Release v3.10.0

Commits

• principled headedness (phase-count) + foot delineation #175 (Ryan Heuser)
• 723b52a: Foot over SYLLABLES, not positions — recover traditional inversions (Ryan Heuser) #175
• 23dbf6a: Inversion map: separate trochaic inversions from promotions by violation type (Ryan Heuser) #175
• d7e7f12: Add caesura test: are mid-line inversions after a punctuation phrase break? (Ryan Heuser) #175
• 19c960d: Foot-parse raw scansion -> footed string (DP, mixed size + headedness) (Ryan Heuser) #175
• 787697b: Wire DP foot-parser into Parse; one head per foot (no pyrrhic), word tiebreak (Ryan Heuser) #175
• 57a33b4: Add foot_examples.py: one sonnet line per foot type from the DP (Ryan Heuser) #175
• 6a863d4: line.metrical_parse: select the most regular co-optimal parse (poem-level) (Ryan Heuser) #175
• derive from co-optimal sets, robust to best_parse tie-break #175 (Ryan Heuser)
• line-local fewest-unique-feet tie-break (drop poem meter + prior) #175 (Ryan Heuser)
• ignore bare (catalectic/extrametrical) feet in the tie-break #175 (Ryan Heuser)
• 8586bb4: Break all co-optimal parse ties by metrical regularity, consistently (Ryan Heuser) #175
• df057ac: Add foot-based tertiary tie-break for residual (score+regularity) ties (Ryan Heuser) #175
• b777923: Set _meter_vals in init for entity/DF parity in the tie-break (Ryan Heuser) #175
• 355b1b3: Foot-parser: catalexis (cheap line-final bare foot) + period-based foot size (Ryan Heuser) #175
• 2916619: Decouple best_parse from the foot-parser: drop the foot-based tertiary tie-break (Ryan Heuser) #175
• 0adf840: Add pseudo-foot tertiary sort key: deterministic, foot-flavored, decoupled (Ryan Heuser) #175
• 5986afa: Document foot parsing: methods doc + roadmap entry (Ryan Heuser) #175
• add fewest-ss / fewest-ww tie-break keys #175 (Ryan Heuser)
• make fewest-ss the primary tie-break (after score) #175 (Ryan Heuser)
• scansion-content final key (w-onset), fully determinative #175 (Ryan Heuser)
• 59a6f1b: Retire the MIDWORD foot-boundary tiebreak (word-boundary footing is a null result) (Ryan Heuser) #175
• ce458f0: Extrametrical foot edges (anacrusis / feminine ending): 93.3% -> 97.5% vs gold (Ryan Heuser) #175
• poem-level footing is a deliberate non-goal (Prosodic is line-scoped) #175 (Ryan Heuser)
• 1534a98: Rewire legacy foot properties onto the DP delineation; metrical_parse = best_parse (Ryan Heuser) #175
• e448fa6: Add Parse.scansion (w/s) + .footed_scansion; free the name by renaming the internal list (Ryan Heuser) #175
• 6f0465a: Promote Foot to a first-class view class + FootList (Option B) (Ryan Heuser) #175
• a4bac08: Mixed-N pooling: keep dominated-length parses as BOUNDED (don't drop them) (Ryan Heuser) #175
• d9e7bb5: Line View: parse single lines via the DF path (shows dominated cross-length parses) (Ryan Heuser) #175
• plan to retire the duplicate entity parse path (unify on DF) #175 (Ryan Heuser)
• parse-path unification plan + spike finding (Parse.concat is the coupling) #176 (Ryan Heuser)
• de8054c: Parse.concat DF-safe + line-level parsing on the DF path (retire-entity-parser) (Ryan Heuser) #176
• e939c50: Audit doc: mark Parse.concat + line-level DF migration done; note bare-WordTokenList remnant (Ryan Heuser) #176
• 439407c: Web linepart parsing on the DF path (parse_units_from_df); no parse_batch in api.py (Ryan Heuser) #176
• 23b22f6: Delete the entity parse path — one parser (DF) remains (Ryan Heuser) #176
• cb8e0af: ParseSlot.wordform DF-safe (None off the DF path); doc: Phase 1 complete, Phase 2 reframed (Ryan Heuser) #176
• mark the entity-parser retirement shipped (PR #176) #176 (Ryan Heuser)
• 4b233e2: Address PR #176 review: fix num_words inflation + Case-2 scoping bug + docs (Ryan Heuser) #176
• 42b27a7: Address PR #177 review: best_parse perf + foot-property edges + docs (Ryan Heuser) #177
• 00528ff: Bump version 3.8.0 -> 3.9.0 (foot delineation, retire entity parser, best_parse determinism) (Ryan Heuser) #177
• b513015: Dominated-length reference readings show their own best pronunciation (Ryan Heuser) #178
• b91698b: Bump version 3.9.0 -> 3.9.1 (dominated-length reference readings show best pronunciation) (Ryan Heuser) #179
• friendlier annotation loading + by-meter foot-gold study #180 (Ryan Heuser)
• b9df211: MaxEnt trains on mixed-N (ragged) lines — recovers all elision lines (Ryan Heuser) #181
• 71437fa: Vectorize the pooling overlay: index-aligned argmin in build_for_N (parse -29%) (Ryan Heuser) #182
• 42ae941: Elite bounding pre-screen on the torch device (MPS/CUDA): parse GPU 5.2s -> 4.1s (Ryan Heuser) #183
• 48dfca4: Profiling table: add measured v1.3.8 reference column (Ryan Heuser) #183
• ed626d2: Track foot-gold-sample.csv (raw annotation export behind foot-gold.csv) (Ryan Heuser) #186
• a4e4d7e: Post-sprint audit sweep: dead code, docs corrections, CI blind-spot test (Ryan Heuser) #184
• 8787ec2: Bump version 3.9.1 -> 3.10.0 (Ryan Heuser) #184
• the foot layer is now a first-class citizen of every how-to surface #185 (Ryan Heuser)
• 9161780: Docs beef-up: rhyme write-up, architecture correction, richer webapp + README (Ryan Heuser) #185
• The grammar of each meter (executable by-meter MaxEnt study) #185 (Ryan Heuser)
• 44a3481: Release-review fixes for 3.10.0 (from /review 186: 8 angles, 6 confirmed defects) (Ryan Heuser) #187