The Map CSV migrated from the legacy ERP carries names with two common
defects: missing French accents (Stephane, Andre, Frederic), and
compound first names that were typed without a separator (Marcandre,
Mariejosee, Jeanphilippe). Sending an email "Bonjour stephane," instead
of "Bonjour Stéphane," reads as sloppy automation. Fix both at parse
time so the user sees the corrected names in Step 2 and can override
inline if the auto-cleaner got it wrong.
Backend (lib/campaigns.js):
- FR_NAME_FIXES — 100+ entry dictionary mapping lowercase no-accent
Québec first names to their canonical accented form (André, Stéphane,
Frédéric, Geneviève, Hélène, Joséée, etc.). Sourced from MIQ baby
names + older-generation curation.
- COMPOUND_PARTS — list of common name parts (jean, marie, anne, marc,
philippe, françois, etc.) that combine into QC compound first names.
When two parts appear concatenated with no separator, the cleaner
splits and hyphenates them. Example: "Marcandre" → ["marc","andre"]
→ "Marc-André" (dictionary then applies accent).
- titleCaseToken — proper Title Case respecting apostrophes (O'Brien,
L'Heureux) and hyphens (Marie-Ève). Uses \p{L} Unicode class so it
works on accented chars correctly.
- cleanName(raw) — full pipeline: trim → Title Case → dictionary
lookup per word → compound split fallback. Applied to firstname AND
lastname in parseMapCsv.
- nameWarning(name) — heuristic flag for cases the cleaner couldn't
confidently handle: digit in name, single letter, abnormally long
without separator (likely two stuck names not in COMPOUND_PARTS).
Returns a short FR description for the UI tooltip.
- parseMapCsv now returns firstname/lastname (cleaned) + firstname_raw/
lastname_raw (original from CSV) + cleaned_changed bool + name_warnings
per recipient. UI uses these to show before/after + flags.
UI (CampaignNewPage Step 2):
- New counter card "Noms à vérifier" — count of recipients with at least
one nameWarning. Only renders if > 0.
- Info banner above the recipients table:
"X nom(s) auto-corrigés (...) Y nom(s) suspects (...)"
- Per-row icons in the firstname + lastname columns:
• ⚠ amber WARNING — cleaner flagged this name as suspicious
(tooltip shows the reason: "deux prénoms collés", "contient un
chiffre", etc.)
• ✨ green AUTO_FIX_HIGH — auto-cleaner changed something at parse
time (tooltip shows the original raw value)
Both icons are tooltip-only — no action required.
- Click any name cell → q-popup-edit opens an inline input. Type the
correction, Enter saves. ESC cancels. This is the manual override
path for any name the auto-cleaner mishandled.
Tests (manual via end-to-end smoke against prod):
STEPHANE TREMBLAY → Stéphane Tremblay ✓ accent + Title Case
marie tremblay → Marie Tremblay ✓ Title Case only
Marcandre Boileau → Marc-André Boileau ✓ compound + accent
Jean Francois Lebrun → Jean François Lebrun ✓ accent only
Mariejosee Lapierre → Marie-Josée Lapierre ✓ compound + double accent
Andre LAPRISE → André Laprise ✓ both fixed
Helene St-Pierre → Hélène St-Pierre ✓ accent, hyphen preserved
Frederic O'Brien → Frédéric O'Brien ✓ accent, apostrophe preserved
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
d897bcedb4