uk.ac.starlink.parquet

Interface Encoder<T>

public interface Encoder<T>

Defines transmission of a java value type to a RecordConsumer for writing to parquet tables.

Since:

25 Feb 2021

Author:

Mark Taylor

Method Summary

Modifier and Type	Method and Description
void	addValue(T value, org.apache.parquet.io.api.RecordConsumer consumer) Passes a supplied non-null value representing a cell of this encoder's column to a record consumer.
String	getColumnName() Returns the name of the column that will be written to the parquet file.
org.apache.parquet.schema.Type	getColumnType() Returns the type of the top-level field that will be written.
T	typedValue(Object obj) Converts an object presented for record consumption to an object of this encoder's parameterised class.

Method Detail

getColumnName

String getColumnName()

Returns the name of the column that will be written to the parquet file. This is the top-level field, which may correspond to a group type containing lower-level elements, which will be assigned subordinate names automatically.

Note this name must obey the (undocumented?) parquet field syntax constraints; it seems that roughly it has to look like a java token.

Returns:

field name

getColumnType

org.apache.parquet.schema.Type getColumnType()

Returns the type of the top-level field that will be written. This may be a primitive type or a group type.

Returns:

column type

typedValue

T typedValue(Object obj)

Converts an object presented for record consumption to an object of this encoder's parameterised class. The presented object ought to be of the right class anyway, so this may just be a cast. However, if the object is of an unexpected type, or if its value corresponds to a blank value, null should be returned.

Parameters:

obj - object for encoding, expected to be of this encoder's type

Returns:

typed object for record consumption, or null if it does not correspond to a definite typed value

addValue

void addValue(T value,
              org.apache.parquet.io.api.RecordConsumer consumer)

Passes a supplied non-null value representing a cell of this encoder's column to a record consumer. This is invoked between RecordConsumer startField and endField calls, but if additional structuring calls are required they must be done by this method.

Parameters:

value - non-null typed value to write

consumer - value consumer