nx_distexec.nx

nx_distexec.nx -- DISTRIBUTED EXECUTION ALGEBRA (distributed-compute, F-dist). Spark's essence is not the cluster -- it is the PARTITION -> MAP -> MERGE algebra: split the data into partitions, compute a PARTIAL result per partition, then MERGE the partials into the whole answer. That algebra is provable and buildable single-process from the first byte up; only its PARALLEL execution (a partition per worker core/node) needs the threading/swarm arc. This organ is the algebra, proven PARTITION-INVARIANT: the merged result is byte-identical for ANY partition count P, which is exactly the correctness a scheduler relies on to run partitions anywhere. Composes nx_vecexec (each partition aggregates vectorized) over the zero-copy nx_colframe. Honest: this is the algebra + a sequential executor; real multi-core/multi-node parallelism is the named finer gap. license_tier: ORIGINAL No hardware writes (Rule 26).