Home › Examples › Tags

Tags from a name

bollhav derives a model's tags from its identity — the catalog, schema, and name — so TAGS= run selection and lineage filtering work without you hand-labelling anything. snake_case is split on underscores; turn on the unpascal_* flags and PascalCase is split into its words too (lowercased).

Take a model whose target is AnalyticsLakehouse.curated_clean_entities.WeeklyChannelAttributionRollupFact:

weekly_rollup.py

from bollhav.model import Database, Model, Tags, Target, Temporality, WriteMode
from bollhav.postgres import PostgresColumn, PostgresType


weekly_rollup = Model(
    temporality=Temporality.TEMPORAL,
    target=Target(
        name="WeeklyChannelAttributionRollupFact",
        schema="curated_clean_entities",
        catalog="AnalyticsLakehouse",
        database=Database.POSTGRES,
        write_mode=WriteMode.APPEND,
        columns=[
            PostgresColumn("week", PostgresType.TIMESTAMPTZ, nullable=False),
            PostgresColumn("channel", PostgresType.TEXT, nullable=False),
            PostgresColumn("attributed_revenue", PostgresType.NUMERIC, nullable=False),
        ],
    ),
    tagging=Tags(
        unpascal_name_for_tags=True,     # default: False
        unpascal_schema_for_tags=True,   # default: False
        unpascal_catalog_for_tags=True,  # default: False
    ),
)

That single declaration produces this tag set:

all

AnalyticsLakehouse                  →  analytics, lakehouse          (catalog, un-pascal'd)
curated_clean_entities              →  curated, clean, entities      (schema, un-snake'd)
WeeklyChannelAttributionRollupFact  →  weekly, channel, attribution, rollup, fact   (name, un-pascal'd)

curated_clean_entities.WeeklyChannelAttributionRollupFact
AnalyticsLakehouse.curated_clean_entities.WeeklyChannelAttributionRollupFact

What gets added

Rule	From the example
the bare catalog / schema / name	AnalyticsLakehouse, curated_clean_entities, WeeklyChannelAttributionRollupFact
unsnake_* — split on _	curated, clean, entities
unpascal_* — split PascalCase (lowercased)	analytics, lakehouse, weekly, channel, attribution, rollup, fact
the schema-qualified name	curated_clean_entities.WeeklyChannelAttributionRollupFact
the fully-qualified name	AnalyticsLakehouse.curated_clean_entities.WeeklyChannelAttributionRollupFact
all (every model gets it)	all

Any explicit tags={...} you pass are merged in on top. Defaults: the bare names, unsnake_*, the qualified names, and all are on; unpascal_* is off (enable it for PascalCase names like above).

Now TAGS="[clean & fact]" selects this model — and so does the tag filter in the GUI, which uses the very same matching. See Tags for the expression syntax and Tagging for the full set of Tags options.