• Home
• airflow.providers.apache.hive
• airflow.providers.apache.hive.hooks
• airflow.providers.apache.hive.hooks.hive

airflow.providers.apache.hive.hooks.hive

Module Contents

Classes

HiveCliHook	Simple wrapper around the hive CLI.
HiveMetastoreHook	Wrapper to interact with the Hive Metastore
HiveServer2Hook	Wrapper around the pyhive library

Functions

get_context_from_env_var()

Extract context from env variable, e.g. dag_id, task_id and execution_date,

Attributes

HIVE_QUEUE_PRIORITIES

airflow.providers.apache.hive.hooks.hive.HIVE_QUEUE_PRIORITIES = ['VERY_HIGH', 'HIGH', 'NORMAL', 'LOW', 'VERY_LOW']

airflow.providers.apache.hive.hooks.hive.get_context_from_env_var()

Extract context from env variable, e.g. dag_id, task_id and execution_date, so that they can be used inside BashOperator and PythonOperator.

Returns

The context of interest.

Return type

Dict[Any, Any]

class airflow.providers.apache.hive.hooks.hive.HiveCliHook(hive_cli_conn_id=default_conn_name, run_as=None, mapred_queue=None, mapred_queue_priority=None, mapred_job_name=None)

Bases: airflow.hooks.base.BaseHook

Simple wrapper around the hive CLI.

It also supports the beeline a lighter CLI that runs JDBC and is replacing the heavier traditional CLI. To enable beeline, set the use_beeline param in the extra field of your connection as in { "use_beeline": true }

Note that you can also set default hive CLI parameters using the hive_cli_params to be used in your connection as in {"hive_cli_params": "-hiveconf mapred.job.tracker=some.jobtracker:444"} Parameters passed here can be overridden by run_cli's hive_conf param

The extra connection parameter auth gets passed as in the jdbc connection string as is.

Parameters

• hive_cli_conn_id (str) -- Reference to the Hive CLI connection id.

• mapred_queue (Optional[str]) -- queue used by the Hadoop Scheduler (Capacity or Fair)

• mapred_queue_priority (Optional[str]) -- priority within the job queue. Possible settings include: VERY_HIGH, HIGH, NORMAL, LOW, VERY_LOW

• mapred_job_name (Optional[str]) -- This name will appear in the jobtracker. This can make monitoring easier.

conn_name_attr = hive_cli_conn_id

default_conn_name = hive_cli_default

conn_type = hive_cli

hook_name = Hive Client Wrapper

run_cli(self, hql, schema=None, verbose=True, hive_conf=None)

Run an hql statement using the hive cli. If hive_conf is specified it should be a dict and the entries will be set as key/value pairs in HiveConf.

Parameters

• hql (str) -- an hql (hive query language) statement to run with hive cli

• schema (Optional[str]) -- Name of hive schema (database) to use

• verbose (bool) -- Provides additional logging. Defaults to True.

• hive_conf (Optional[Dict[Any, Any]]) -- if specified these key value pairs will be passed to hive as -hiveconf "key"="value". Note that they will be passed after the hive_cli_params and thus will override whatever values are specified in the database.

>>> hh = HiveCliHook()
>>> result = hh.run_cli("USE airflow;")
>>> ("OK" in result)
True

test_hql(self, hql)

Test an hql statement using the hive cli and EXPLAIN

load_df(self, df, table, field_dict=None, delimiter=',', encoding='utf8', pandas_kwargs=None, **kwargs)

Loads a pandas DataFrame into hive.

Hive data types will be inferred if not passed but column names will not be sanitized.

Parameters

• df (pandas.DataFrame) -- DataFrame to load into a Hive table

• table (str) -- target Hive table, use dot notation to target a specific database

• field_dict (Optional[Dict[Any, Any]]) -- mapping from column name to hive data type. Note that it must be OrderedDict so as to keep columns' order.

• delimiter (str) -- field delimiter in the file

• encoding (str) -- str encoding to use when writing DataFrame to file

• pandas_kwargs (Any) -- passed to DataFrame.to_csv

• kwargs (Any) -- passed to self.load_file

load_file(self, filepath, table, delimiter=',', field_dict=None, create=True, overwrite=True, partition=None, recreate=False, tblproperties=None)

Loads a local file into Hive

Note that the table generated in Hive uses STORED AS textfile which isn't the most efficient serialization format. If a large amount of data is loaded and/or if the tables gets queried considerably, you may want to use this operator only to stage the data into a temporary table before loading it into its final destination using a HiveOperator.

Parameters

• filepath (str) -- local filepath of the file to load

• table (str) -- target Hive table, use dot notation to target a specific database

• delimiter (str) -- field delimiter in the file

• field_dict (Optional[Dict[Any, Any]]) -- A dictionary of the fields name in the file as keys and their Hive types as values. Note that it must be OrderedDict so as to keep columns' order.

• create (bool) -- whether to create the table if it doesn't exist

• overwrite (bool) -- whether to overwrite the data in table or partition

• partition (Optional[Dict[str, Any]]) -- target partition as a dict of partition columns and values

• recreate (bool) -- whether to drop and recreate the table at every execution

• tblproperties (Optional[Dict[str, Any]]) -- TBLPROPERTIES of the hive table being created

kill(self)

Kill Hive cli command

class airflow.providers.apache.hive.hooks.hive.HiveMetastoreHook(metastore_conn_id=default_conn_name)

Bases: airflow.hooks.base.BaseHook

Wrapper to interact with the Hive Metastore

Parameters

metastore_conn_id (str) -- reference to the :ref: metastore thrift service connection id <howto/connection:hive_metastore>.

MAX_PART_COUNT = 32767

conn_name_attr = metastore_conn_id

default_conn_name = metastore_default

conn_type = hive_metastore

hook_name = Hive Metastore Thrift

__getstate__(self)

__setstate__(self, d)

get_metastore_client(self)

Returns a Hive thrift client.

get_conn(self)

Returns connection for the hook.

check_for_partition(self, schema, table, partition)

Checks whether a partition exists

Parameters

• schema (str) -- Name of hive schema (database) @table belongs to

• table (str) -- Name of hive table @partition belongs to

• partition (str) -- Expression that matches the partitions to check for (eg a = 'b' AND c = 'd')

Return type

bool

>>> hh = HiveMetastoreHook()
>>> t = 'static_babynames_partitioned'
>>> hh.check_for_partition('airflow', t, "ds='2015-01-01'")
True

check_for_named_partition(self, schema, table, partition_name)

Checks whether a partition with a given name exists

Parameters

• schema (str) -- Name of hive schema (database) @table belongs to

• table (str) -- Name of hive table @partition belongs to

• partition_name (str) -- Name of the partitions to check for (eg a=b/c=d)

Return type

bool

>>> hh = HiveMetastoreHook()
>>> t = 'static_babynames_partitioned'
>>> hh.check_for_named_partition('airflow', t, "ds=2015-01-01")
True
>>> hh.check_for_named_partition('airflow', t, "ds=xxx")
False

get_table(self, table_name, db='default')

Get a metastore table object

>>> hh = HiveMetastoreHook()
>>> t = hh.get_table(db='airflow', table_name='static_babynames')
>>> t.tableName
'static_babynames'
>>> [col.name for col in t.sd.cols]
['state', 'year', 'name', 'gender', 'num']

get_tables(self, db, pattern='*')

Get a metastore table object

get_databases(self, pattern='*')

Get a metastore table object

get_partitions(self, schema, table_name, partition_filter=None)

Returns a list of all partitions in a table. Works only for tables with less than 32767 (java short max val). For subpartitioned table, the number might easily exceed this.

>>> hh = HiveMetastoreHook()
>>> t = 'static_babynames_partitioned'
>>> parts = hh.get_partitions(schema='airflow', table_name=t)
>>> len(parts)
1
>>> parts
[{'ds': '2015-01-01'}]

max_partition(self, schema, table_name, field=None, filter_map=None)

Returns the maximum value for all partitions with given field in a table. If only one partition key exist in the table, the key will be used as field. filter_map should be a partition_key:partition_value map and will be used to filter out partitions.

Parameters

• schema (str) -- schema name.

• table_name (str) -- table name.

• field (Optional[str]) -- partition key to get max partition from.

• filter_map (Optional[Dict[Any, Any]]) -- partition_key:partition_value map used for partition filtering.

>>> hh = HiveMetastoreHook()
>>> filter_map = {'ds': '2015-01-01'}
>>> t = 'static_babynames_partitioned'
>>> hh.max_partition(schema='airflow',        ... table_name=t, field='ds', filter_map=filter_map)
'2015-01-01'

table_exists(self, table_name, db='default')

Check if table exists

>>> hh = HiveMetastoreHook()
>>> hh.table_exists(db='airflow', table_name='static_babynames')
True
>>> hh.table_exists(db='airflow', table_name='does_not_exist')
False

drop_partitions(self, table_name, part_vals, delete_data=False, db='default')

Drop partitions from the given table matching the part_vals input

Parameters

• table_name -- table name.

• part_vals -- list of partition specs.

• delete_data -- Setting to control if underlying data have to deleted in addition to dropping partitions.

• db -- Name of hive schema (database) @table belongs to

>>> hh = HiveMetastoreHook()
>>> hh.drop_partitions(db='airflow', table_name='static_babynames',
part_vals="['2020-05-01']")
True

class airflow.providers.apache.hive.hooks.hive.HiveServer2Hook(*args, schema=None, **kwargs)

Bases: airflow.hooks.dbapi.DbApiHook

Wrapper around the pyhive library

Notes: * the default authMechanism is PLAIN, to override it you can specify it in the extra of your connection in the UI * the default for run_set_variable_statements is true, if you are using impala you may need to set it to false in the extra of your connection in the UI

Parameters

• hiveserver2_conn_id -- Reference to the :ref: Hive Server2 thrift service connection id <howto/connection:hiveserver2>.

• schema (Optional[str]) -- Hive database name.

conn_name_attr = hiveserver2_conn_id

default_conn_name = hiveserver2_default

conn_type = hiveserver2

hook_name = Hive Server 2 Thrift

supports_autocommit = False

get_conn(self, schema=None)

Returns a Hive connection object.

get_results(self, hql, schema='default', fetch_size=None, hive_conf=None)

Get results of the provided hql in target schema.

Parameters

• hql (str) -- hql to be executed.

• schema (str) -- target schema, default to 'default'.

• fetch_size (Optional[int]) -- max size of result to fetch.

• hive_conf (Optional[Dict[Any, Any]]) -- hive_conf to execute alone with the hql.

Returns

results of hql execution, dict with data (list of results) and header

Return type

dict

to_csv(self, hql, csv_filepath, schema='default', delimiter=',', lineterminator='\r\n', output_header=True, fetch_size=1000, hive_conf=None)

Execute hql in target schema and write results to a csv file.

Parameters

• hql (str) -- hql to be executed.

• csv_filepath (str) -- filepath of csv to write results into.

• schema (str) -- target schema, default to 'default'.

• delimiter (str) -- delimiter of the csv file, default to ','.

• lineterminator (str) -- lineterminator of the csv file.

• output_header (bool) -- header of the csv file, default to True.

• fetch_size (int) -- number of result rows to write into the csv file, default to 1000.

• hive_conf (Optional[Dict[Any, Any]]) -- hive_conf to execute alone with the hql.

get_records(self, hql, schema='default', hive_conf=None)

Get a set of records from a Hive query.

Parameters

• hql (str) -- hql to be executed.

• schema (str) -- target schema, default to 'default'.

• hive_conf (Optional[Dict[Any, Any]]) -- hive_conf to execute alone with the hql.

Returns

result of hive execution

Return type

list

>>> hh = HiveServer2Hook()
>>> sql = "SELECT * FROM airflow.static_babynames LIMIT 100"
>>> len(hh.get_records(sql))
100

get_pandas_df(self, hql, schema='default', hive_conf=None, **kwargs)

Get a pandas dataframe from a Hive query

Parameters

• hql (str) -- hql to be executed.

• schema (str) -- target schema, default to 'default'.

• hive_conf (Optional[Dict[Any, Any]]) -- hive_conf to execute alone with the hql.

• kwargs -- (optional) passed into pandas.DataFrame constructor

Returns

result of hive execution

Return type

DataFrame

>>> hh = HiveServer2Hook()
>>> sql = "SELECT * FROM airflow.static_babynames LIMIT 100"
>>> df = hh.get_pandas_df(sql)
>>> len(df.index)
100

Returns

pandas.DateFrame

Return type

pandas.DataFrame

Previous Next

Was this entry helpful?