Standard backend · import "parquet"

Parquet

Read and write typed columnar files for compact storage and efficient reloads. The standard hosts link the bundled backend directly.

I/O guide

Functions

Read and write Parquet

`parquet::read`

parquet::read(path: String) -> DataFrame

Reads a Parquet file into a DataFrame, preserving supported column types. In addition to local files, the bundled backend supports HTTP(S) locations and S3 URIs.

`parquet::write`

parquet::write(df: DataFrame, path: String) -> Int

Writes the DataFrame to a Parquet file and returns the number of rows written. Use this return value to record the result of a batch conversion.

Examples

Local and remote reads

import "parquet";

let local_prices = parquet::read("data/prices.parquet");
let public_prices = parquet::read("https://data.example.com/prices.parquet");
let cloud_prices = parquet::read("s3://market-data/prices.parquet?region=us-east-1");

Write a query result

Parquet works well as an output for computed tables that will be read again later.

let daily = local_prices[select { average = mean(price) }, by symbol];
let rows_written = parquet::write(daily, "out/daily.parquet");

For a directory conversion using fs::list, see the filesystem reference and the I/O guide.