Sessions are one of the key parts of interacting with the database. They provide a wrapper around a transaction object, providing an API which uses table row instances and query objects to interacting with the database connected to your application.

Creating a session

Creating a new Session that is bound to the current database interface is simple via the usage of DatabaseInterface.get_session().

# create a session bound to our current database
# this will automatically provide the ability to get transactions from the db
sess = db.get_session()

# alternatively, you can create your own sessions bound to the database interface
# providng a custom subclass or so on
session = Session(bind=db)

Using the session requires beginning it; behind the scenes this will acquire a new transaction object from the current connector, and emit a BEGIN statement to start the transaction.

# begin and connect the session
await session.begin()

The session object also supports the async with protocol, meaning you can have it automatically open and close without calling the begin/close methods.

async with session:

# or alternatively
async with db.get_session() as sess:

Running SQL

The most basic thing you can do with a session is to run some SQL code, using either Session.execute() or Session.cursor(). The former is used for queries without a result, the latter is used to execute and return a result.

For example, to fetch the result of the sum 1 + 1, you would use:

cursor = await session.cursor("SELECT 1+1;")

This returns an instance of the the low-level object BaseResultSet. To fetch the result, you can use BaseResultSet.fetch_row():

result = await cursor.fetch_row()
answer = result["?column?"]  # postgres example
answer = list(result.values())[0]  # or the list form for cross-db compatability

Inserting Rows

The session is the one-stop gateway to inserting, updating, or even deleting Row Objects. There are several methods used: Session.add(), Session.merge(), and Session.remove() are the high level methods.

  • Session.add() is used for new rows, or rows that have been retrieved from a query.
  • Session.merge() is used for rows that already exist in the database.
  • Session.remove() is used to delete rows that exist in the database.

For example, to add a user to the DB:

u = User(id=1, name="heck")
await session.add(u)

You can also update a user in the database as long as the row you’re providing has a primary key, and you use the merge method:

u = User(id=1) = "not heck"
await session.merge(u)

Querying with the Session

See querying for an explanation of how to query using the session object.

class asyncqlio.orm.session.Session(bind, **kwargs)[source]

Bases: asyncqlio.orm.session.SessionBase

Sessions act as a temporary window into the database. They are responsible for creating queries, inserting and updating rows, etc.

Sessions are bound to a DatabaseInterface instance which they use to get a transaction and execute queries in.

# get a session from our db interface
sess = db.get_session()
Parameters:bind (DatabaseInterface) – The DatabaseInterface instance we are bound to.

Creates a new SELECT query that can be built upon.

Return type:SelectQuery
Returns:A new SelectQuery.

Creates a new INSERT INTO query that can be built upon.

Return type:InsertQuery
Returns:A new InsertQuery.

Creates a new UPDATE query that can be built upon.

Return type:BulkUpdateQuery
Returns:A new BulkUpdateQuery.

Creates a new DELETE query that can be built upon.

Return type:BulkDeleteQuery
Returns:A new BulkDeleteQuery.
coroutine add(self, row)[source]

Adds a row to the current transaction. This will emit SQL that will generate an INSERT or UPDATE statement, and then update the primary key of this row.


This will only generate the INSERT statement for the row now. Only Session.commit() will actually commit the row to storage.

Parameters:row (Table) – The Table instance object to add to the transaction.
Return type:Table
Returns:The Table instance with primary key filled in, if applicable.
close(*, has_error=False)

Closes the current session.

Parameters:has_error (bool) – If this session had an error. Internal usage only.


This will NOT COMMIT ANY DATA. Old data will die.


Commits the current session, running inserts/updates/deletes.

This will not close the session; it can be re-used after a commit.

Return type:SessionBase
cursor(sql, params=None)

Executes SQL inside the current session, and returns a new BaseResultSet.

Return type:



Deletes a row NOW.

Return type:Table
execute(sql, params=None)

Executes SQL inside the current session.

This is part of the low-level API.

fetch(sql, params=None)

Fetches a single row.

Return type:Mapping[~KT, +VT_co]

Inserts a row NOW.


This will only generate the INSERT statement for the row now. Only Session.commit() will actually commit the row to storage.

Also, tables with auto-incrementing fields will only have their first field filled in outside of Postgres databases.

Parameters:row (Table) – The Table instance to insert.
Return type:Any
Returns:The row, with primary key included.
coroutine merge(self, row)[source]

Merges a row with a row that already exists in the database.

This should be used for rows that have a primary key, but were not returned from

Parameters:row (Table) – The Table instance to merge.
Return type:Table
Returns:The Table instance once updated.
coroutine remove(self, row)[source]

Removes a row from the database.

Parameters:row (Table) – The Table instance to remove.
Return type:Table

Rolls the current session back. This is useful if an error occurs inside your code.

Parameters:checkpoint (Optional[str]) – The checkpoint to roll back to, if applicable.
Return type:SessionBase
coroutine run_delete_query(self, query)[source]

Executes a delete query.

Parameters:query (RowDeleteQuery) – The RowDeleteQuery or BulkDeleteQuery to execute.
coroutine run_insert_query(self, query)[source]

Executes an insert query.

Parameters:query (InsertQuery) – The InsertQuery to use.
Returns:The list of rows that were inserted.
coroutine run_select_query(self, query)[source]

Executes a select query.


Unlike the other run_*_query methods, this method should not be used without a good reason; it creates a special class that is used for the query.

Use SelectQuery.first or SelectQuery.all.

Parameters:query (SelectQuery) – The SelectQuery to use.
Returns:A _ResultGenerator for this query.
coroutine run_update_query(self, query)[source]

Executes an update query.

Parameters:query (BaseQuery) – The RowUpdateQuery or BulkUpdateQuery to execute.
coroutine start(self)

Starts the session, acquiring a transaction connection which will be used to modify the DB. This must be called before using the session.

sess = db.get_session()  # or get_ddl_session etc
await sess.start()


When using async with, this is automatically called.

Return type:SessionBase
coroutine truncate(self, table, *, cascade=False)[source]

Truncates a table.

  • table (Type[Table]) – The table to truncate.
  • cascade (bool) – If this truncate should cascade to other tables.

Updates a row NOW.


This will only generate the UPDATE statement for the row now. Only Session.commit() will actually commit the row to storage.

Parameters:row (Table) – The Table instance to update.
Return type:Table
Returns:The Table instance that was updated.