Quickstart

Full code here

The most simple graph is two nodes with no time component, no aggregations, and no labels. We are using customer (cust.csv) and order data (orders.csv).

In plain terms what the below code does is as follows:

1. Create prefixes for each node so we always know where the column originated from after joinining the two datasets.
2. Instantiate two DynamicNode instances: one node for cust.csv and another for orders.csv.
3. Instantiate the GraphReduce object to house the compute graph. We are specifying that the cust.csv node is the parent_node, which means all data will be joined to and aggregated to the cust.csv node. In cases where we reduce the data the resulting dataset should be at the ganularity of the parent_node dimension.
4. Add the nodes.
5. Add the edge between the nodes.
6. Execute the computation with GraphReduce.do_transformations() the primary entrypoint to execution.
7. Dump out the head of the computed dataframe.

from graphreduce.node import GraphReduceNode, DynamicNode
from graphreduce.graph_reduce import GraphReduce
from graphreduce.enum import ComputeLayerEnum as GraphReduceComputeLayerEnum, PeriodUnit

# Need unique prefixes for all nodes
# so when columns are merged we know
# where they originate from.
prefixes = {
    'cust.csv' : {'prefix':'cu'},
    'orders.csv':{'prefix':'ord'}
}

# create graph reduce nodes
gr_nodes = {
    f.split('/')[-1]: DynamicNode(
        fpath=f,
        fmt='csv',
        pk='id',
        prefix=prefixes[f]['prefix'],
        date_key=None,
        compute_layer=GraphReduceComputeLayerEnum.pandas,
        compute_period_val=730,
        compute_period_unit=PeriodUnit.day,
    )
    for f in files.keys()
}

# Instantiate GraphReduce with params.
# We are using 'cust.csv' as parent node
# so the granularity should be at the customer
# dimension.
gr = GraphReduce(
    name='starter_graph',
    parent_node=gr_nodes['cust.csv'],
    fmt='csv',
    cut_date=datetime.datetime(2023,9,1),
    compute_layer=GraphReduceComputeLayerEnum.pandas,
    auto_features=True,
    auto_feature_hops_front=1,
    auto_feature_hops_back=2,
    label_node=gr_nodes['orders.csv'],
    label_operation='count',
    label_field='id',
    label_period_val=60,
    label_period_unit=PeriodUnit.day
)

# Add the nodes to the GraphReduce instance.
gr.add_node(gr_nodes['cust.csv'])
gr.add_node(gr_nodes['orders.csv'])

gr.add_entity_edge(
    parent_node=gr_nodes['cust.csv'],
    relation_node=gr_nodes['orders.csv'],
    parent_key='id',
    relation_key='customer_id',
    reduce=True
)

gr.do_transformations()

gr.parent_node.df.head()
    cu_id   cu_name ord_customer_id ord_id_count    ord_customer_id_count   ord_ts_min  ord_ts_max  ord_amount_count    ord_customer_id_dupe    ord_id_label
0   1   wes 1   3   3   2023-05-12  2023-09-02  3   1   3
1   2   ana 2   3   3   2022-08-05  2023-10-15  3   2   3
2   3   caleb   3   1   1   2023-06-01  2023-06-01  1   3   1
3   4   luly    4   2   2   2024-01-01  2024-02-01  2   4   2