2c8367109c
Implementeaza app/import_parse.py (modul nou) cu toate cele 4 taskuri parser: T14 (task #7): schelet parser + arhitectura 2-treceri (Issue 2 consens cross-model) - Trecerea 1 read_only=True: dim-check FileTooLarge (>5000 randuri / >5MB) + detectie multi-sheet - Trecerea 2 normal-mode: header + merged cells + body - CSV delimiter sniff (csv.Sniffer + proba explicita {; , \t}) — export RO foloseste ; - Encoding: UTF-8-sig -> UTF-8 -> cp1250 -> latin2 (fallback RO) - Exceptii custom: FileTooLarge, HeaderError, MultipleSheets - Coloane duplicate din merge: sufixate _2/_3 (nu HeaderError) T3 (task #8): coercion guard + needs_review + mesaj formule-None - VIN numeric (openpyxl citeste ca int/float) -> coercion_flags[row] needs_review - Odometru float 123456.0 -> tunde .0 inainte de validare (§3.4bis, ordonare critica) - Rata None > 60% pe coloana obligatorie -> formula_columns (Issue 3 mesaj specific) - Datetime nativ -> convertit la YYYY-MM-DD (neambiguu) T10 (task #9): dezambiguizare data la nivel de COLOANA (OV-8) - Datetime nativ -> "native" (direct, fara ambiguitate) - String: detectie format din intreaga coloana — daca oricare rand are token[1]>12 -> DD-first - Daca toti zi<=12 -> "ambiguous" -> needs_review per rand - parse_date_value() helper pentru preview resolve T13 (task #10): robustete export RO - Multi-sheet: >1 sheet non-gol -> MultipleSheets([...]) cu lista; sheet ales -> parse normal - Merged header: _unmerge_header propaga valoarea topleft la toate coloanele din grup - Footer trim: randuri trailing cu VIN + data ambele goale -> skip structural (nu needs_data) Teste: 37 teste verzi in tests/test_import_parse.py (fixture-uri xlsx generate in-memory). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>