Parquet functions

This page introduces the Apache Parquet read function.

info

Apache Parquet support is in beta.

It may not be fit for production use.

Please let us know if you run into issues.

Either:

  1. Email us at support@questdb.io
  2. Join our public Slack
  3. Post on our Discourse community

read_parquet​

Reads a parquet file as a table.

read_parquet(parquet_file_path)

Usage​

With this function, query a Parquet file located at the QuestDB copy root directory. Both relative and absolute file paths are supported.

read_parquet example
SELECT
  *
FROM
  read_parquet('trades.parquet')
WHERE
  side = 'buy'
LIMIT 1;
symbolsidepriceamounttimestamp
BTC-USDbuy62755.60.000433672024-07-01T00:46:39.754075Z

The query above:

  • Reads all columns from the file trades.parquet located at the server copy root directory, i.e. import/trades.parquet in the QuestDB copy root directory by default.
  • Filters rows, keeping only the first row where the side column equals buy.

Configuration​

For security reason, reading is only allowed from a specified directory. It defaults to the import directory inside the QuestDB copy root directory. To change the allowed directory, set the cairo.sql.copy.root configuration by using one of the following settings:

  • The environment variable QDB_CAIRO_SQL_COPY_ROOT.
  • The cairo.sql.copy.root key in server.conf.

Limitations​

Parquet format support rich set of data types, including structural types. QuestDB only can read data types that match QuestDB data types:

  • Varchar
  • Int
  • Long
  • Short
  • Byte
  • Boolean
  • UUID
  • Double
  • Float
  • Timestamp
  • Binary

Parquet columns with unsupported data types are ignored.

Multiple files are not suppored, only a single file.

Nested data and/or arrays are not supported.