Batches partition people, keeping all their windows together. Batch size is
not a result-row limit. No writes or temporary DB tables are used. A failed
batch raises an error, never a partial successful result. All returned records
are ultimately assembled in R memory; batching does not make collect out of core.
Usage
# S3 method for class 'thip_window_plan'
collect(x, ..., batch_size = 1000L, max_query_bytes = 1048576L)
Arguments
- x
A plan from thip_window().
- ...
Reserved; unused arguments are rejected.
- batch_size
Maximum distinct people per batch, default 1000.
- max_query_bytes
Maximum generated SQL bytes per batch. Oversized batches
fail with guidance to reduce batch_size, rather than silently truncating.
Value
A data frame with a window_report attribute containing batch counts,
completion and unmatched-entry counts. Output row order is unspecified.