27391924006
85%

Ran 12 Jun 2026 03:09AM UTC

Jobs 1

Files 9

Run time 1min

Badge

Embed ▾

Committed 12 Jun 2026 03:08AM UTC coverage: 85.562%. Remained the same

Build # 27391924006

Build Type

push

github

Committed by

freeeve

Commit Message

feat(rust): TermIndexStreamWriter + slice_term_monolith — geometric splits at disk speed

The batch RRTI writer is now a thin wrapper over a streaming writer (terms
pushed in sorted order, postings region streamed to the caller's sink, dict
blocks resident) — the existing goldens prove the refactor byte-identical.
parse_dict_block gains head_size so a sequential consumer can pace a
lockstep read of the postings region.

slice_term_monolith builds a term split set by SLICING the monolith .rrt
along doc ranges instead of re-tokenizing the corpus: one sequential pass
over dictionary + postings fanned into N streaming split writers — the
shared rank-ordered doc-ID space makes a split's posting the monolith
posting restricted to [lo, hi), rebased. Doubling doc-count ranges mirror
the builders' byte_cap_max tiering. --selftest proves builder and slicer
emit byte-identical split sets given the same ranges. (The greedy
TermSplitSetBuilder degrades badly at multi-GiB caps — its open-split map
ran at ~1.9K docs/s, a 71 h full-corpus ETA; the slice is ~disk speed.)

Coverage Stats

966 of 1129 relevant lines covered (85.56%)

28.47 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	27391924006.1	12 Jun 2026 03:09AM UTC	9	85.56	GitHub Action Run

freeeve / roaringrange / 27391924006
85%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 27391924006

freeeve / roaringrange / 27391924006 85%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 27391924006

freeeve / roaringrange / 27391924006
85%

README BADGES
x