Chart Configuration Serialization#

Chart configuration in Superset is stored in the Slice model (slices table) across two MediumText columns:

• params — the full chart form configuration as a JSON string 1. Exposed at runtime as the form_data property, which deserializes the JSON and merges in live fields like datasource, slice_id, and viz_type.
• query_context — a serialized QueryContext blob used by the query pipeline 2.

The set of fields included in export/import is defined by two class-level lists on Slice 3:

ImportExportMixin.import_from_dict() gates field acceptance to export_fields + extra_import_fields, preventing injection of arbitrary columns.

Export: ExportChartsCommand#

ExportChartsCommand extends ExportModelsCommand and produces a ZIP containing metadata.yaml, per-chart YAML files under charts/, and optionally related dataset and tag YAMLs.

The core serialization happens in _file_content():

1. export_to_dict() extracts export_fields with recursive=False, include_defaults=True, export_uuids=True.
2. Key removal — REMOVE_KEYS = ["datasource_type", "datasource_name", "url_params"] are stripped 4. datasource_type/datasource_name are redundant once dataset_uuid is present; url_params are environment-local.
3. params JSON expansion — the params string is decoded from JSON into a Python dict before YAML serialization, producing human-readable YAML rather than an escaped string 5.
4. Extra fields appended: version ("1.0.0"), dataset_uuid (the linked SqlaTable.uuid), and tags when the TAGGING_SYSTEM feature flag is enabled 6.

Related datasets are exported transitively via ExportDatasetsCommand.

Import: ImportChartsCommand Pipeline#

ImportChartsCommand._import() orchestrates a multi-step pipeline that resolves UUID-based references to local database IDs before writing any rows:

1. Collect dataset_uuids from chart configs
2. Collect database_uuids from dataset configs
3. import_database() for each referenced database
4. import_dataset() with resolved database_id
5. update_chart_config_dataset() — rewrite datasource refs
6. import_chart() — write the Slice row
7. import_tag() if TAGGING_SYSTEM enabled

Datasource Reference Rewriting#

Before import_chart() is called, update_chart_config_dataset() rewrites all datasource references using the newly imported dataset's local ID. It patches three levels simultaneously 7:

If query_context JSON cannot be parsed, it is set to None 8.

import_chart() Steps#

import_chart() in superset/commands/chart/importers/v1/utils.py:

1. Looks up the chart by uuid; checks write permissions and ownership before overwriting 9.
2. Calls filter_chart_annotations() — mutates params to keep only FORMULA-type annotation layers; EVENT and other types that depend on external objects are dropped.
3. Re-serializes params from dict back to JSON string 10.
4. Calls migrate_chart() to upgrade legacy viz types via the processors registry, which also rewrites query_context.form_data to match the migrated params.
5. Calls Slice.import_from_dict() to upsert the row 11.

filter_box charts are silently skipped by the outer import loop before import_chart() is reached 12.

Key Source Files#