Created: 2026-07-09 Purpose: Resolve every branch name cited in this repository's scientific ledgers to a durable, permanent Git reference.
The scientific ledgers in docs/research/ and docs/optimization/ cite the
branch on which each experiment was performed — this is what makes a recorded
result reproducible rather than merely asserted. Branch names, however, are
mutable local pointers: once a branch is deleted, its name no longer resolves,
and a citation such as **Branch**: feat/wallbreaker-freq-split becomes a
dangling reference.
On 2026-07-09 the repository's 19 stray feature branches were retired. This table preserves the mapping from every cited branch name to the immutable commit SHA it pointed at, together with the durable reference through which that commit remains reachable. No commit referenced by any ledger was lost.
A commit is reachable — and therefore permanently safe from garbage
collection — if it is an ancestor of master, or if an annotated tag points at
it. Every retired branch satisfies one of these two conditions.
| Retired branch | Commit | Reachable via | Content status |
|---|---|---|---|
feat/wallbreaker-freq-split | 3d372ef | tag experiment/wallbreaker-freq-split | Unique code, never merged |
feat/wallbreaker-substring-opt | 4a62248 | tag experiment/wallbreaker-substring-opt | Unique docs, superseded |
perf/phonetic-normalized-opt-5 | 9a57cca | tag archive/phonetic-normalized-opt-5 | Patch-equivalent to master |
release-rebuild-v0.9.0 | 66de57c | tag archive/release-rebuild-v0.9.0 | Patch-equivalent to master |
feat/wallbreaker-benchmarks | 7543b75 | ancestor of master | Merged |
feat/wallbreaker-simd | 5acf623 | ancestor of master | Merged |
dylon/ci-sibling-fix | 1968089 | ancestor of master | Merged |
fix-nodup-definition | 198b525 | ancestor of master | Merged |
opt/baseline | 845a288 | ancestor of master | Merged |
opt/llev-h1-intern-class-names | 1ec7f71 | ancestor of master | Merged |
opt/llev-h2-named-class-lookup | 5080be7 | ancestor of master | Merged |
opt/llev-h4-smallvec-charclass | 3db880e | ancestor of master | Merged |
perf/phonetic-normalized-benchmarks | dd1438c | ancestor of master | Merged |
perf/phonetic-normalized-opt-1 | 62db70d | ancestor of master | Merged |
perf/phonetic-normalized-opt-2 | ea54db5 | ancestor of master | Merged |
perf/phonetic-normalized-opt-3 | 50ceebb | ancestor of master | Merged |
perf/phonetic-normalized-opt-4 | 3418d30 | ancestor of master | Merged |
perf/phonetic-normalized-opt-6-bktree | 316f874 | ancestor of master | Merged |
proof-multirule-axiom | 9fa2fc9 | ancestor of master | Merged |
Surviving branches: master and release (the latter carries one intentional
commit, 32c9613, disabling PathMap in Cargo.toml).
Only two retired branches held anything absent from master. Both are now
annotated tags whose messages record the full hypothesis, verdict, and
measurements; read them with git show <tag>.
experiment/wallbreaker-freq-splitThe sole surviving copy of FrequencyPatternSplitter and
FrequencyWallBreaker (≈786 insertions across
src/wallbreaker/pattern_splitter.rs, src/wallbreaker/mod.rs, and
benches/wallbreaker_benchmarks.rs). It is the evidence behind Experiment 3
of the WallBreaker scientific ledger, whose
verdict was ❌ REJECTED (Overall regression).
git show experiment/wallbreaker-freq-split # hypothesis, results, analysis
git show experiment/wallbreaker-freq-split:src/wallbreaker/pattern_splitter.rs
git log --oneline master..experiment/wallbreaker-freq-split # the two unmerged commits
This code must not be deleted: it is the reproducible basis of a documented
negative result. It forked 170 commits behind master and predates the
migration of the dictionary module to the libdictenstein crate (cd82727),
so any attempt to rebase it onto master will conflict in src/dictionary/.
experiment/wallbreaker-substring-optDocuments the Phase 2 suffix-link substring-search optimization as rejected on
the grounds that the then-current SCDAWG was a prefix DAWG rather than a true
suffix automaton. This conclusion was subsequently overturned: master
implemented precisely that conversion in Experiment 5, "True SCDAWG
Implementation (𝒪(∣pattern∣) Substring Search)". The tag is retained as the
historical record of the intermediate negative result.
Its source edits touch src/dictionary/scdawg.rs and
src/dictionary/scdawg_char.rs, neither of which exists on master any longer
(migrated to libdictenstein in cd82727). They must not be cherry-picked.
Tags are immutable pointers; recreating a working branch from one is a single command.
git switch -c feat/wallbreaker-freq-split experiment/wallbreaker-freq-split
To confirm nothing has been orphaned:
git fsck --unreachable --no-reflogs | grep commit # expect no output
Can you improve this documentation?Edit on GitHub
cljdoc builds & hosts documentation for Clojure/Script libraries
| Ctrl+k | Jump to recent docs |
| ← | Move to previous article |
| → | Move to next article |
| Ctrl+/ | Jump to the search field |