_schema.models[0].values == ["customers+"]
assert len(filter_schema.owners) == 0
assert len(filter_schema.node_names) == 0
assert len(filter_schema.statuses) == 1
assert sorted(filter_schema.statuses[0].values) == sorted(
["fail", "error", "runtime error", "warn"]
)
assert len(filter_schema.resource_types) == 0


def test_exclude_models_with_graph_operators_from_cli_params():
"""Test that graph operators in excludes are NOT resolved to node_names"""
config = MockConfig("mock_project_dir")
tracking = MockAnonymousTracking()

cli_filter = ()
cli_excludes = ("models:customers+",)
filter_schema = FiltersSchema.from_cli_params(
cli_filter, cli_excludes, config, tracking
)
assert len(filter_schema.tags) == 0
assert len(filter_schema.models) == 1
assert filter_schema.models[0].values == ["customers+"]
assert filter_schema.models[0].type == FilterType.IS_NOT
assert len(filter_schema.owners) == 0
assert len(filter_schema.node_names) == 0
assert len(filter_schema.statuses) == 1
assert sorted(filter_schema.statuses[0].values) == sorted(
["fail", "error", "runtime error", "warn"]
)
assert len(filter_schema.resource_types) == 0
62 changes: 62 additions & 0 deletions tests/unit/monitor/data_monitoring/test_selector_filter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -265,3 +265,65 @@ def dbt_runner_with_models_mock() -> Generator[MagicMock, None, None]:
) as mock_ls:
mock_ls.return_value = ["node_name_1", "node_name_2"]
yield mock_ls


def test_parse_selector_with_graph_operators_downstream(
dbt_runner_with_models_mock, anonymous_tracking_mock
):
config = MockConfig("mock_project_dir")

data_monitoring_filter = SelectorFilter(
tracking=anonymous_tracking_mock,
config=config,
selector="model:customers+",
)

assert data_monitoring_filter.get_filter().node_names == [
"node_name_1",
"node_name_2",
]
assert data_monitoring_filter.get_filter().selector == "model:customers+"


def test_parse_selector_with_graph_operators_upstream(
dbt_runner_with_models_mock, anonymous_tracking_mock
):
config = MockConfig("mock_project_dir")

data_monitoring_filter = SelectorFilter(
tracking=anonymous_tracking_mock,
config=config,
selector="model:+customers",
)

assert data_monitoring_filter.get_filter().node_names == [
"node_name_1",
"node_name_2",
]
assert data_monitoring_filter.get_filter().selector == "model:+customers"


def test_parse_selector_with_graph_operators_both(
dbt_runner_with_models_mock, anonymous_tracking_mock
):
config = MockConfig("mock_project_dir")

data_monitoring_filter = SelectorFilter(
tracking=anonymous_tracking_mock,
config=config,
selector="model:+customers+",
)

assert data_monitoring_filter.get_filter().node_names == [
"node_name_1",
"node_name_2",
]
assert data_monitoring_filter.get_filter().selector == "model:+customers+"


Comment on lines +270 to +323
Copy link
Copy Markdown

@coderabbitai coderabbitai Bot Nov 19, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

⚠️ Potential issue | 🟡 Minor

Mark dbt_runner_with_models_mock as used to satisfy Ruff (ARG001)

In the three graph-operator tests, dbt_runner_with_models_mock is required for fixture injection/patching but never referenced in the body, so Ruff flags it as an unused argument.

One simple pattern is to reference the argument at the top of each test:

-def test_parse_selector_with_graph_operators_downstream(
-    dbt_runner_with_models_mock, anonymous_tracking_mock
-):
+def test_parse_selector_with_graph_operators_downstream(
+    dbt_runner_with_models_mock, anonymous_tracking_mock
+):
+    _ = dbt_runner_with_models_mock  # ensure fixture is used for linters

@@
-def test_parse_selector_with_graph_operators_upstream(
-    dbt_runner_with_models_mock, anonymous_tracking_mock
-):
+def test_parse_selector_with_graph_operators_upstream(
+    dbt_runner_with_models_mock, anonymous_tracking_mock
+):
+    _ = dbt_runner_with_models_mock  # ensure fixture is used for linters

@@
-def test_parse_selector_with_graph_operators_both(
-    dbt_runner_with_models_mock, anonymous_tracking_mock
-):
+def test_parse_selector_with_graph_operators_both(
+    dbt_runner_with_models_mock, anonymous_tracking_mock
+):
+    _ = dbt_runner_with_models_mock  # ensure fixture is used for linters

Alternatively, you could rename the fixture parameter to _dbt_runner_with_models_mock and adjust the fixture definition name accordingly if you prefer that style.

🧰 Tools
🪛 Ruff (0.14.5)

271-271: Unused function argument: dbt_runner_with_models_mock

(ARG001)

289-289: Unused function argument: dbt_runner_with_models_mock

(ARG001)

307-307: Unused function argument: dbt_runner_with_models_mock

(ARG001)

🤖 Prompt for AI Agents 
In tests/unit/monitor/data_monitoring/test_selector_filter.py around lines 270
to 323, the fixture parameter dbt_runner_with_models_mock is injected but not
referenced which triggers Ruff ARG001; fix by either referencing it at the top
of each test (e.g., assign it to a throwaway like _ =
dbt_runner_with_models_mock) or rename the parameter to
_dbt_runner_with_models_mock (and if renaming, update the fixture definition
name accordingly) so Ruff no longer treats it as an unused argument.

def test_has_graph_operators():
assert SelectorFilter._has_graph_operators("customers+") is True
assert SelectorFilter._has_graph_operators("+customers") is True
assert SelectorFilter._has_graph_operators("+customers+") is True
assert SelectorFilter._has_graph_operators("customers") is False
assert SelectorFilter._has_graph_operators("my_model") is False
Loading