Superset analytic connection checks¶
Start with the shared setup and dataset. The versioned observations below use the MatrixOne build identified in the overview.
Connect with the MySQL dialect¶
Use the existing monitoring tutorial for the older deployment example. This page records Superset 6.1.0 behavior using its MySQL/PyMySQL path; it does not extend that older tutorial’s support claim to this newer combination. The separate view-reflection control used SQLAlchemy 2.1.4 and PyMySQL 1.2.3, rather than establishing those as the dependencies bundled with Superset.
Ensure the Superset environment includes a MySQL driver. In Settings > Database Connections, add a MySQL connection using the SQLAlchemy URI:
mysql+pymysql://MO_USER:PERCENT_ENCODED_PASSWORD@MO_HOST:6001/bi_tutorial?charset=utf8mb4
Set TLS through your deployment’s SQLAlchemy/driver connection arguments. Run SELECT 1 in SQL Lab before checking the shared KPI queries.
Check type compatibility before creating charts¶
Run
SELECT COUNT(*) FROM bi_tutorial.order_metrics; expect4.Run the small Decimal check separately:
SELECT CAST(1.123456 AS DECIMAL(24,6)) AS metric_value;
If it succeeds, run the category query and verify exact values. Register the physical
order_metricstable as a dataset; defineSUM(net_amount)as a metric andcategoryas a dimension. Continue to a chart only if these query and registration checks pass on your deployed versions.
In the tested combination, Decimal SQL Lab and chart requests failed with 'tuple' object does not support item assignment. MatrixOne returned legacy Decimal protocol type 0, whereas the MySQL control returned NEWDECIMAL type 246. Type 0 is not inherently invalid and the database value was correct; it activates a Superset Decimal mutation path that attempts to modify an immutable PyMySQL result tuple. This is an interoperability problem, not proof of incorrect Decimal arithmetic. See the evidence on #26928.
Physical tables and view reflection¶
Registering MatrixOne views also failed with NullType() takes no arguments. Uppercase type names in DESCRIBE were not recognized by the SQLAlchemy MySQL view reflection path; physical-table reflection passed the independent control. See #30045.
The shared tutorial deliberately uses a physical snapshot. This avoids that particular view-reflection path but does not resolve the Decimal chart failure. Lowercasing returned type names in an offline control was diagnostic only; it is not a supported application setting. Do not patch result tuples or advertise a working Decimal dashboard without retesting the complete chart path.
Validation exercised the unmodified SQL Lab, dataset registration and chart API handlers through the application test client; a complete working Superset dashboard was not established for this combination.