[m-hu]

Closes #1 — ports the clojure-only
src/string_related/suffix_tree/string_function_calculation.clj (audited
line by line: build-tree/follow-link/tree-value) to
src/string_related/suffix_tree/string_function_calculation.py (140 lines).

Algorithm

Ukkonen online suffix tree over input + terminal $; answer = max over
nodes of string-depth × leaf-count. O(n) time / O(n) space, same as the
.clj.

Validation

• Difftest vs brute-force oracle: new port 143/143 (edges + 120 seeded
  fuzz + structured), old suffix_array twin 143/143; 5/5 IO edge cases.
• JVM cross-check of the .clj: 8/8 values byte-identical (incl.
  mississippi→11, aabaaabaaab→14); blank/malformed stdout+stderr
  byte-identical.
• Scaling: identical outputs everywhere; ~1.3–2.1x faster than the old
  twin on adversarial inputs.
• run_tests.py --only suffix_tree/string_function_calculation: 8 passed,
  0 failed locally.

Note for the merger

The 8 hand-built test cases could not be committed: tests/resources/ is
gitignored and the encrypted bundle needs RESOURCES_PASSPHRASE. A
passphrase holder must add cases to resource/shared and run
tools/build_bundle.sh before CI perf-callgrind picks up the benchmark.

AI-Tool: Muse Code (Meta Muse Spark)

[michael-h-hu]

Approve — round 1

Inspected the full diff (1 new file, 140 lines, no other files touched) against the linked issue #1 and the in-repo suffix_array twin.

What was checked:

• Read src/string_related/suffix_tree/string_function_calculation.py end to end: Ukkonen online construction over input + $ terminal, answer = max over nodes of string-depth x leaf-count, explicit stacks (no recursion), ans = n init and empty-input guard matching the twin.
• Ran an independent 137-case difftest of string_value vs a brute-force oracle (edges, mississippi, aabaaabaaab, 120 seeded fuzz over small alphabets, adversarial repeats): 0 mismatches.
• CLI: mississippi -> 11, blank input -> silent exit 0, non-lowercase -> stderr error with empty stdout (stderr is ignored by the run_tests.py stdout-only comparison, so this is safe; it mirrors the .clj malformed-input behavior cited in the PR body).
• CI: test + all 4 test shards green; all 4 perf-callgrind shards green after a retry of an unrelated environmental wall-clock timeout (dp/edit_distance 45B-Ir case, pre-existing knife-edge, passed on rerun with no code change); report green. New benchmark measured 11/11 cases in perf with 0 failures.
• Issue #1 acceptance: same algorithm/complexity as the .clj (Ukkonen O(n)/O(n)); shared-layout cases resolve from the bundle; run_tests.py green; benchmark present in the perf sweep.

Residual notes (non-blocking):

• The PR body's "cases could not be committed" caveat is stale: the shared-layout suffix_tree cases already exist in the encrypted bundle (same resource path as the clojure side) and CI measured 11 of them. No passphrase-holder action needed.
• .clj line-by-line fidelity relies on the crew's audit plus the behavioral evidence above; I did not re-read the .clj source in this pass.

No findings. Merging is the author's call.

AI-Tool: Muse Code (Meta Muse Spark)