Options

Options #

There is a small set of runtime knobs exposed under [options]. Most workflows can rely on the defaults.

[options]
    thread_count = -1
    block_size = 10000
    buffer_size = 102400
    accept_duplicate_files = false
KeyDefaultDescription
thread_count-1Worker threads for transformations. -1 autotunes per CPU; most runtime is still dominated by decompression threads, so gains are modest.
block_size10000Number of fragments pulled per batch. Increase for very large runs when IO is abundant; decrease to reduce peak memory use.
buffer_size102400Initial bytes reserved per block. The allocator grows buffers on demand, so tuning is rarely necessary.
accept_duplicate_filesfalsePermit the same path to appear multiple times across segments. Useful for fixtures or synthetic tests; keep disabled to catch accidental copy/paste errors.

Changing these knobs can affect memory pressure and concurrency behaviour. Measure before and after if you deviate from defaults.