The declarative shape most datasets need: one sample row per row of one event
table, keyed dataset_key:sample_type:id, with no parent. Exported because a
dataset's projection belongs in its own ingest notebook — this is what keeps that
a one-line declaration rather than copied SQL.
Usage
sample_arm_self(
dataset_key,
tbl,
id_col,
sample_type,
dt_col = "datetime_start_utc",
grid_expr = "grid_key",
site_expr = "NULL::VARCHAR",
ord_expr = "NULL::INTEGER",
depth_min = "0::DOUBLE",
depth_max = "0::DOUBLE"
)Value
a SQL SELECT string for append_sample()