Delta File Relations#
The deltaFile
relation is used for creating Delta Lake tables stored only in some Hadoop
compatible file system, without registering the table in the Hive metastore. If you want to use a Delta Lake table,
and you also want to store its metadata in Hive, then use the deltaFile
relation instead.
Plugin#
This relation type is provided as part of the flowman-delta
plugin, which needs to be enabled in your
namespace.yml
file. See namespace documentation for more information for configuring plugins.
Example#
relations:
some_delta_table:
kind: deltaFile
# Specify the physical location where the data files should be stored at.
location: /warehouse/default/financial_transactions
# Optional add partition column
partitions:
- name: business_date
type: string
# Specify the default key to use in upsert operations. Normally this should match the primary key
# (except partition columns, which will be added implicitly)
mergeKey:
- id
# Specify a schema, which is mandatory for creating the table during CREATE phase
schema:
kind: inline
fields:
- name: id
type: string
- name: amount
type: double
Fields#
kind
(mandatory) (string):deltaFile
schema
(optional) (schema) (default: empty): Explicitly specifies the schema of the Delta Lake table. Alternatively Flowman will automatically use the schema of the Delta relation, if it already exists.description
(optional) (string) (default: empty): A description of the relation. This is purely for informational purpose.options
(optional) (map:string) (default: empty): All key-value pairs specified in options are directly passed to Apache Spark for reading and/or writing to this relation. Theoptions
will not be persisted in the Hive metastore. If that is what you want, then have a closer look atproperties
below.location
(mandatory) (string) (default: empty): Specifies the location of where the Delta files will be created.partitions
(optional) (list:partition) (default: empty): Specifies all partition columns. This is used both for creating Hive tables, but also for writing and reading to and from them. Therefore, if you are working with partitioned Hive tables you have to specify partition columns, even if Flowman is not used for creating the table. Normally the partition columns are separate from the schema, but you may also include the partition column in the schema, although this is not considered to be best practice. But it turns out to be quite useful in combination with dynamically writing to multiple partitions.properties
(optional) (map:string) (default: empty): Specifies additional properties of the Hive table. This setting is only used when Flowman is used to create the Delta table and is ignored otherwise.mergeKey
(optional) (list:string) (default: empty): List of column names specifying the key to identify matching records onupdate
operations.migrationPolicy
(optional) (string) (default: empty) Can be one ofRELAXED
orSTRICT
. If left empty, then the value of the Flowman configuration propertyflowman.default.relation.migrationPolicy
will be used instead.migrationStrategy
(optional) (string) (default: empty) Can be one ofALTER
,ALTER_REPLACE
,REPLACE
,NEVER
orFAIL
. If left empty, then the value of the Flowman configuration propertyflowman.default.relation.migrationStrategy
will be used instead.
Automatic Migrations#
Flowman supports some automatic migrations, specifically with the migration
strategies ALTER
, ALTER_REPLACE
and REPLACE
(those can be set via the property migrationStrategy
or the
global config variable flowman.default.relation.migrationStrategy
, see configuration
for more details).
The migration strategy ALTER
supports the following alterations:
Changing nullability
Changing the comment
Adding new columns
Other changes (like changing the data type or dropping columns) is not supported in the ALTER
strategy and
will require either REPLACE
or ALTER_REPLACE
- but this will remove all existing data in that table!
Schema Conversion#
The Delta file relation fully supports automatic schema conversion on input and output operations as described in the corresponding section of relations.
Output Modes#
Batch Writing#
The deltaFile
relation supports the following output modes in a relation
target:
Output Mode | Supported | Comments |
---|---|---|
errorIfExists |
yes | Throw an error if the Delta table already exists |
ignoreIfExists |
yes | Do nothing if the Delta table already exists |
overwrite |
yes | Overwrite the whole table or the specified partitions |
overwrite_dynamic |
yes | Overwrites all partitions where records are provided in the input data. Other partitions remain untouched |
append |
yes | Append new records to the existing table |
update |
yes | Updates existing records, either using mergeKey or the primary key of the specified schema |
In addition, the deltaFile
relation also supports complex merge operations in a merge
target.
Note that support overwrite_dynamic
is not perfect and is implemented in a two phase approach: First all partition
values are calculated, then the data is inserted. This implies that the input data is scanned twice, which might be an
expensive operation depending on the computational complexity and the amount of data.
Stream Writing#
In addition to batch writing, the Delta file relation also supports stream writing via the
stream
target with the following semantics:
Output Mode | Supported | Comments |
---|---|---|
append |
yes | Append new records from the streaming process once they don't change anymore. |
update |
yes | Append records every time they are updated |
complete |
yes | - |
Remarks#
Schema Inference#
Note that Flowman will rely on schema inference in some important situations, like mocking and generally for describing the schema of a relation. This might create unwanted connections to the physical data source, particular in case of self-contained tests. To prevent Flowman from creating a connection to the physical data source, you simply need to explicitly specify a schema, which will then be used instead of the physical schema in all situations where only schema information is required.
Writing to Dynamic Partitions#
Beside explicitly writing to a single Hive partition, Flowman also supports to write to multiple partitions where
the records need to contain values for the partition columns. Note that currently Delta table do not support the
relation output mode dynamic_overwrite
with dynamic partitions, instead you can only use append
and overwrite
.
This means, that whenever you write to a partitioned Delta table without explicitly specifying the target partition
in the relation target, then the table is truncated first, and all normally unchanged partitions
will be lost.