Standard

Apache ORC

Optimized columnar format with indexing.

1 connections

Definition

What it is

Optimized columnar format with indexing.

Recent developments

Latest signals

  • Reading and Writing the Apache ORC Format. Apache Arrow supports reading and writing ORC files. PyArrow built with ORC support bundled. Ideal in-memory representation layer for ORC data. Per arrow.apache.org.

  • Apache ORC. Official Apache ORC project site. ACID support, built-in indexes (min/max, bloom filters), complex types support. Per orc.apache.org.

  • Columnar storage formats: Parquet, ORC, and Arrow explained. May 2026 ClickHouse engineering article. ORC ecosystem narrower than Parquet; dominant in Hive/HDFS. Spark, Trino, Presto read ORC well. ClickHouse, DuckDB, BigQuery, Snowflake have varying levels of support. Per clickhouse.com (2026-05-09).

  • GitHub. Apache ORC Format 1.0.0 protobuf definitions for Apache ORC 2.0+. Standardized open-source columnar storage format spec. Per GitHub (apache/orc-format) (2023-12-05).

  • ORC Adopters. Apache ORC adopters page listing Hadoop, Spark, Arrow, Flink, Iceberg, Druid, Hive, Impala, Gobblin, Nifi, Pig, EEL, and Facebook (300+ PB). Per orc.apache.org (2015-07-16).

  • ClickHouse documents full ORC compression codec support. Self-describing columnar format with zstd, zlib, snappy, and lz4 compression. Per ORC file format | ClickHouse Resource Hub (2026-06-16).

  • ClickHouse reads ORC natively alongside Iceberg, Delta Lake, and Hudi. ClickHouse's 2026 columnar-database roundup lists raw Parquet, ORC, and Avro on S3/GCS/Azure Blob as first-class read targets, and notes ClickHouse speaks major catalogs including Hive Metastore. Per Best columnar databases in 2026 | ClickHouse (2026-06-06).

  • Iceberg treats ORC as one of only three native data file formats. Alongside Parquet and Avro — Parquet is the dominant/recommended choice, with ORC used mainly for workloads migrating off Hive. Per Iceberg Data Files | Apache Iceberg Knowledge Base (2026-05-14).

  • ORC's predicate pushdown works at three levels: file footer, stripe metadata, and row indexes. Combined with dictionary/RLE encoding and zlib/Snappy/LZO compression. Tertiary source, no better coverage found. Per A Deep-Dive in Open-Data Formats: Parquet, ORC, and Avro (IBM Community) (2026-05-05). Sources: arrow.apache.org · orc.apache.org · clickhouse.com · GitHub (apache/orc-format) · orc.apache.org

Connections 1

Outbound 1
scoped_to1
Lakehouse