About Parquet to JSON

Parse and convert Parquet file metadata to JSON. Part of the DevTools Surf developer suite. Browse more tools in the Data / SQL collection.

Use Cases

• Inspect Parquet file schema, row count, and column statistics without a Spark cluster.
• Convert Parquet file contents to JSON for debugging or data inspection.
• Check compression codec and encoding metadata before deciding on storage optimization.
• Validate a Parquet file's schema against an expected schema before loading into a data warehouse.

Tips

• Inspect the Parquet schema before converting — column types (INT64, BINARY, DOUBLE) map differently to JSON number vs. string types depending on encoding.
• Check the row group size metadata — large row groups (128MB+) are optimal for analytics queries but may indicate the file was created for Spark rather than streaming reads.
• Use the metadata viewer to inspect compression codec (Snappy, GZIP, LZ4, ZSTD) before deciding whether to re-compress for your use case.

Fun Facts

• Apache Parquet was created by Twitter and Cloudera in 2013, donated to Apache in the same year. It was designed to be the columnar storage format for the Hadoop ecosystem, addressing the row-based limitations of CSV and JSON.
• Parquet's columnar storage compresses 5-7x better than row-based formats like CSV for typical analytics data because similar values are stored adjacent — integers in a column compress far better than interspersed mixed-type rows.
• AWS, Google, and Azure all use Parquet as the default format for their data lake services (S3 Athena, BigQuery, Azure Data Lake Storage) — making Parquet the de facto standard for cloud analytics storage as of 2020.

FAQ

Does it handle nested/repeated Parquet columns?

Yes — Dremel encoding (nested repeated fields) is decoded and represented as nested JSON objects/arrays, matching how Spark and BigQuery surface nested Parquet columns.

Is there a file size limit?

Files up to 100MB can be processed in-browser. Larger files should be processed with a command-line tool like parquet-tools or pyarrow — the browser memory footprint of large Parquet files is significant.

Does it support all Parquet logical types?

Standard logical types are supported: STRING, INT (8/16/32/64), FLOAT, DOUBLE, BOOLEAN, DATE, TIME, TIMESTAMP, DECIMAL, UUID, and JSON. Custom logical types may be shown as raw bytes.