Complete DataAI Function Coverage for Tableau

The dataai-spark-tableau Maven artifact depends on dataai-spark-functions, so adding the Tableau artifact makes the complete DataAI computational catalog available transitively. Algorithms are not copied into Tableau code; they remain tested once in the portable functions module.

Every DataAI function result is already a Spark Dataset<Row> or a small record containing DataFrames. Pass any result DataFrame to TableauFunctionOutputs.withRunMetadata(...), persist it under a governed table/view, and connect Tableau through Spark SQL or Databricks.

Coverage

Function family	DataAI APIs available to Tableau pipelines	Suggested result name
ETL pipeline	`DataAiPipeline`, declarative `RuleSpec` validation	`pipeline_*`
Data quality	`DataQualityFunctions` automatic, missing, duplicate, date, outlier, category, and text checks	`quality_*`
General analytics	`AnalyticsFunctions` summaries, pivot, variance, ranking, correlations, regression	`analytics_*`
Time analysis	`TimeSeriesFunctions` summaries, moving averages, rolling totals, period changes	`time_*`
Business analytics	`BusinessFunctions` outliers, anomalies, drift, Pareto, cohort, funnel, KPI	`business_*`
Market models	`MarketFunctions` demand, pricing, elasticity, basket, segments, churn, risk, inventory, profit, scenario	`market_*`
Geographic	`MapFunctions.readiness(...)` and all returned readiness DataFrames	`map_*`
Matrix	`MatrixFunctions.crossTab(...)`, `MatrixFunctions.balance(...)`, `MatrixFunctions.balanceWithMinimumDeparture(...)`, and `MatrixFunctions.minimumDepartureAdjustment(...)`	`matrix_cross_tab`, `matrix_balance`
Insights	`InsightFunctions` dictionary, chart recommendations, narratives, and alerts	`insight_*`

Universal adapter

Dataset<Row> tableauResult = TableauFunctionOutputs.withRunMetadata(
        dataAiFunctionResult,
        "market_demand",
        runId,
        completedAt,
        libraryVersion);

tableauResult.write()
        .format("delta")
        .mode(SaveMode.Overwrite)
        .saveAsTable("analytics.dataai_market_demand");

The added reserved columns are:

_dataai_result_name
_dataai_run_id
_dataai_completed_at
_dataai_library_version

The adapter rejects a result that already contains one of these names instead of silently overwriting customer data.

Matrix balancing

Matrix balancing is explicitly supported. Use balanceWithMinimumDeparture(...) when the balanced matrix also requires DataAI's minimum-departure adjustment, or call minimumDepartureAdjustment(...) on an already balanced matrix. Then use TableauFunctionOutputs.matrixBalance(...) to expose balancedCells() plus:

balance_iterations
balance_maximum_error
balance_converged

The output retains the DataAI matrix fields: original value, balanced value, row and column targets, balanced totals, differences, and coefficients. The complete persistence example is ../examples/java/TableauMatrixBalancingExample.java.

Tableau can visualize the balanced values as a heat map, highlight target differences, filter to non-converged runs, and compare original versus balanced values. The customer's Spark job remains responsible for convergence policy and must decide whether a non-converged result may be published.