SEPURI-SAI-KRISHNA opened a new issue, #45300:
URL: https://github.com/apache/superset/issues/45300
### Bug description
Importing a dashboard fails with "Import dashboard failed for an unknown
reason" when the dashboard's `json_metadata` names a chart ID that the import
cannot look up. The message gives no hint about which chart or which field is
at fault.
`update_id_refs` in `superset/commands/dashboard/importers/v1/utils.py`
remaps old chart IDs to new ones. Four metadata fields carry chart IDs, and two
of them index the map directly:
```python
if "timed_refresh_immune_slices" in metadata:
metadata["timed_refresh_immune_slices"] = [
id_map[old_id] for old_id in metadata["timed_refresh_immune_slices"]
]
...
if "expanded_slices" in metadata:
metadata["expanded_slices"] = {
str(id_map[int(old_id)]): value
for old_id, value in metadata["expanded_slices"].items()
}
```
The other two, `filter_scopes` and `default_filters`, check first, because
#17741 added that check for this exact case. The test it shipped still states
the intent:
> A chart might be removed from a dashboard but still remain in the list of
charts immune to filters. The missing chart ID should be simply ignored when
the dashboard is imported.
`timed_refresh_immune_slices` and `expanded_slices` never got the same
treatment.
**Reproduction**
Call `update_id_refs` with one chart in the layout, which is how the
existing tests for it are written:
```python
config = {
"position": {
"CHART1": {"id": "CHART1", "type": "CHART",
"meta": {"chartId": 101, "uuid": "uuid1"}},
},
"metadata": {"timed_refresh_immune_slices": [101, 999]},
}
update_id_refs(config, {"uuid1": 1}, {})
```
Results on master, where 101 is in the bundle and 999 is not:
| metadata | result |
| --- | --- |
| `timed_refresh_immune_slices: [101, 999]` | `KeyError: 999` |
| `timed_refresh_immune_slices: ["101"]` | `KeyError: '101'` |
| `expanded_slices: {"101": true, "999": false}` | `KeyError: 999` |
| `expanded_slices: {"abc": true}` | `ValueError: invalid literal for int()`
|
| `filter_scopes: {"999": {}}` | fine, entry dropped |
| `default_filters: '{"999": {}}'` | fine, entry dropped |
The second row is a reference to a chart that *is* in the bundle.
`timed_refresh_immune_slices` can hold string-typed IDs:
`superset/commands/dashboard/export.py` coerces these fields with `int()` on
the way out, and the MCP `remove_chart_from_dashboard` tool has a test named
"timed_refresh_immune_slices entries may be string-typed IDs". The import side
does no coercion, so a string entry misses the integer-keyed map and raises.
The exception is caught by the blanket `except Exception` in
`superset/commands/importers/v1/__init__.py` and re-raised as
`DashboardImportError`, which is where the "unknown reason" message comes from.
**How an unresolvable ID gets there**
Nothing prunes these two fields when a chart is deleted or leaves a
dashboard. `DeleteChartCommand` does not touch dashboard metadata, and the only
pruning code in the tree is the MCP `remove_chart_from_dashboard` tool. So
stale entries are normal. Three ways one then meets an incomplete map:
1. A chart in the layout is skipped at import when its dataset is not
resolvable in the bundle.
`superset/commands/dashboard/importers/v1/__init__.py` only imports a chart if
it has a `datasource_ref` or a `dataset_uuid` present in `dataset_info`, and
`dataset_info` only receives datasets whose database is also in the bundle.
Validation does not require either, so the chart is dropped quietly and its old
ID never enters the map.
2. `/api/v1/assets/import/` goes through the same function and lets you
import a dashboard without every chart yaml it references.
3. The exporter only started pruning these two fields in #40588 (June 2026).
Bundles produced before that keep their stale entries, and importing older
bundles is supported.
### Superset version
master / latest-dev
### Python version
3.11
### Node version
Not applicable
### Browser
Not applicable
### Additional context
Verified on master at `ffee0a843f`. The failures above come from calling
`update_id_refs` directly, so there is no server stacktrace to paste. The call
chain that reaches it is:
- `superset/commands/importers/v1/__init__.py:103` calls `self._import(...)`
inside a `try`
- `superset/commands/dashboard/importers/v1/__init__.py:277` calls
`update_id_refs(config, chart_ids, dataset_info)`
- `superset/commands/dashboard/importers/v1/utils.py:90` raises `KeyError`
from `id_map[old_id]`
- `superset/commands/importers/v1/__init__.py:107` catches it with `except
Exception` and raises `DashboardImportError`, whose message is "Import
dashboard failed for an unknown reason"
PR #37042 found the two unguarded sites in January and was closed by its own
author with no human review. It also bundled an unrelated change to
`DeleteChartCommand`. This report covers the import side only, and adds the
string-typed ID case that PR did not.
### Checklist
- [x] I have searched Superset docs and Slack and didn't find a solution to
my problem.
- [x] I have searched the GitHub issue tracker and didn't find a similar bug
report.
- [ ] I have checked Superset's logs for errors and if I found a relevant
Python stacktrace, I included it here as text in the "additional context"
section.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
