27321540146
65%

Ran 11 Jun 2026 03:20AM UTC

Jobs 1

Files 7

Run time 1min

Badge

Embed ▾

Committed 11 Jun 2026 03:20AM UTC coverage: 85.137%. Remained the same

Build # 27321540146

Build Type

push

github

Committed by

freeeve

Commit Message

feat(rust): out-of-core full-corpus RRSB builder + bm25_query harness

build_impacts builds a .rrb over an RRSR record store and its finished .rrt
without holding the corpus stats in RAM: doc-chunked spill of term-sorted
(doc, tf) runs (resumable — complete run files are skipped), then a k-way
merge in lockstep with a streaming dictionary scan, failing loudly on any
dictionary/run divergence. Quantization and layout share the in-RAM writer's
code (bm25 format constants and quantize_impact are now pub), and --selftest
proves both paths byte-identical on a synthetic corpus. Verified on the
first 1M real docs against openalex-1m-stem.rrt (89 s, 72 MB sidecar, all
531,680 dictionary terms matched).

TermIndex gains dict_block_locs + parse_dict_block (streaming dictionary
scan for vocabularies too large to materialize — the 484M corpus has ~187M
terms) and Tokenizer gains spec() so build tooling can construct identical
per-thread tokenizers. bm25_query is a local .rrt+.rrb query harness;
build_global_bloom and build_trigram_splitset get their missing example
feature gates.

Coverage Stats

653 of 767 relevant lines covered (85.14%)

11.68 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	27321540146.1	11 Jun 2026 03:20AM UTC	7	85.14	GitHub Action Run

freeeve / roaringrange / 27321540146
65%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 27321540146

freeeve / roaringrange / 27321540146 65%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 27321540146

freeeve / roaringrange / 27321540146
65%

README BADGES
x