PySpark

https://spark.apache.org/docs/latest/api/python

Install

Install Ibis and dependencies for the PySpark backend:

Install with the pyspark extra:

pip install 'ibis-framework[pyspark]'

And connect:

import ibis

con = ibis.pyspark.connect()
1
Adjust connection parameters as needed.

Install for PySpark:

conda install -c conda-forge ibis-pyspark

And connect:

import ibis

con = ibis.pyspark.connect()
1
Adjust connection parameters as needed.

Install for PySpark:

mamba install -c conda-forge ibis-pyspark

And connect:

import ibis

con = ibis.pyspark.connect()
1
Adjust connection parameters as needed.

Connect

ibis.pyspark.connect

con = ibis.pyspark.connect(session=session)
Note

ibis.pyspark.connect is a thin wrapper around ibis.backends.pyspark.Backend.do_connect.

Connection Parameters

do_connect

do_connect(self, session=None, mode='batch', **kwargs)

Create a PySpark Backend for use with Ibis.

Parameters
Name Type Description Default
session SparkSession | None A SparkSession instance. None
mode ConnectionMode Can be either “batch” or “streaming”. If “batch”, every source, sink, and query executed within this connection will be interpreted as a batch workload. If “streaming”, every source, sink, and query executed within this connection will be interpreted as a streaming workload. 'batch'
kwargs Additional keyword arguments used to configure the SparkSession. {}
Examples
>>> import ibis
>>> from pyspark.sql import SparkSession
>>> session = SparkSession.builder.getOrCreate()
>>> ibis.pyspark.connect(session)
<ibis.backends.pyspark.Backend at 0x...>

ibis.connect URL format

In addition to ibis.pyspark.connect, you can also connect to PySpark by passing a properly-formatted PySpark connection URL to ibis.connect:

con = ibis.connect(f"pyspark://{warehouse-dir}?spark.app.name=CountingSheep&spark.master=local[2]")

pyspark.Backend

compile

compile(self, expr, /, *, limit=None, params=None, pretty=False)

Compile an expression to a SQL string.

Parameters

Name Type Description Default
expr ir.Expr An ibis expression to compile. required
limit str | int | None An integer to effect a specific row limit. A value of None means no limit. None
params Mapping[ir.Expr, Any] | None Mapping of scalar parameter expressions to value. None
pretty bool Pretty print the SQL query during compilation. False

Returns

Name Type Description
str Compiled expression

compute_stats

compute_stats(self, name, database=None, noscan=False)

Issue a COMPUTE STATISTICS command for a given table.

Parameters

Name Type Description Default
name str Table name required
database str | None Database name None
noscan bool If True, collect only basic statistics for the table (number of rows, size in bytes). False

connect

connect(self, *args, **kwargs)

Connect to the database.

Parameters

Name Type Description Default
*args Mandatory connection parameters, see the docstring of do_connect for details. ()
**kwargs Extra connection parameters, see the docstring of do_connect for details. {}

Notes

This creates a new backend instance with saved args and kwargs, then calls reconnect and finally returns the newly created and connected backend instance.

Returns

Name Type Description
BaseBackend An instance of the backend

create_database

create_database(self, name, /, *, catalog=None, path=None, force=False)

Create a new Spark database.

Parameters

Name Type Description Default
name str Database name required
catalog str | None Catalog to create database in (defaults to current_catalog) None
path str | Path | None Path where to store the database data; otherwise uses Spark default None
force bool Whether to append IF NOT EXISTS to the database creation SQL False

create_table

create_table(self, name, /, obj=None, *, schema=None, database=None, temp=None, overwrite=False, format='parquet', partition_by=None)

Create a new table in Spark.

Parameters

Name Type Description Default
name str Name of the new table. required
obj ir.Table | pd.DataFrame | pa.Table | pl.DataFrame | pl.LazyFrame | None If passed, creates table from SELECT statement results None
schema sch.SchemaLike | None Mutually exclusive with obj, creates an empty table with a schema None
database str | None Database name To specify a table in a separate catalog, you can pass in the catalog and database as a string "catalog.database", or as a tuple of strings ("catalog", "database"). None
temp bool | None Whether the new table is temporary (unsupported) None
overwrite bool If True, overwrite existing data False
format str Format of the table on disk 'parquet'
partition_by str | list[str] | None Name(s) of partitioning column(s) None

Returns

Name Type Description
Table The newly created table.

Examples

>>> con.create_table("new_table_name", table_expr)  # quartodoc: +SKIP

create_view

create_view(self, name, /, obj, *, database=None, overwrite=False)

Create a temporary Spark view from a table expression.

Parameters

Name Type Description Default
name str View name required
obj ir.Table Expression to use for the view required
database str | None Database name None
overwrite bool Replace an existing view of the same name if it exists False

Returns

Name Type Description
Table The created view

disconnect

disconnect(self)

Disconnect from the backend.

drop_database

drop_database(self, name, /, *, catalog=None, force=False)

Drop a Spark database.

Parameters

Name Type Description Default
name str Database name required
catalog str | None Catalog containing database to drop (defaults to current_catalog) None
force bool If False, Spark throws exception if database is not empty or database does not exist False

drop_table

drop_table(self, name, /, *, database=None, force=False)

Drop a table from the backend.

Parameters

Name Type Description Default
name str The name of the table to drop required
database tuple[str, str] | str | None The database that the table is located in. None
force bool If True, do not raise an error if the table does not exist. False

drop_view

drop_view(self, name, /, *, database=None, force=False)

Drop a view from the backend.

Parameters

Name Type Description Default
name str The name of the view to drop. required
database str | None The database that the view is located in. None
force bool If True, do not raise an error if the view does not exist. False

execute

execute(self, expr, /, *, params=None, limit=None, **kwargs)

Execute an expression.

from_connection

from_connection(cls, session, /, *, mode='batch', **kwargs)

Create a PySpark Backend from an existing SparkSession instance.

Parameters

Name Type Description Default
session SparkSession A SparkSession instance. required
mode ConnectionMode Can be either “batch” or “streaming”. If “batch”, every source, sink, and query executed within this connection will be interpreted as a batch workload. If “streaming”, every source, sink, and query executed within this connection will be interpreted as a streaming workload. 'batch'
kwargs Any Additional keyword arguments used to configure the SparkSession. {}

get_schema

get_schema(self, table_name, *, catalog=None, database=None)

Return a Schema object for the indicated table and database.

Parameters

Name Type Description Default
table_name str Table name. May be fully qualified required
catalog str | None Catalog to use None
database str | None Database to use to get the active database. None

Returns

Name Type Description
Schema An ibis schema

has_operation

has_operation(cls, operation, /)

Return whether the backend supports the given operation.

Parameters

Name Type Description Default
operation type[ops.Value] Operation type, a Python class object. required

insert

insert(self, name, /, obj, *, database=None, overwrite=False)

Insert data into a table.

NoteIbis does not use the word schema to refer to database hierarchy.

A collection of table is referred to as a database. A collection of database is referred to as a catalog.

These terms are mapped onto the corresponding features in each backend (where available), regardless of whether the backend itself uses the same terminology.

Parameters

Name Type Description Default
name str The name of the table to which data needs will be inserted required
obj pd.DataFrame | ir.Table | list | dict The source data or expression to insert required
database str | None Name of the attached database that the table is located in. For backends that support multi-level table hierarchies, you can pass in a dotted string path like "catalog.database" or a tuple of strings like ("catalog", "database"). None
overwrite bool If True then replace existing contents of table False

list_catalogs

list_catalogs(self, *, like=None)

List existing catalogs in the current connection.

NoteIbis does not use the word schema to refer to database hierarchy.

A collection of table is referred to as a database. A collection of database is referred to as a catalog.

These terms are mapped onto the corresponding features in each backend (where available), regardless of the terminology the backend uses.

See the Table Hierarchy Concepts Guide for more info.

Parameters

Name Type Description Default
like str | None A pattern in Python’s regex format to filter returned catalog names. None

Returns

Name Type Description
list[str] The catalog names that exist in the current connection, that match the like pattern if provided.

list_databases

list_databases(self, *, like=None, catalog=None)

List existing databases in the current connection.

NoteIbis does not use the word schema to refer to database hierarchy.

A collection of table is referred to as a database. A collection of database is referred to as a catalog.

These terms are mapped onto the corresponding features in each backend (where available), regardless of the terminology the backend uses.

See the Table Hierarchy Concepts Guide for more info.

Parameters

Name Type Description Default
like str | None A pattern in Python’s regex format to filter returned database names. None
catalog str | None The catalog to list databases from. If None, the current catalog is searched. None

Returns

Name Type Description
list[str] The database names that exist in the current connection, that match the like pattern if provided.

list_tables

list_tables(self, *, like=None, database=None)

The table names that match like in the given database.

For some backends, the tables may be files in a directory, or other equivalent entities in a SQL database.

Parameters

Name Type Description Default
like str | None A pattern in Python’s regex format. None
database tuple[str, str] | str | None The database, or (catalog, database) from which to list tables. For backends that support a single-level table hierarchy, you can pass in a string like "bar". For backends that support multi-level table hierarchies, you can pass in a dotted string path like "catalog.database" or a tuple of strings like ("catalog", "database"). If not provided, the current database (and catalog, if applicable for this backend) is used. See the Table Hierarchy Concepts Guide for more info. None

Returns

Name Type Description
list[str] The list of the table names that match the pattern like.

Examples

This example uses the DuckDB backend, but the list_tables API works the same for other backends.

>>> import ibis
>>> con = ibis.duckdb.connect()
>>> foo = con.create_table("foo", schema=ibis.schema(dict(a="int")))
>>> con.list_tables()
['foo']
>>> bar = con.create_view("bar", foo)
>>> con.list_tables()
['bar', 'foo']
>>> con.create_database("my_database")
>>> con.list_tables(database="my_database")
[]
>>> con.raw_sql("CREATE TABLE my_database.baz (a INTEGER)")
<duckdb.duckdb.DuckDBPyConnection object at 0x...>
>>> con.list_tables(database="my_database")
['baz']

raw_sql

raw_sql(self, query, **kwargs)

read_csv

read_csv(self, paths, /, *, table_name=None, **kwargs)

Register a CSV file as a table in the current database.

Parameters

Name Type Description Default
paths str | list[str] | tuple[str] The data source(s). May be a path to a file or directory of CSV files, or an iterable of CSV files. required
table_name str | None An optional name to use for the created table. This defaults to a random generated name. None
kwargs Any Additional keyword arguments passed to PySpark loading function. https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.DataFrameReader.csv.html {}

Returns

Name Type Description
ir.Table The just-registered table

read_csv_dir

read_csv_dir(self, path, /, *, table_name=None, watermark=None, **kwargs)

Register a CSV directory as a table in the current database.

Parameters

Name Type Description Default
path str | Path The data source. required
table_name str | None An optional name to use for the created table. This defaults to a random generated name. None
watermark Watermark | None Watermark strategy for the table. None
kwargs Any Additional keyword arguments passed to PySpark loading function. https://spark.apache.org/docs/latest/api/python/reference/pyspark.ss/api/pyspark.sql.streaming.DataStreamReader.csv.html {}

Returns

Name Type Description
ir.Table The just-registered table

read_delta

read_delta(self, path, /, *, table_name=None, **kwargs)

Register a Delta Lake table as a table in the current database.

Parameters

Name Type Description Default
path str | Path The path to the Delta Lake table. required
table_name str | None An optional name to use for the created table. This defaults to a random generated name. None
kwargs Any Additional keyword arguments passed to PySpark. https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.DataFrameReader.load.html {}

Returns

Name Type Description
ir.Table The just-registered table

read_json

read_json(self, paths, /, *, table_name=None, **kwargs)

Register a JSON file as a table in the current database.

Parameters

Name Type Description Default
paths str | Sequence[str] The data source(s). May be a path to a file or directory of JSON files, or an iterable of JSON files. required
table_name str | None An optional name to use for the created table. This defaults to a random generated name. None
kwargs Any Additional keyword arguments passed to PySpark loading function. https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.DataFrameReader.json.html {}

Returns

Name Type Description
ir.Table The just-registered table

read_json_dir

read_json_dir(self, path, /, *, table_name=None, watermark=None, **kwargs)

Register a JSON file as a table in the current database.

Parameters

Name Type Description Default
path str | Path The data source. A directory of JSON files. required
table_name str | None An optional name to use for the created table. This defaults to a random generated name. None
watermark Watermark | None Watermark strategy for the table. None
kwargs Any Additional keyword arguments passed to PySpark loading function. https://spark.apache.org/docs/latest/api/python/reference/pyspark.ss/api/pyspark.sql.streaming.DataStreamReader.json.html {}

Returns

Name Type Description
ir.Table The just-registered table

read_kafka

read_kafka(self, *, table_name=None, watermark=None, auto_parse=False, schema=None, options=None)

Register a Kafka topic as a table.

Parameters

Name Type Description Default
table_name str | None An optional name to use for the created table. This defaults to a sequentially generated name. None
watermark Watermark | None Watermark strategy for the table. None
auto_parse bool Whether to parse Kafka messages automatically. If False, the source is read as binary keys and values. If True, the key is discarded and the value is parsed using the provided schema. False
schema sch.Schema | None Schema of the value of the Kafka messages. None
options Mapping[str, str] | None Additional arguments passed to PySpark as .option(“key”, “value”). https://spark.apache.org/docs/latest/structured-streaming-kafka-integration.html None

Returns

Name Type Description
ir.Table The just-registered table

read_parquet

read_parquet(self, path, /, *, table_name=None, **kwargs)

Register a parquet file as a table in the current database.

Parameters

Name Type Description Default
path str | Path The data source. May be a path to a file or directory of parquet files. required
table_name str | None An optional name to use for the created table. This defaults to a random generated name. None
kwargs Any Additional keyword arguments passed to PySpark. https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.DataFrameReader.parquet.html {}

Returns

Name Type Description
ir.Table The just-registered table

read_parquet_dir

read_parquet_dir(self, path, /, *, table_name=None, watermark=None, schema=None, **kwargs)

Register a parquet file as a table in the current database.

Parameters

Name Type Description Default
path str | Path The data source. A directory of parquet files. required
table_name str | None An optional name to use for the created table. This defaults to a random generated name. None
watermark Watermark | None Watermark strategy for the table. None
schema sch.Schema | None Schema of the parquet source. None
kwargs Any Additional keyword arguments passed to PySpark. https://spark.apache.org/docs/latest/api/python/reference/pyspark.ss/api/pyspark.sql.streaming.DataStreamReader.parquet.html {}

Returns

Name Type Description
ir.Table The just-registered table

reconnect

reconnect(self)

Reconnect to the database already configured with connect.

register_options

register_options(cls)

Register custom backend options.

rename_table

rename_table(self, old_name, new_name)

Rename an existing table.

Parameters

Name Type Description Default
old_name str The old name of the table. required
new_name str The new name of the table. required

sql

sql(self, query, /, *, schema=None, dialect=None)

Create an Ibis table expression from a SQL query.

Parameters

Name Type Description Default
query str A SQL query string required
schema SchemaLike | None The schema of the query. If not provided, Ibis will try to infer the schema of the query. None
dialect str | None The SQL dialect of the query. If not provided, the backend’s dialect is assumed. This argument can be useful when the query is written in a different dialect from the backend. None

Returns

Name Type Description
ir.Table The table expression representing the query

table

table(self, name, /, *, database=None)

to_csv

to_csv(self, expr, /, path, *, params=None, **kwargs)

Write the results of executing the given expression to a CSV file.

This method is eager and will execute the associated expression immediately.

Parameters

Name Type Description Default
expr ir.Table The ibis expression to execute and persist to CSV. required
path str | Path The data source. A string or Path to the CSV file. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
kwargs Any Additional keyword arguments passed to pyarrow.csv.CSVWriter {}
https required

to_csv_dir

to_csv_dir(self, expr, /, path, *, params=None, limit=None, options=None)

Write the results of executing the given expression to a CSV directory.

Parameters

Name Type Description Default
expr ir.Expr The ibis expression to execute and persist to CSV. required
path str | Path The data source. A string or Path to the CSV directory. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means “no limit”. The default is in ibis/config.py. None
options Mapping[str, str] | None Additional keyword arguments passed to pyspark.sql.streaming.DataStreamWriter None

Returns

Name Type Description
StreamingQuery | None Returns a Pyspark StreamingQuery object if in streaming mode, otherwise None

to_delta

to_delta(self, expr, /, path, *, params=None, limit=None, **kwargs)

Write the results of executing the given expression to a Delta Lake table.

This method is eager and will execute the associated expression immediately.

Parameters

Name Type Description Default
expr ir.Table The ibis expression to execute and persist to a Delta Lake table. required
path str | Path The data source. A string or Path to the Delta Lake table. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means “no limit”. The default is in ibis/config.py. None
**kwargs Any Additional keyword arguments passed to pyspark.sql.DataFrameWriter. {}

to_json

to_json(self, expr, /, path, **kwargs)

Write the results of expr to a json file of [{column -> value}, …] objects.

This method is eager and will execute the associated expression immediately.

Parameters

Name Type Description Default
expr ir.Table The ibis expression to execute and persist to Delta Lake table. required
path str | Path The data source. A string or Path to the Delta Lake table. required
kwargs Any Additional, backend-specifc keyword arguments. {}

to_kafka

to_kafka(self, expr, /, *, auto_format=False, options=None, params=None, limit='default')

Write the results of executing the given expression to a Kafka topic.

This method does not return outputs. Streaming queries are run continuously in the background.

Parameters

Name Type Description Default
expr ir.Expr The ibis expression to execute and persist to a Kafka topic. required
auto_format bool Whether to format the Kafka messages before writing. If False, the output is written as-is. If True, the output is converted into JSON and written as the value of the Kafka messages. False
options Mapping[str, str] | None PySpark Kafka write arguments. https://spark.apache.org/docs/latest/structured-streaming-kafka-integration.html None
params Mapping | None Mapping of scalar parameter expressions to value. None
limit str | None An integer to effect a specific row limit. A value of None means “no limit”. The default is in ibis/config.py. 'default'

Returns

Name Type Description
StreamingQuery A Pyspark StreamingQuery object

to_pandas

to_pandas(self, expr, /, *, params=None, limit=None, **kwargs)

Execute an Ibis expression and return a pandas DataFrame, Series, or scalar.

Note

This method is a wrapper around execute.

Parameters

Name Type Description Default
expr ir.Expr Ibis expression to execute. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means no limit. The default is in ibis/config.py. None
kwargs Any Keyword arguments {}

to_pandas_batches

to_pandas_batches(self, expr, /, *, params=None, limit=None, chunk_size=1000000, **kwargs)

Execute an Ibis expression and return an iterator of pandas DataFrames.

Parameters

Name Type Description Default
expr ir.Expr Ibis expression to execute. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means no limit. The default is in ibis/config.py. None
chunk_size int Maximum number of rows in each returned DataFrame batch. This may have no effect depending on the backend. 1000000
kwargs Any Keyword arguments {}

Returns

Name Type Description
Iterator[pd.DataFrame] An iterator of pandas DataFrames.

to_parquet

to_parquet(self, expr, /, path, *, params=None, limit=None, **kwargs)

Write the results of executing the given expression to a Parquet file.

This method is eager and will execute the associated expression immediately.

Parameters

Name Type Description Default
expr ir.Table The ibis expression to execute and persist to a Parquet file. required
path str | Path The data source. A string or Path to the Parquet file. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means “no limit”. The default is in ibis/config.py. None
**kwargs Any Additional keyword arguments passed to pyspark.sql.DataFrameWriter. {}

to_parquet_dir

to_parquet_dir(self, expr, /, path, *, params=None, limit=None, options=None)

Write the results of executing the given expression to a parquet directory.

Parameters

Name Type Description Default
expr ir.Expr The ibis expression to execute and persist to parquet. required
path str | Path The data source. A string or Path to the parquet directory. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means “no limit”. The default is in ibis/config.py. None
options Mapping[str, str] | None Additional keyword arguments passed to pyspark.sql.streaming.DataStreamWriter None

Returns

Name Type Description
StreamingQuery | None Returns a Pyspark StreamingQuery object if in streaming mode, otherwise None

to_polars

to_polars(self, expr, /, *, params=None, limit=None, **kwargs)

Execute expression and return results in as a polars DataFrame.

This method is eager and will execute the associated expression immediately.

Parameters

Name Type Description Default
expr ir.Expr Ibis expression to export to polars. required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means no limit. The default is in ibis/config.py. None
kwargs Any Keyword arguments {}

Returns

Name Type Description
dataframe A polars DataFrame holding the results of the executed expression.

to_pyarrow

to_pyarrow(self, expr, /, *, params=None, limit=None, **kwargs)

Execute expression to a pyarrow object.

This method is eager and will execute the associated expression immediately.

Parameters

Name Type Description Default
expr ir.Expr Ibis expression to export to pyarrow required
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
limit int | str | None An integer to effect a specific row limit. A value of None means no limit. The default is in ibis/config.py. None
kwargs Any Keyword arguments {}

Returns

Name Type Description
result If the passed expression is a Table, a pyarrow table is returned. If the passed expression is a Column, a pyarrow array is returned. If the passed expression is a Scalar, a pyarrow scalar is returned.

to_pyarrow_batches

to_pyarrow_batches(self, expr, /, *, params=None, limit=None, chunk_size=1000000, **kwargs)

Execute expression and return an iterator of PyArrow record batches.

This method is eager and will execute the associated expression immediately.

Parameters

Name Type Description Default
expr ir.Expr Ibis expression to export to pyarrow required
limit int | str | None An integer to effect a specific row limit. A value of None means “no limit”. The default is in ibis/config.py. None
params Mapping[ir.Scalar, Any] | None Mapping of scalar parameter expressions to value. None
chunk_size int Maximum number of rows in each returned record batch. 1000000

Returns

Name Type Description
RecordBatchReader Collection of pyarrow RecordBatchs.

to_torch

to_torch(self, expr, /, *, params=None, limit=None, **kwargs)

Execute an expression and return results as a dictionary of torch tensors.

Parameters

Name Type Description Default
expr ir.Expr Ibis expression to execute. required
params Mapping[ir.Scalar, Any] | None Parameters to substitute into the expression. None
limit int | str | None An integer to effect a specific row limit. A value of None means no limit. None
kwargs Any Keyword arguments passed into the backend’s to_torch implementation. {}

Returns

Name Type Description
dict[str, torch.Tensor] A dictionary of torch tensors, keyed by column name.

truncate_table

truncate_table(self, name, /, *, database=None)

Delete all rows from a table.

NoteIbis does not use the word schema to refer to database hierarchy.

A collection of tables is referred to as a database. A collection of database is referred to as a catalog. These terms are mapped onto the corresponding features in each backend (where available), regardless of whether the backend itself uses the same terminology.

Parameters

Name Type Description Default
name str Table name required
database str | None Name of the attached database that the table is located in. For backends that support multi-level table hierarchies, you can pass in a dotted string path like "catalog.database" or a tuple of strings like ("catalog", "database"). None
Back to top