feat: Support reading Parquet ENUM logical type as String

[leochen4891]

Closes #7723

Background

Parquet's ENUM logical type is physically identical to STRING: both annotate a BINARY column with UTF-8 encoded bytes. The only difference is the label. External tools such as Spark and PyArrow use ENUM to indicate a column holds a finite set of string values, but the wire format is the same.

Deephaven's read pipeline has three stages where logical type is dispatched. All three previously had no handling for EnumLogicalTypeAnnotation, causing ENUM-annotated columns from externally produced files to fail on read.

Changes

Stage 1 — Schema to Java type (ParquetSchemaReader)

Before: visit(EnumLogicalTypeAnnotation) set an error string and returned Optional.empty(), so the column was unresolvable.

After: Returns Optional.of(String.class), the same result as visit(StringLogicalTypeAnnotation).

Stage 2 — Column data to chunk (ParquetColumnLocation)

Before: No visit(EnumLogicalTypeAnnotation) override existed, so the visitor returned Optional.empty() and the read failed at runtime.

After: A new override delegates to ToStringPage.create(...), the same decoder used for STRING columns.

Stage 3 — Pushdown statistics (MinMaxFromStatistics)

Before: getMinMaxForStrings only accepted StringLogicalTypeAnnotation, so ENUM columns returned false and forced a full scan on every filter.

After: The condition adds || instanceof EnumLogicalTypeAnnotation, enabling min/max pushdown for ENUM columns.

Tests

• MinMaxFromStatisticsTest.enumLogicalStatisticsAreMaterialised — unit test verifying ENUM statistics are extracted as strings.
• ParquetTableReadWriteTest.testReadEnumLogicalTypeAsString — end-to-end test that writes a Parquet file with a BINARY+ENUM column and reads it back, verifying the column materializes as String with correct values.

[github-actions]

No docs changes detected for 84ea207

[Copilot]

Pull request overview

Adds support for Parquet ENUM logical type by treating it like STRING across Deephaven’s Parquet read pipeline, enabling successful reads and statistics pushdown for externally-produced Parquet files.

Changes:

• Map EnumLogicalTypeAnnotation to String during schema-to-Java type resolution.
• Decode ENUM-annotated BINARY columns using the existing string page decoder.
• Enable min/max statistics extraction (pushdown) for ENUM logical type and add unit + end-to-end tests.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
extensions/parquet/table/src/main/java/io/deephaven/parquet/table/ParquetSchemaReader.java	Maps Parquet ENUM logical type to String in schema interpretation.
extensions/parquet/table/src/main/java/io/deephaven/parquet/table/location/ParquetColumnLocation.java	Adds ENUM logical type visitor to decode values via ToStringPage.
extensions/parquet/table/src/main/java/io/deephaven/parquet/table/location/MinMaxFromStatistics.java	Allows string min/max extraction for ENUM logical type to support pushdown.
extensions/parquet/table/src/test/java/io/deephaven/parquet/table/location/MinMaxFromStatisticsTest.java	Adds unit test asserting min/max materialization for ENUM-annotated stats.
extensions/parquet/table/src/test/java/io/deephaven/parquet/table/ParquetTableReadWriteTest.java	Adds end-to-end test writing a BINARY+ENUM column and reading it back as String.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.