Refactor SQL parsing components to Use CoralTable Instead of Hive Table

[aastha25]

This refactoring impacts multiple components that are currently coupled to Hive's Table class:

(1)ParseTreeBuilder (coral-hive) is tightly coupled to Hive's org.apache.hadoop.hive.metastore.api.Table class, preventing it from working with other table formats (e.g., Iceberg). This needs to be refactored to use the unified CoralTable interface to enable multi-format support.

(2) HiveFunctionResolver (coral-hive) - Resolves Hive and Dali function names to Calcite operators

• Method: tryResolveAsDaliFunction(String functionName, @Nonnull Table table, int numOfOperands)
• Uses: Calls HiveCalciteTableAdapter.getDaliFunctionParams() and getDaliUdfDependencies() for UDF resolution

(3) HiveCalciteTableAdapter (coral-common) - Calcite adapter for Hive tables

• Provides: getDaliFunctionParams() - parses "functions" table property (format: "f:c1 g:c2")
• Provides: getDaliUdfDependencies() - parses "dependencies" table property (format: "o1:m1:v1 o2:m2:v2?transitive=false")
• Used by: HiveFunctionResolver for Dali UDF resolution
• (previously called HiveTable)

(4) HiveCalciteViewAdapter (coral-common) - Calcite adapter for Hive views (extends HiveCalciteTableAdapter)

• Inherits: Dali UDF metadata extraction methods
• Used by: View expansion in Calcite's relational algebra conversion
• (previously called HiveView)

(5) IcebergCalciteTableAdapter (coral-common) - Calcite adapter for Iceberg tables

• Current workaround: Uses IcebergHiveTableConverter to convert Iceberg tables to Hive Table objects for UDF resolution
• Blocks: Direct Iceberg table support in ParseTreeBuilder
• (Introduced as a new class in PR Add native Apache Iceberg table support with CoralCatalog abstraction #556 since (3) & (4) exist)

(6) IcebergHiveTableConverter (coral-common) - Temporary bridge (to be removed)

• Converts: IcebergCoralTable → Hive Table for backward compatibility
• Why exists: ParseTreeBuilder and HiveFunctionResolver can't accept CoralTable yet
• Will be removed: Once this issue is resolved
• • (Introduced as a new class in PR Add native Apache Iceberg table support with CoralCatalog abstraction #556 since (1) through (5) exist)

Refactoring Scope:

• Replace org.apache.hadoop.hive.metastore.api.Table parameters with CoralTable in all affected APIs
• Move Dali UDF metadata extraction logic from HiveCalciteTableAdapter to work with CoralTable.properties()
• Update HiveFunctionResolver to accept CoralTable instead of Hive Table
• Remove IcebergHiveTableConverter once ParseTreeBuilder accepts CoralTable
• Ensure both HiveCoralTable and IcebergCoralTable can provide UDF metadata through the unified interface

This will enable Coral to parse and translate SQL from both Hive and Iceberg tables using a unified API.

Background:
Coral is introducing CoralTable as a unified abstraction for representing tables across different formats (Hive, Iceberg, etc.) as part of #556. However, ParseTreeBuilder - a core component that converts view SQL -> Hive AST (no schema dependency, only view metadatata - definition & properties), - is still deeply integrated with Hive's metastore Table class throughout its API surface and internal implementation.

Minimal Example:

Before (Current - Hive-specific):

// Client code
import org.apache.hadoop.hive.metastore.api.Table;

Table hiveTable = hiveMetastore.getTable("mydb", "mytable");
ParseTreeBuilder builder = new ParseTreeBuilder(functionResolver);
SqlNode sqlNode = builder.process("SELECT * FROM mytable", hiveTable);

After (Desired - Format-agnostic):

// Client code - works with both Hive and Iceberg
import com.linkedin.coral.common.catalog.CoralTable;
import com.linkedin.coral.common.catalog.HiveCoralTable;

// For Hive tables
org.apache.hadoop.hive.metastore.api.Table hiveTable = hiveMetastore.getTable("mydb", "mytable");
CoralTable coralTable = new HiveCoralTable(hiveTable);

// OR for Iceberg tables (when available)
// CoralTable coralTable = new IcebergCoralTable(icebergTable);

ParseTreeBuilder builder = new ParseTreeBuilder(functionResolver);
SqlNode sqlNode = builder.process("SELECT * FROM mytable", coralTable);

[aastha25]

Additionally, Coral uses Calcite Schema adapters to bridge catalog metadata to Calcite's query planner. These current exist with legacy integration -

1. HiveSchema (coral-common) - Legacy root-level Calcite adapter

   • Uses: HiveMetastoreClient directly for database/table lookups
   • Returns: HiveDbSchema instances for each database
   • Status: Legacy API for backward compatibility

2. HiveDbSchema (coral-common) - Legacy database-level Calcite adapter

   • Uses: HiveMetastoreClient to fetch Hive tables
   • Returns: HiveCalciteTableAdapter or HiveCalciteViewAdapter based on table type
   • Status: Legacy API for backward compatibility

These will also be cleaned up as part of the same issue