bollhav 1.1.3

Standardizing models

Bollhav

A lightweight config class for defining data pipeline models.

Usage

Basic table

model = Model(
    name="orders",
    source_table="raw.orders",
    table="orders",
    schema="public",
    write_mode=WriteMode.TRUNCATE_INSERT,
    database=Database.POSTGRES,
    columns=[
        PostgresColumn(name="id", data_type=PostgresType.BIGINT, primary_key=True, nullable=False),
        PostgresColumn(name="customer_id", data_type=PostgresType.BIGINT),
        PostgresColumn(name="amount", data_type=PostgresType.NUMERIC, precision=10, scale=2),
        PostgresColumn(name="created_at", data_type=PostgresType.TIMESTAMPTZ),
    ],
)

View

model = Model(
    name="orders_view",
    source_table="raw.orders",
    table="orders_view",
    schema="public",
    model_type=ModelType.VIEW,
    write_mode=WriteMode.VIEW,
    database=Database.POSTGRES,
    columns=[
        PostgresColumn(name="id", data_type=PostgresType.BIGINT),
        PostgresColumn(name="amount", data_type=PostgresType.NUMERIC, precision=10, scale=2),
    ],
)

With a schedule

model = Model(
    name="orders",
    source_table="raw.orders",
    database=Database.POSTGRES,
    columns=[
        PostgresColumn(name="id", data_type=PostgresType.BIGINT),
        PostgresColumn(name="amount", data_type=PostgresType.NUMERIC, precision=10, scale=2),
    ],
    cron="0 3 * * *",
)
model.batch_size  # BatchSize.DAILY

batch_size is inferred from the cron expression and is read-only.

BatchSize	Example cron
YEARLY	0 0 1 1 *
MONTHLY	0 0 1 * *
WEEKLY	0 0 * * 0
DAILY	0 3 * * *
HOURLY	0 * * * *

With dynamic kwargs

def my_ddl(table_name: str, schema: str, **kwargs) -> str:
    return f"CREATE TABLE {schema}.{table_name} (id SERIAL PRIMARY KEY);"

model = Model(
    name="orders",
    source_table="raw.orders",
    database=Database.POSTGRES,
    columns=[
        PostgresColumn(name="id", data_type=PostgresType.BIGINT, primary_key=True, nullable=False),
    ],
    ddl=my_ddl,
    table_name="orders",
    schema="public",
)
model.extra["ddl"]  # resolved DDL string

Callables in **kwargs are resolved at init using the non-callable kwargs as arguments.

With debug

model = Model(
    name="orders",
    source_table="raw.orders",
    debug=True,
)

Write modes

WriteMode	ModelType	Description
APPEND	TABLE	Insert without truncating
TRUNCATE_INSERT	TABLE	Truncate then insert
OVERWRITE_INSERT	TABLE	Overwrite matching rows
MERGE	TABLE	Upsert based on keys
VIEW	VIEW	Create or replace view

ModelType and WriteMode are validated against each other at init.

PostgresColumn

Field	Type	Default	Notes
name	str	required
data_type	PostgresType	required
nullable	bool	True
order	int | None	None
primary_key	bool	False	Implies nullable=False
unique	bool	False
precision	int | None	None	For NUMERIC, DECIMAL
scale	int | None	None	For NUMERIC, DECIMAL
length	int | None	None	For VARCHAR, CHAR, BIT

primary_key=True with nullable=True raises at init.

Tags

Optional and freeform.

model = Model(
    name="orders",
    source_table="raw.orders",
    database=Database.POSTGRES,
    columns=[
        PostgresColumn(name="id", data_type=PostgresType.BIGINT),
    ],
    tags=["finance", "critical"],
)