Join Example

This section will introduce some of the basic concepts of merging and joining streams of data.

Multiple input sources

Bytewax dataflows can receive input from multiple sources. In the following example, we create two TestingSource sources and add them to our Dataflow as input.

from bytewax import operators as op

from bytewax.connectors.stdio import StdOutSink
from bytewax.dataflow import Dataflow
from bytewax.testing import TestingSource

flow = Dataflow("join")

src_1 = [
    {"user_id": "123", "name": "Bumble"},
]
inp1 = op.input("inp1", flow, TestingSource(src_1))

src_2 = [
    {"user_id": "123", "email": "bee@bytewax.com"},
    {"user_id": "456", "email": "hive@bytewax.com"},
]
inp2 = op.input("inp2", flow, TestingSource(src_2))

In order for our dataflow to process input from either of these sources, we’ll need to create a Stream that combines input from both of them, we can use the merge operator to do so:

merged_stream = op.merge("merge", inp1, inp2)

Now that we have our merged stream, we can write it to standard out:

op.inspect("debug", merged_stream)

$ python -m bytewax.run merge_example

join.debug: {'user_id': '123', 'name': 'Bumble'}
join.debug: {'user_id': '123', 'email': 'bee@bytewax.com'}
join.debug: {'user_id': '456', 'email': 'hive@bytewax.com'}

The dataflow will stop once all input sources are completely exhausted. Even though the input sources have different numbers of items, we see all of them.

Joining streams

To create a streaming join of data from both of our input sources, we’ll need to first choose a key that we want to join our streams on. In our example data, we’ll use the user_id field.

from bytewax import operators as op

from bytewax.connectors.stdio import StdOutSink
from bytewax.dataflow import Dataflow
from bytewax.testing import TestingSource

flow = Dataflow("join")

src_1 = [
    {"user_id": "123", "name": "Bumble"},
]
inp1 = op.input("inp1", flow, TestingSource(src_1))
keyed_inp_1 = op.key_on("key_stream_1", inp1, lambda x: x["user_id"])
src_2 = [
    {"user_id": "123", "email": "bee@bytewax.com"},
    {"user_id": "456", "email": "hive@bytewax.com"},
]
inp2 = op.input("inp2", flow, TestingSource(src_2))
keyed_inp_2 = op.key_on("key_stream_2", inp2, lambda x: x["user_id"])

Now that we have our two keyed streams of data, we can join them together with the join operator.

When creating a dataflow, you can use the inspect operator to view the data in a stream. The inspect operator can be used multiple times and counts as an output (recall that every dataflow requires an output).

merged_stream = op.join("join", keyed_inp_1, keyed_inp_2)
op.inspect("debug", merged_stream)

Running this example, we should see the following output for our stream, which includes the step_id for our inspect operator.

$ python -m bytewax.run join_example

join.debug: ('123', ({'user_id': '123', 'name': 'Bumble'}, {'user_id': '123', 'email': 'bee@bytewax.com'}))

Notice that we don’t see any output for user_id 456. Since we didn’t receive any input for that key from inp2, we won’t see any output for that user until we do.

For more details about the behavior of the join operator, see the Joins section of the documentation.