atoti.session module

class atoti.session.Session(name, *, config, detached_process)

Holds a connection to the Java gateway.

close()

Close this session and free all the associated resources.

Return type

None

property closed: bool

Return whether the session is closed or not.

Return type

bool

create_cube(base_table, name=None, *, mode='auto')

Create a cube based on the passed table.

Parameters
  • base_table (Table) – The base table of the cube.

  • name (Optional[str]) – The name of the created cube. Defaults to the name of the base table.

  • mode (Literal[‘auto’, ‘manual’, ‘no_measures’]) –

    The cube creation mode:

    • auto: Creates hierarchies for every key column or non-numeric column of the table, and measures for every numeric column.

    • manual: Does not create any hierarchy or measure (except from the count).

    • no_measures: Creates the hierarchies like auto but does not create any measures.

    For tables with hierarchized_columns specified, these will be converted into hierarchies regardless of the cube creation mode.

Example

>>> table = session.create_table(
...     "Table",
...     types={"id": tt.type.STRING, "value": tt.type.NULLABLE_DOUBLE},
... )
>>> cube_auto = session.create_cube(table)
>>> sorted(cube_auto.measures)
['contributors.COUNT', 'update.TIMESTAMP', 'value.MEAN', 'value.SUM']
>>> list(cube_auto.hierarchies)
[('Table', 'id')]
>>> cube_no_measures = session.create_cube(table, mode="no_measures")
>>> sorted(cube_no_measures.measures)
['contributors.COUNT', 'update.TIMESTAMP']
>>> list(cube_no_measures.hierarchies)
[('Table', 'id')]
>>> cube_manual = session.create_cube(table, mode="manual")
>>> sorted(cube_manual.measures)
['contributors.COUNT', 'update.TIMESTAMP']
>>> list(cube_manual.hierarchies)
[]

See also

Hierarchies and measures created by a join().

Return type

Cube

create_scenario(name, *, origin='Base')

Create a new source scenario.

Parameters
  • name (str) – The name of the scenario.

  • origin (str) – The scenario to fork.

Return type

None

create_table(name, *, types, keys=(), partitioning=None, hierarchized_columns=None, **kwargs)

Create a table from a schema.

Parameters
  • name (str) – The name of the table to create.

  • types (Mapping[str, DataType]) – Types for all columns of the table. This defines the columns which will be expected in any future data loaded into the table.

  • keys (Iterable[str]) – The columns that will become keys of the table.

  • partitioning (Optional[str]) –

    The description of how the data will be split across partitions of the table.

    Joined tables can only use a sub-partitioning of the table referencing them.

    Example

    hash4(country) splits the data across 4 partitions based on the country column’s hash value.

  • hierarchized_columns (Optional[Iterable[str]]) –

    The list of columns which will automatically be converted into hierarchies no matter which creation mode is used for the cube.

    The different behaviors based on the passed value are:

    • None: all non-numeric columns are converted into hierarchies, depending on the cube’s creation mode.

    • Empty collection: no columns are converted into hierarchies.

    • Non-empty collection: only the columns in the collection will be converted into hierarchies.

    For partial joins, the un-mapped key columns of the target table are always converted into hierarchies, regardless of the value of this parameter.

Example

>>> from datetime import date
>>> table = session.create_table(
...     "Product",
...     types={"Date": tt.type.LOCAL_DATE, "Product": tt.type.STRING, "Quantity": tt.type.NULLABLE_DOUBLE},
...     keys=["Date"],
... )
>>> table.head()
Empty DataFrame
Columns: [Product, Quantity]
Index: []
>>> table.append((date(2021, 5, 19), "TV", 15.0))
>>> table.head()
           Product  Quantity
Date
2021-05-19      TV      15.0
Return type

Table

property cubes: atoti.cubes.Cubes

Cubes of the session.

Return type

Cubes

delete_scenario(scenario)

Delete the source scenario with the provided name if it exists.

Return type

None

endpoint(route, *, method='GET')

Create a custom endpoint at /atoti/pyapi/{route}".

This is useful to reuse atoti’s built-in server instead of adding a FastAPI or Flask server to the project. This way, when deploying the project in a container or a VM, only one port (the one of the atoti server) can be exposed instead of two. Since custom endpoints are exposed by atoti’s server, they automatically inherit from the configured atoti.config.session_config.SessionConfig.authentication and atoti.config.session_config.SessionConfig.https parameters.

The decorated function must take three parameters with types User, HttpRequest, and Session and return a response body as a Python data structure that can be converted to JSON.

Parameters
  • route (str) –

    The path suffix after /atoti/pyapi/. For instance, if custom/search is passed, a request to /atoti/pyapi/custom/search?query=test#results will match. The route should not contain the query (?) or fragment (#).

    Path parameters can be configured by wrapping their name in curly braces in the route.

  • method (Literal[‘POST’, ‘GET’, ‘PUT’, ‘DELETE’]) – The HTTP method the request must be using to trigger this endpoint. DELETE, POST, and PUT requests can have a body but it must be JSON.

Example

>>> import requests
>>> df = pd.DataFrame(
...     columns=["Year", "Month", "Day", "Quantity"],
...     data=[
...         (2019, 7, 1, 15),
...         (2019, 7, 2, 20),
...     ],
... )
>>> table = session.read_pandas(df, table_name="Quantity")
>>> table.head()
Year  Month  Day  Quantity
0  2019      7    1        15
1  2019      7    2        20
>>> endpoints_base_url = f"http://localhost:{session.port}/atoti/pyapi"
>>> @session.endpoint("tables/{table_name}/size", method="GET")
... def get_table_size(request, user, session):
...     table_name = request.path_parameters["table_name"]
...     return len(session.tables[table_name])
>>> requests.get(f"{endpoints_base_url}/tables/Quantity/size").json()
2
>>> @session.endpoint("tables/{table_name}/rows", method="POST")
... def append_rows_to_table(request, user, session):
...     rows = request.body
...     table_name = request.path_parameters["table_name"]
...     session.tables[table_name].append(*rows)
>>> requests.post(
...     f"{endpoints_base_url}/tables/Quantity/rows",
...     json=[
...         {"Year": 2021, "Month": 5, "Day": 19, "Quantity": 50},
...         {"Year": 2021, "Month": 5, "Day": 20, "Quantity": 6},
...     ],
... ).status_code
200
>>> requests.get(f"{endpoints_base_url}/tables/Quantity/size").json()
4
>>> table.head()
Year  Month  Day  Quantity
0  2019      7    1        15
1  2019      7    2        20
2  2021      5   19        50
3  2021      5   20         6
Return type

Callable[[Callable[…, Any]], Callable[…, Any]]

explain_mdx_query(mdx, *, timeout=30)

Run the query but return an explanation of how the query was executed instead of its result.

See also

query_mdx() for the roles of the parameters.

Return type

QueryAnalysis

Returns

An explanation containing a summary, global timings, and the query plan with all the retrievals.

export_translations_template(path)

Export a template containing all translatable values in the session’s cubes.

Parameters

path (Union[str, Path]) – The path at which to write the template.

Return type

None

Display a link to this session.

Clicking on the link will open it in a new browser tab.

Note

This method requires the atoti-jupyterlab plugin.

The extension will try to access the session through (in that order):

  1. Jupyter Server Proxy if it is enabled.

  2. f"{session_protocol}//{jupyter_server_hostname}:{session.port}" for Session and session.url for QuerySession.

Parameters

path (str) – The path to append to the session base URL. Defaults to the session home page.

Example

Pointing directly to an existing dashboard:

dashboard_id = "92i"
session.link(path=f"#/dashboard/{dashboard_id}")
Return type

Any

property logs_path: pathlib.Path

Path to the session logs file.

Return type

Path

property name: str

Name of the session.

Return type

str

property port: int

Port on which the session is exposed.

Can be configured with port.

Return type

int

query_mdx(mdx, *, keep_totals=False, timeout=30, mode='pretty')

Execute an MDX query and return its result as a pandas DataFrame.

Parameters
  • mdx (str) – The MDX SELECT query to execute. Regardless of the axes on which levels and measures appear in the MDX, the returned DataFrame will have all levels on rows and measures on columns.

  • keep_totals (bool) – Whether the resulting DataFrame should contain, if they are present in the query result, the grand total and subtotals. Totals can be useful but they make the DataFrame harder to work with since its index will have some empty values.

  • timeout (int) – The query timeout in seconds.

  • mode (Literal[‘pretty’, ‘raw’]) –

    The query mode.

    • "pretty" is best for queries returning small results:

      • A QueryResult will be returned and its rows will be sorted according to the level comparators.

    • "raw" is best for benchmarks or large exports:

    • A faster and more efficient endpoint reducing the data transfer from Java to Python will be used.

    • A classic pandas.DataFrame will be returned.

    • include_totals="True" will not be allowed.

    • The Convert to Widget Below action provided by the atoti-jupyterlab plugin will not be available.

Example

>>> from datetime import date
>>> df = pd.DataFrame(
...     columns=["Country", "Date", "Price"],
...     data=[
...         ("China", date(2020, 3, 3), 410.0),
...         ("China", date(2020, 4, 4), 350.0),
...         ("France", date(2020, 1, 1), 480.0),
...         ("France", date(2020, 2, 2), 500.0),
...         ("France", date(2020, 3, 3), 400.0),
...         ("France", date(2020, 4, 4), 420.0),
...         ("India", date(2020, 1, 1), 360.0),
...         ("India", date(2020, 2, 2), 400.0),
...         ("UK", date(2020, 2, 2), 960.0),
...     ],
... )
>>> table = session.read_pandas(
...     df, keys=["Country", "Date"], table_name="Prices"
... )
>>> cube = session.create_cube(table)

This MDX:

>>> mdx = (
...     "SELECT"
...     "  NON EMPTY Hierarchize("
...     "    DrilldownLevel("
...     "      [Prices].[Country].[ALL].[AllMember]"
...     "    )"
...     "  ) ON ROWS,"
...     "  NON EMPTY Crossjoin("
...     "    [Measures].[Price.SUM],"
...     "    Hierarchize("
...     "      DrilldownLevel("
...     "        [Prices].[Date].[ALL].[AllMember]"
...     "      )"
...     "    )"
...     "  ) ON COLUMNS"
...     "  FROM [Prices]"
... )

would display this pivot table:

Country

Price.sum

Total

2020-01-01

2020-02-02

2020-03-03

2020-04-04

Total

2,280.00

840.00

1,860.00

810.00

770.00

China

760.00

410.00

350.00

France

1,800.00

480.00

500.00

400.00

420.00

India

760.00

360.00

400.00

UK

960.00

960.00

but will return this DataFrame:

>>> session.query_mdx(mdx).sort_index()
                    Price.SUM
Date       Country
2020-01-01 France       480.0
           India        360.0
2020-02-02 France       500.0
           India        400.0
           UK           960.0
2020-03-03 China        410.0
           France       400.0
2020-04-04 China        350.0
           France       420.0
Return type

DataFrame

read_csv(path, *, keys=(), table_name=None, separator=None, encoding='utf-8', process_quotes=None, partitioning=None, types={}, array_separator=None, hierarchized_columns=None, date_patterns={}, client_side_encryption=None, **kwargs)

Read a CSV file into a table.

Parameters
  • path (Union[str, Path]) –

    The path to the CSV file to load.

    .gz, .tar.gz and .zip files containing compressed CSV(s) are also supported.

    The path can also be a glob pattern (e.g. path/to/directory/**.*.csv).

  • keys (Iterable[str]) – The columns that will become keys of the table.

  • table_name (Optional[str]) – The name of the table to create. Required when path is a glob pattern. Otherwise, defaults to the final component of the path argument.

  • separator (Optional[str]) – The character separating the values of each line. the separator will be detected automatically.

  • encoding (str) – The encoding to use to read the CSV.

  • process_quotes (Optional[bool]) –

    Whether double quotes should be processed to follow the official CSV specification:

    • True:

      • Each field may or may not be enclosed in double quotes (however some programs, such as Microsoft Excel, do not use double quotes at all). If fields are not enclosed with double quotes, then double quotes may not appear inside the fields.

      • A double quote appearing inside a field must be escaped by preceding it with another double quote.

      • Fields containing line breaks, double quotes, and commas should be enclosed in double-quotes.

    • False: all double-quotes within a field will be treated as any regular character, following Excel’s behavior. In this mode, it is expected that fields are not enclosed in double quotes. It is also not possible to have a line break inside a field.

    • None: The behavior will be inferred from the first lines of the CSV file.

  • partitioning (Optional[str]) –

    The description of how the data will be split across partitions of the table.

    Joined tables can only use a sub-partitioning of the table referencing them.

    Example

    hash4(country) splits the data across 4 partitions based on the country column’s hash value.

  • types (Mapping[str, DataType]) – Types for some or all columns of the table. Types for non specified columns will be inferred from the first 1,000 lines.

  • array_separator (Optional[str]) – The character separating array elements. Setting it to a non-None value will parse all the columns containing this separator as arrays.

  • hierarchized_columns (Optional[Iterable[str]]) –

    The list of columns which will automatically be converted into hierarchies no matter which creation mode is used for the cube.

    The different behaviors based on the passed value are:

    • None: all non-numeric columns are converted into hierarchies, depending on the cube’s creation mode.

    • Empty collection: no columns are converted into hierarchies.

    • Non-empty collection: only the columns in the collection will be converted into hierarchies.

    For partial joins, the un-mapped key columns of the target table are always converted into hierarchies, regardless of the value of this parameter.

  • date_patterns (Mapping[str, str]) – A column name to date pattern mapping that can be used when the built-in date parsers fail to recognize the formatted dates in the passed files.

  • client_side_encryption (Optional[ClientSideEncryption]) – The client side encryption configuration to use when loading data.

Return type

Table

Returns

The created table holding the content of the CSV file(s).

read_numpy(array, *, columns, table_name, keys=(), partitioning=None, types={}, hierarchized_columns=None, **kwargs)

Read a NumPy 2D array into a new table.

Parameters
  • array (ndarray) – The NumPy 2D ndarray to read the data from.

  • columns (Sequence[str]) – The names to use for the table’s columns. They must be in the same order as the values in the NumPy array.

  • table_name (str) – The name of the table to create.

  • keys (Iterable[str]) – The columns that will become keys of the table.

  • partitioning (Optional[str]) –

    The description of how the data will be split across partitions of the table.

    Joined tables can only use a sub-partitioning of the table referencing them.

    Example

    hash4(country) splits the data across 4 partitions based on the country column’s hash value.

  • types (Mapping[str, DataType]) – Types for some or all columns of the table. Types for non specified columns will be inferred from numpy data types.

  • hierarchized_columns (Optional[Iterable[str]]) –

    The list of columns which will automatically be converted into hierarchies no matter which creation mode is used for the cube.

    The different behaviors based on the passed value are:

    • None: all non-numeric columns are converted into hierarchies, depending on the cube’s creation mode.

    • Empty collection: no columns are converted into hierarchies.

    • Non-empty collection: only the columns in the collection will be converted into hierarchies.

    For partial joins, the un-mapped key columns of the target table are always converted into hierarchies, regardless of the value of this parameter.

Return type

Table

Returns

The created table holding the content of the array.

read_pandas(dataframe, *, table_name, keys=(), partitioning=None, types={}, hierarchized_columns=None, **kwargs)

Read a pandas DataFrame into a table.

All the named indices of the DataFrame are included into the table. Multilevel columns are flattened into a single string name.

Parameters
  • dataframe (DataFrame) – The DataFrame to load.

  • table_name (str) – The name of the table to create.

  • keys (Iterable[str]) – The columns that will become keys of the table.

  • partitioning (Optional[str]) –

    The description of how the data will be split across partitions of the table.

    Joined tables can only use a sub-partitioning of the table referencing them.

    Example

    hash4(country) splits the data across 4 partitions based on the country column’s hash value.

  • types (Mapping[str, DataType]) – Types for some or all columns of the table. Types for non specified columns will be inferred from pandas dtypes.

  • hierarchized_columns (Optional[Iterable[str]]) –

    The list of columns which will automatically be converted into hierarchies no matter which creation mode is used for the cube.

    The different behaviors based on the passed value are:

    • None: all non-numeric columns are converted into hierarchies, depending on the cube’s creation mode.

    • Empty collection: no columns are converted into hierarchies.

    • Non-empty collection: only the columns in the collection will be converted into hierarchies.

    For partial joins, the un-mapped key columns of the target table are always converted into hierarchies, regardless of the value of this parameter.

Return type

Table

Returns

The created table holding the content of the DataFrame.

read_parquet(path, *, keys=(), table_name=None, partitioning=None, hierarchized_columns=None, client_side_encryption=None, **kwargs)

Read a Parquet file into a table.

Parameters
  • path (Union[str, Path]) – The path to the Parquet file. If a path pointing to a directory is provided, all of the files with the .parquet extension in the directory will be loaded into the same table and, as such, they are all expected to share the same schema. The path can also be a glob pattern (e.g. path/to/directory/**.*.parquet).

  • keys (Iterable[str]) – The columns that will become keys of the table.

  • table_name (Optional[str]) – The name of the table to create. Required when path is a glob pattern. Otherwise, defaults to the final component of the path argument.

  • partitioning (Optional[str]) –

    The description of how the data will be split across partitions of the table.

    Joined tables can only use a sub-partitioning of the table referencing them.

    Example

    hash4(country) splits the data across 4 partitions based on the country column’s hash value.

  • hierarchized_columns (Optional[Iterable[str]]) –

    The list of columns which will automatically be converted into hierarchies no matter which creation mode is used for the cube.

    The different behaviors based on the passed value are:

    • None: all non-numeric columns are converted into hierarchies, depending on the cube’s creation mode.

    • Empty collection: no columns are converted into hierarchies.

    • Non-empty collection: only the columns in the collection will be converted into hierarchies.

    For partial joins, the un-mapped key columns of the target table are always converted into hierarchies, regardless of the value of this parameter.

  • client_side_encryption (Optional[ClientSideEncryption]) – The client side encryption configuration to use when loading data.

Return type

Table

Returns

The created table holding the content of the Parquet file(s).

read_spark(dataframe, *, table_name, keys=(), partitioning=None, hierarchized_columns=None, **kwargs)

Read a Spark DataFrame into a table.

Parameters
  • dataframe – The DataFrame to load.

  • table_name (str) – The name of the table to create.

  • keys (Iterable[str]) – The columns that will become keys of the table.

  • partitioning (Optional[str]) –

    The description of how the data will be split across partitions of the table.

    Joined tables can only use a sub-partitioning of the table referencing them.

    Example

    hash4(country) splits the data across 4 partitions based on the country column’s hash value.

  • hierarchized_columns (Optional[Iterable[str]]) –

    The list of columns which will automatically be converted into hierarchies no matter which creation mode is used for the cube.

    The different behaviors based on the passed value are:

    • None: all non-numeric columns are converted into hierarchies, depending on the cube’s creation mode.

    • Empty collection: no columns are converted into hierarchies.

    • Non-empty collection: only the columns in the collection will be converted into hierarchies.

    For partial joins, the un-mapped key columns of the target table are always converted into hierarchies, regardless of the value of this parameter.

Return type

Table

Returns

The created table holding the content of the DataFrame.

read_sql(query, *, url, table_name, driver=None, keys=(), partitioning=None, types={}, hierarchized_columns=None)

Create a table from the result of the passed SQL query.

Note

This method requires the atoti-sql plugin.

Parameters
  • query (str) – The result of this SQL query will be loaded into the table.

  • url (str) –

    The JDBC connection URL of the database. The jdbc: prefix is optional but the database specific part (such as h2: or mysql:) is mandatory. For instance:

    • h2:file:/home/user/database/file/path;USER=username;PASSWORD=passwd

    • mysql://localhost:7777/example?user=username&password=passwd

    • postgresql://postgresql.db.server:5430/example?user=username&password=passwd

    More examples can be found here.

  • driver (Optional[str]) – The JDBC driver used to load the data. If None, the driver is inferred from the URL. Drivers can be found in the atoti_sql.drivers module.

  • table_name (str) – The name of the table to create.

  • keys (Iterable[str]) – The columns that will become keys of the table.

  • partitioning (Optional[str]) –

    The description of how the data will be split across partitions of the table.

    Joined tables can only use a sub-partitioning of the table referencing them.

    Example

    hash4(country) splits the data across 4 partitions based on the country column’s hash value.

  • types (Mapping[str, DataType]) – Types for some or all columns of the table. Types for non specified columns will be inferred from the SQL types.

  • hierarchized_columns (Optional[Iterable[str]]) –

    The list of columns which will automatically be converted into hierarchies no matter which creation mode is used for the cube.

    The different behaviors based on the passed value are:

    • None: all non-numeric columns are converted into hierarchies, depending on the cube’s creation mode.

    • Empty collection: no columns are converted into hierarchies.

    • Non-empty collection: only the columns in the collection will be converted into hierarchies.

    For partial joins, the un-mapped key columns of the target table are always converted into hierarchies, regardless of the value of this parameter.

Example

>>> table = session.read_sql(
...     "SELECT * FROM MYTABLE;",
...     url=f"h2:file:{RESOURCES}/h2-database;USER=root;PASSWORD=pass",
...     table_name="Cities",
...     keys=["ID"],
... )
>>> len(table)
5
Return type

Table

property scenarios: Sequence[str]

Collection of source scenarios of the session.

Return type

Sequence[str]

property security: atoti_plus.security.Security
Return type

Security

start_transaction(scenario_name='Base')

Start a transaction to batch several table operations.

  • It is more efficient than doing each table operation one after the other.

  • It avoids possibly incorrect intermediate states (e.g. if loading some new data requires dropping existing rows first).

Note

Some operations are not allowed during a transaction:

Parameters

scenario_name (str) – The name of the source scenario impacted by all the table operations inside the transaction.

Example

>>> df = pd.DataFrame(
...     columns=["City", "Price"],
...     data=[
...         ("Berlin", 150.0),
...         ("London", 240.0),
...         ("New York", 270.0),
...         ("Paris", 200.0),
...     ],
... )
>>> table = session.read_pandas(
...     df, keys=["City"], table_name="start_transaction example"
... )
>>> cube = session.create_cube(table)
>>> extra_df = pd.DataFrame(
...     columns=["City", "Price"],
...     data=[
...         ("Singapore", 250.0),
...     ],
... )
>>> with session.start_transaction():
...     table += ("New York", 100.0)
...     table.drop({"City": "Paris"})
...     table.load_pandas(extra_df)
...
>>> table.head().sort_index()
           Price
City
Berlin     150.0
London     240.0
New York   100.0
Singapore  250.0
Return type

Transaction

property tables: atoti.tables.Tables

Tables of the session.

Return type

Tables

visualize(name=None)

Display an atoti widget to explore the session interactively.

Note

This method requires the atoti-jupyterlab plugin.

The widget state will be stored in the cell metadata. This state should not have to be edited but, if desired, it can be found in JupyterLab by opening the “Notebook tools” sidebar and expanding the “Advanced Tools” section.

Parameters

name (Optional[str]) – The name to give to the widget.

wait()

Wait for the underlying server subprocess to terminate.

This will prevent the Python process to exit.

Return type

None