A :class:`~google.cloud.spanner.snapshot.Snapshot` represents a read-only transaction: when multiple read operations are peformed via a Snapshot, the results are consistent as of a particular point in time.

To begin using a snapshot using the default "bound" (which is "strong"), meaning all reads are performed at a timestamp where all previously-committed transactions are visible:

snapshot = session.snapshot()

You can also specify a weaker bound, which can either be to perform all reads as of a given timestamp:

import datetime
from pytz import UTC
TIMESTAMP = datetime.utcnow().replace(tzinfo=UTC)
snapshot = session.snapshot(read_timestamp=TIMESTAMP)

or as of a given duration in the past:

import datetime
DURATION = datetime.timedelta(seconds=5)
snapshot = session.snapshot(exact_staleness=DURATION)

Read data for selected rows from a table in the session's database. Calls the Read API, which returns all rows specified in key_set, or else fails if the result set is too large,

result = snapshot.read(
    table='table-name', columns=['first_name', 'last_name', 'age'],
    key_set=['phred@example.com', 'bharney@example.com'])

for row in result.rows:
    print(row)

Read data from a query against tables in the session's database. Calls the ExecuteSql API, which returns all rows matching the query, or else fails if the result set is too large,

QUERY = (
    'SELECT e.first_name, e.last_name, p.telephone '
    'FROM employees as e, phones as p '
    'WHERE p.employee_id == e.employee_id')
result = snapshot.execute_sql(QUERY)

for row in result.rows:
    print(row)

Next, learn about :doc:`transaction-usage`.