Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Minor: Add more documentation about table_partition_columns #5576

Merged
merged 5 commits into from
Mar 15, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
49 changes: 42 additions & 7 deletions datafusion/core/src/datasource/listing/table.rs
Original file line number Diff line number Diff line change
Expand Up @@ -212,10 +212,7 @@ pub struct ListingOptions {
/// The file format
pub format: Arc<dyn FileFormat>,
/// The expected partition column names in the folder structure.
/// For example `Vec["a", "b"]` means that the two first levels of
/// partitioning expected should be named "a" and "b":
/// - If there is a third level of partitioning it will be ignored.
/// - Files that don't follow this partitioning will be ignored.
/// See [Self::with_table_partition_cols] for details
pub table_partition_cols: Vec<(String, DataType)>,
/// Set true to try to guess statistics from the files.
/// This can add a lot of overhead as it will usually require files
Expand Down Expand Up @@ -297,16 +294,54 @@ impl ListingOptions {
self
}

/// Set table partition column names on [`ListingOptions`] and returns self.
/// Set `table partition columns` on [`ListingOptions`] and returns self.
///
/// You may use [`wrap_partition_type_in_dict`] to request a dictionary-encoded type.
/// "partition columns," used to support [Hive Partitioning], are
/// columns added to the data that is read, based on the folder
/// structure where the data resides.
///
/// For example, give the following files in your filesystem:
///
/// ```text
/// /mnt/nyctaxi/year=2022/month=01/tripdata.parquet
/// /mnt/nyctaxi/year=2021/month=12/tripdata.parquet
/// /mnt/nyctaxi/year=2021/month=11/tripdata.parquet
/// ```
///
/// A [`ListingTable`] created at `/mnt/nyctaxi/` with partition
/// columns "year" and "month" will include new `year` and `month`
/// columns while reading the files. The `year` column would have
/// value `2022` and the `month` column would have value `01` for
/// the rows read from
/// `/mnt/nyctaxi/year=2022/month=01/tripdata.parquet`
///
///# Notes
///
/// - If only one level (e.g. `year` in the example above) is
/// specified, the other levels are ignored but the files are
/// still read.
///
/// - Files that don't follow this partitioning scheme will be
/// ignored.
///
/// - Since the columns have the same value for all rows read from
/// each individual file (such as dates), they are typically
/// dictionary encoded for efficiency. You may use
/// [`wrap_partition_type_in_dict`] to request a
/// dictionary-encoded type.
///
/// - The partition columns are solely extracted from the file path. Especially they are NOT part of the parquet files itself.
///
/// # Example
///
/// ```
/// # use std::sync::Arc;
/// # use arrow::datatypes::DataType;
/// # use datafusion::prelude::col;
/// # use datafusion::datasource::{listing::ListingOptions, file_format::parquet::ParquetFormat};
///
/// // listing options for files with paths such as `/mnt/data/col_a=x/col_b=y/data.parquet`
/// // `col_a` and `col_b` will be included in the data read from those files
/// let listing_options = ListingOptions::new(Arc::new(
/// ParquetFormat::default()
/// ))
Expand All @@ -317,7 +352,7 @@ impl ListingOptions {
/// ("col_b".to_string(), DataType::Utf8)]);
/// ```
///
///
/// [Hive Partitioning]: https://docs.cloudera.com/HDPDocuments/HDP2/HDP-2.1.3/bk_system-admin-guide/content/hive_partitioned_tables.html
/// [`wrap_partition_type_in_dict`]: crate::physical_plan::file_format::wrap_partition_type_in_dict
pub fn with_table_partition_cols(
mut self,
Expand Down
21 changes: 11 additions & 10 deletions datafusion/core/src/physical_plan/file_format/mod.rs
Original file line number Diff line number Diff line change
Expand Up @@ -69,22 +69,23 @@ use std::{

use super::{ColumnStatistics, Statistics};

/// Convert logical type of partition column to physical type: `Dictionary(UInt16, val_type)`.
/// Convert type to a type suitable for use as a [`ListingTable`]
/// partition column. Returns `Dictionary(UInt16, val_type)`, which is
/// a reasonable trade off between a reasonable number of partition
/// values and space efficiency.
///
/// You CAN use this to specify types for partition columns. However you MAY also choose not to dictionary-encode the
/// data or to use a different dictionary type.
/// This use this to specify types for partition columns. However
/// you MAY also choose not to dictionary-encode the data or to use a
/// different dictionary type.
///
/// Use [`wrap_partition_value_in_dict`] to wrap the values.
/// Use [`wrap_partition_value_in_dict`] to wrap a [`ScalarValue`] in the same say.
pub fn wrap_partition_type_in_dict(val_type: DataType) -> DataType {
DataType::Dictionary(Box::new(DataType::UInt16), Box::new(val_type))
}

/// Convert scalar value of partition columns to physical type: `Dictionary(UInt16, val_type)` .
///
/// You CAN use this to specify types for partition columns. However you MAY also choose not to dictionary-encode the
/// data or to use a different dictionary type.
///
/// Use [`wrap_partition_type_in_dict`] to wrap the types.
/// Convert a [`ScalarValue`] of partition columns to a type, as
/// decribed in the documentation of [`wrap_partition_type_in_dict`],
/// which can wrap the types.
pub fn wrap_partition_value_in_dict(val: ScalarValue) -> ScalarValue {
ScalarValue::Dictionary(Box::new(DataType::UInt16), Box::new(val))
}
Expand Down