Byron94odhiambo opened a new issue, #61594:
URL: https://github.com/apache/airflow/issues/61594
### Description
When triggering a DAG via the Airflow web UI, there is no way to define
**dependent (cascading) relationships** between `Param` dropdowns — where the
selected value in one dropdown dynamically filters the available options in
another.
For example, if a DAG has a `country` param and a `region` param, selecting
**"Kenya"** should only show Kenyan regions (`Nandi County`, `Nyandarua`,
`Nakuru`), and selecting **"Ethiopia"** should only show Ethiopian regions
(`Jimma`, `Arsi`, etc.). Currently, all regions from all countries are shown in
a **flat list** regardless of which country is selected.
---
### Current Workaround
The only option today is to flatten all child values into a single `enum`
list and validate the parent-child pairing at runtime:
```python
from airflow.models.param import Param
from airflow.models import Variable
COUNTRY_REGIONS = Variable.get("country_region_mapping",
deserialize_json=True)
# {
# "Kenya": {"Nandi County": "21802", "Nyandarua": "26727", "Nakuru":
"26199"},
# "Ethiopia": {"Jimma": "123", "Arsi": "456"}
# }
COUNTRY_LIST = list(COUNTRY_REGIONS.keys())
# Flatten ALL regions into one list — no filtering by country
ALL_REGIONS = ["All Regions"]
for country in COUNTRY_REGIONS:
ALL_REGIONS.extend(list(COUNTRY_REGIONS[country].keys()))
dag_params = {
"country": Param(default="Kenya", enum=COUNTRY_LIST, type="string"),
"region": Param(
default="All Regions",
enum=ALL_REGIONS, # Shows ALL regions from ALL countries
description=" Ensure region matches the selected country",
),
}
```
> **Problems:**
> - Users can select mismatched pairs (e.g. `Kenya` + `Jimma`) — the DAG
fails at runtime instead of preventing the mistake in the UI
> - The dropdown is cluttered with irrelevant options from other parent
values
> - Does not scale as more countries/regions are added
> - Requires a dedicated validation task in every DAG that uses hierarchical
params
---
### Proposed Solution — `depends_on` Attribute
Add a `depends_on` keyword to `Param` that references a parent param and
maps each parent value to its valid child options:
```python
dag_params = {
"country": Param(
default="Kenya",
enum=["Kenya", "Ethiopia"],
type="string",
),
"region": Param(
default="Nandi County",
type="string",
depends_on={
"param": "country",
"mapping": {
"Kenya": ["All Regions", "Nandi County", "Nyandarua",
"Nakuru"],
"Ethiopia": ["All Regions", "Jimma", "Arsi", "Shashemene"],
},
},
),
}
```
**How it would work in the Trigger Form UI:**
1. `country` renders as a normal dropdown
2. When `country` changes → the UI reads
`region.depends_on.mapping[selectedCountry]` and updates `region`'s options
3. If the current `region` value is not in the new options → it resets to
the first available option
4. Server-side validation on submit still validates the final values
---
### Implementation Notes
**Backend (Python):**
- Extend `airflow.models.param.Param` to accept a `depends_on` attribute
- Serialize the dependency metadata into the JSON Schema returned by the
trigger form API endpoint
**Frontend (React — Trigger Form):**
- When rendering params, check if a param has `depends_on`
- Register an `onChange` listener on the parent param's dropdown
- On parent value change → update child dropdown options, reset child value
if invalid
### Use case/motivation
This is a common requirement in **multi-country / multi-tenant data
pipelines**.
Our team runs a project, which operates automated data quality pipelines
across Kenya and Ethiopia. Each country has multiple sub-regions, and our DAGs
need users to select a valid country → region pair when triggering reports.
Without cascading support, we must flatten all regions into a single dropdown
and add runtime validation — leading to avoidable DAG failures and a poor user
experience.
This pattern appears frequently across many Airflow deployments:
| Parent Param | Child Param | Example |
|:---|:---|:---|
| Country | Region / Province | Multi-country data pipelines |
| Cloud Provider | Region / Zone | Infrastructure DAGs (`AWS` → `us-east-1`,
`GCP` → `us-central1`) |
| Database | Schema / Table | ETL pipelines |
| Environment | Cluster | Deployment DAGs (`prod` → `prod-cluster-1`) |
| Department | Team | Org-scoped reporting DAGs |
| Data Source | Dataset | Analytics pipelines |
### Related issues
- #56632 — Dynamic DAG parameter options from external config (addresses
dynamic enum values, but **not** cascading/dependent relationships between
params)
- #56427 — DAG Params Enum not working as expected
- #42524 — DAG Param incorrectly converts enum objects to strings
- #31399 — Trigger UI Form Dropdowns with enums do not set default correctly
- #39904 — Dynamic DAG Params behave differently in manually triggered vs
scheduled runs
- https://github.com/apache/airflow/discussions/48481 — Multiple-choice
Params discussion
### Are you willing to submit a PR?
- [x] Yes I am willing to submit a PR!
### Code of Conduct
- [x] I agree to follow this project's [Code of
Conduct](https://github.com/apache/airflow/blob/main/CODE_OF_CONDUCT.md)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
