v1.5.0

New

• Guard-aware normalization — in :abstract mode, all calls inside
  when guard clauses are abstracted so that functions differing only in
  guard predicates are detected as clones. Covers Kernel guards, Erlang
  BIFs, defguard macros, and library guards like Integer.is_even/1.
• Boolean operator canonicalization — &&/||/! are rewritten to
  and/or/not so stylistic choice between short-circuit and keyword
  operators doesn’t prevent clone matching.
• Sigil ~w expansion — ~w(foo bar)a is expanded to [:foo, :bar]
  so sigil word-lists match their literal equivalents.
• MinHash-accelerated fuzzy detection — large posting lists (>50
  entries) now use MinHash signatures for O(k) approximate Jaccard instead
  of O(|A|+|B|) exact set operations. Removes the hard posting-list cap,
  improving recall for large monorepos without sacrificing precision.
• HTML report syntax highlighting via Makeup — proper Elixir
  tokenization with dark/light theme support, replacing the regex-based
  highlighter.
• Configurable detection tuning — previously hardcoded constants are
  now available as config options and CLI flags:
  • max_window_size (--max-window-size, default: 4) — max consecutive
    sibling functions combined into a single fingerprint for cross-module
    clone detection.
  • mass_tolerance (--mass-tolerance, default: 0.3) — max relative size
    difference for Type-III comparison.

Changed

• ignored_attributes default — derived from
  Module.reserved_attributes/0 instead of a hardcoded list. Picks up
  5 previously missing attributes and stays current with future Elixir
  versions automatically.

Performance

• Fused normalizer — metadata stripping, boolean canonicalization, sigil
  expansion, pipe normalization, and variable renaming run in a single AST
  walk instead of 4 separate traversals. Ash (572 files) ~14% faster.

Benchmarked on real-world projects with full Type-I/II/III detection
(literal_mode: :abstract, min_similarity: 0.85, normalize_pipes: true):

Project	Files	Clones	Time
Broadway	22	1	45ms
Nx	42	12	674ms
Nerves	50	2	172ms
Ecto	56	19	525ms
Commanded	63	8	147ms
Oban	66	16	193ms
Phoenix	74	14	607ms
Elixir stdlib	105	84	1.6s
Surface	109	31	513ms
Absinthe	263	63	590ms
Livebook	265	62	2.1s
Plausible	465	80	2.4s
Ash	572	535	5.8s