Trait datafusion::datasource::physical_plan::parquet::ParquetFileReaderFactory

source ·

pub trait ParquetFileReaderFactory:
    Debug
    + Send
    + Sync
    + 'static {
    // Required method
    fn create_reader(
        &self,
        partition_index: usize,
        file_meta: FileMeta,
        metadata_size_hint: Option<usize>,
        metrics: &ExecutionPlanMetricsSet,
    ) -> Result<Box<dyn AsyncFileReader + Send>>;
}

Expand description

Interface for reading parquet files.

The combined implementations of ParquetFileReaderFactory and AsyncFileReader can be used to provide custom data access operations such as pre-cached metadata, I/O coalescing, etc.

See DefaultParquetFileReaderFactory for a simple implementation.

Required Methods§

source

fn create_reader( &self, partition_index: usize, file_meta: FileMeta, metadata_size_hint: Option<usize>, metrics: &ExecutionPlanMetricsSet, ) -> Result<Box<dyn AsyncFileReader + Send>>

Provides an AsyncFileReader for reading data from a parquet file specified

§Notes

If the resulting AsyncFileReader returns ParquetMetaData without page index information, the reader will load it on demand. Thus it is important to ensure that the returned ParquetMetaData has the necessary information if you wish to avoid a subsequent I/O

§Arguments

partition_index - Index of the partition (for reporting metrics)
file_meta - The file to be read
metadata_size_hint - If specified, the first IO reads this many bytes from the footer
metrics - Execution metrics

Implementors§

source §