letsbe/pn-new-crm
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

feat(berths): per-berth PDF storage (versioned) + reverse parser

Browse Source 
Phase 6b of the berth-recommender refactor (see
docs/berth-recommender-and-pdf-plan.md §3.2, §3.3, §4.7b, §11.1, §14.6).
Builds on the Phase 6a pluggable storage backend (commit 83693dd) — every
file write goes through `getStorageBackend()`; no direct minio imports.

Schema (migration 0030_berth_pdf_versions):
  - new table `berth_pdf_versions` with monotonic `version_number` per
    berth, `storage_key` (renamed convention from §4.7a), sha256, size,
    `download_url_expires_at` cache slot for §11.1 signed-URL throttling,
    and `parse_results` jsonb for the audit trail.
  - new column `berths.current_pdf_version_id` (deferred from Phase 0)
    with FK to `berth_pdf_versions(id)` ON DELETE SET NULL.
  - relations + types exported from `schema/berths.ts`.

3-tier reverse parser (`lib/services/berth-pdf-parser.ts`):
  1. AcroForm via pdf-lib — pulls named fields (`length_ft`,
     `mooring_number`, etc.) at confidence 1. Sample PDF has 0 such
     fields, so this is defensive coverage for future templates.
  2. OCR via Tesseract.js — positional/regex heuristics keyed off the
     §9.2 layout (Length/Width/Water Depth as `<imperial> / <metric>`,
     `WEEK HIGH / LOW`, `CONFIRMED THROUGH UNTIL <date>`, etc.). Returns
     per-field confidence + global mean; flags imperial-vs-metric drift
     >1% in `warnings`.
  3. AI fallback — gated via `getResolvedOcrConfig()` (existing
     openai/claude provider). Surfaced from the diff dialog only when
     `shouldOfferAiTier()` returns true (mean OCR confidence below
     0.55 threshold), so OPENAI_API_KEY isn't burned on every upload.

Service layer (`lib/services/berth-pdf.service.ts`):
  - `uploadBerthPdf()` — magic-byte check, size cap, version-number
    bump + current pointer in one transaction.
  - `reconcilePdfWithBerth()` — auto-applies fields where CRM is null;
    flags conflicts when CRM and PDF disagree; tolerates ±1% on numeric
    columns; warns on mooring-number-in-PDF mismatch (§14.6).
  - `applyParseResults()` — hard allowlist of writable columns;
    stamps `appliedFields` onto `parse_results` for audit.
  - `rollbackToVersion()` — pointer flip only, never re-parses (§14.6).
  - `listBerthPdfVersions()` — version list with 15-min signed URLs.
  - `getMaxUploadMb()` — port-override → global → default 15 lookup
    on `system_settings.berth_pdf_max_upload_mb`.

§14.6 critical mitigations:
  - Magic-byte check (`%PDF-`) on every upload; mismatch deletes the
    storage object and rejects the request.
  - Size cap from `system_settings.berth_pdf_max_upload_mb` (default
    15 MB); enforced in the upload-url presign AND server-side.
  - 0-byte uploads rejected.
  - Mooring-number mismatch surfaces as a `warnings[]` entry on the
    reconcile result so the rep sees it in the diff dialog.
  - Imperial vs metric ±1% tolerance in both the parser warnings and
    the reconcile equality check.
  - Path traversal already blocked at the storage layer (Phase 6a).

API + UI:
  - `POST /api/v1/berths/[id]/pdf-upload-url` — presigned URL (S3) or
    HMAC-signed proxy URL (filesystem) sized to the per-port cap.
  - `POST /api/v1/berths/[id]/pdf-versions` — verifies the upload via
    `backend.head()`, writes the row, bumps `current_pdf_version_id`.
  - `GET /api/v1/berths/[id]/pdf-versions` — version list + signed URLs.
  - `POST /api/v1/berths/[id]/pdf-versions/[versionId]/rollback`.
  - `POST /api/v1/berths/[id]/pdf-versions/parse-results/apply` —
    rep-confirmed diff payload.
  - New "Documents" tab on the berth detail page (`berth-tabs.tsx`)
    with current-PDF panel, version history, Replace PDF button, and
    `<PdfReconcileDialog>` for the auto-applied + conflicts UX.

System settings:
  - `berth_pdf_max_upload_mb` (default 15) — caps presigned-upload size
    + server-side validation. Resolved port-override → global → default.

Tests:
  - `tests/unit/services/berth-pdf-parser.test.ts` — magic bytes,
    feet-inches, human dates, full §9.2-shaped OCR text → 18 fields,
    drift warning, AI-tier gate.
  - `tests/unit/services/berth-pdf-acroform.test.ts` — synthetic
    pdf-lib AcroForm round-trip.
  - `tests/integration/berth-pdf-versions.test.ts` — upload, version-
    number bump, magic-byte rejection, reconcile auto-applied vs
    conflicts vs ±1% tolerance, mooring-number warning,
    applyParseResults allowlist enforcement, rollback semantics.

Acceptance: `pnpm exec tsc --noEmit` clean, `pnpm exec vitest run`
green at 1103/1103.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Matt Ciaccio
2026-05-05 03:34:24 +02:00
parent 83693dd993
commit 249ffe3e4a
22 changed files with 13349 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

499
src/lib/services/berth-pdf-parser.ts Normal file
View File

@@ -0,0 +1,499 @@
/**
* Reverse parser for per-berth PDFs (Phase 6b — see plan §4.7b and §9.2).
*
* Three tiers, each falling back to the next:
*
* 1. AcroForm — read named text fields via pdf-lib. The sample
* `Berth_Spec_Sheet_A1.pdf` has 0 AcroForm fields (designers export the
* PDF flat), so this tier is built defensively for future templates that
* may include named form fields. When fields exist, this is the highest-
* confidence path because there's no OCR loss.
*
* 2. OCR — Tesseract.js extracts text from the page; positional/regex
* heuristics keyed off the labels documented in §9.2 pull out values.
* Returns per-field confidence scores.
*
* 3. AI fallback — gated on `getResolvedOcrConfig(...)` returning a usable
* OpenAI/Claude config. Only invoked when OCR confidence is below
* threshold for too many fields AND the rep opts in via the diff dialog.
* A null `apiKey` causes this tier to return a clear "not configured"
* error rather than silently falling back to OCR-only.
*/
import { PDFDocument } from 'pdf-lib';
// ─── shared types ────────────────────────────────────────────────────────────
export type ParserEngine = 'acroform' | 'ocr' | 'ai';
/**
* Canonical extracted shape. Keys map 1:1 to nullable columns on the `berths`
* table; `mooringNumber` is special (used for the §14.6 mismatch warning).
*/
export interface ExtractedBerthFields {
mooringNumber?: string | null;
lengthFt?: number | null;
lengthM?: number | null;
widthFt?: number | null;
widthM?: number | null;
/** Water depth at the berth (separate from a vessel's max draft). */
waterDepth?: number | null;
waterDepthM?: number | null;
/** Max draught of vessel — falls back to the berth's draft column. */
draftFt?: number | null;
draftM?: number | null;
bowFacing?: string | null;
sidePontoon?: string | null;
powerCapacity?: number | null;
voltage?: number | null;
mooringType?: string | null;
cleatType?: string | null;
cleatCapacity?: string | null;
bollardType?: string | null;
bollardCapacity?: string | null;
access?: string | null;
weeklyRateHighUsd?: number | null;
weeklyRateLowUsd?: number | null;
dailyRateHighUsd?: number | null;
dailyRateLowUsd?: number | null;
/** ISO date YYYY-MM-DD. */
pricingValidUntil?: string | null;
price?: number | null;
}
export interface ParsedField<T = unknown> {
value: T;
/** 0..1 confidence; 1 means "absolute match" (AcroForm or unambiguous regex). */
confidence: number;
/** Engine that produced this field; helps the diff dialog explain itself. */
engine: ParserEngine;
}
export interface ParseResult {
engine: ParserEngine;
/** Sparse — only fields the parser was able to extract. */
fields: Partial<Record<keyof ExtractedBerthFields, ParsedField>>;
/** Mean confidence across all extracted fields (0..1). */
meanConfidence: number;
/** Raw text the OCR or AI tier produced — useful for the diff dialog audit. */
rawText?: string;
/** Set when a tier degraded; the API surface uses this to decide whether to
* surface the "AI parse" button. */
warnings: string[];
}
// ─── magic-byte check (§14.6 critical) ───────────────────────────────────────
/** Reads first 5 bytes; returns true iff they are `%PDF-`. */
export function isPdfMagic(buffer: Buffer): boolean {
if (buffer.length < 5) return false;
return (
buffer[0] === 0x25 && // %
buffer[1] === 0x50 && // P
buffer[2] === 0x44 && // D
buffer[3] === 0x46 && // F
buffer[4] === 0x2d // -
);
}
// ─── tier 1: AcroForm ────────────────────────────────────────────────────────
/**
* AcroForm field name → ExtractedBerthFields key. Mirrors the names §4.7b
* mentions ("length_ft", "mooring_number"…) plus a couple of tolerant aliases.
*/
const ACROFORM_FIELD_MAP: Record<string, keyof ExtractedBerthFields> = {
mooring_number: 'mooringNumber',
berth_number: 'mooringNumber',
length_ft: 'lengthFt',
length_m: 'lengthM',
width_ft: 'widthFt',
width_m: 'widthM',
draft_ft: 'draftFt',
draft_m: 'draftM',
water_depth: 'waterDepth',
water_depth_m: 'waterDepthM',
bow_facing: 'bowFacing',
side_pontoon: 'sidePontoon',
pontoon: 'sidePontoon',
power_capacity: 'powerCapacity',
voltage: 'voltage',
mooring_type: 'mooringType',
cleat_type: 'cleatType',
cleat_capacity: 'cleatCapacity',
bollard_type: 'bollardType',
bollard_capacity: 'bollardCapacity',
access: 'access',
weekly_rate_high_usd: 'weeklyRateHighUsd',
weekly_rate_low_usd: 'weeklyRateLowUsd',
daily_rate_high_usd: 'dailyRateHighUsd',
daily_rate_low_usd: 'dailyRateLowUsd',
pricing_valid_until: 'pricingValidUntil',
price: 'price',
};
async function tryAcroForm(buffer: Buffer): Promise<ParseResult | null> {
let doc: PDFDocument;
try {
doc = await PDFDocument.load(buffer, { ignoreEncryption: true });
} catch {
return null;
}
let form: ReturnType<PDFDocument['getForm']>;
try {
form = doc.getForm();
} catch {
return null;
}
const fields = form.getFields();
if (fields.length === 0) return null;
const out: Partial<Record<keyof ExtractedBerthFields, ParsedField>> = {};
for (const field of fields) {
const name = field.getName().toLowerCase();
const target = ACROFORM_FIELD_MAP[name];
if (!target) continue;
// pdf-lib doesn't expose a generic "get value" — narrow to text fields.
let raw: string | undefined;
try {
const tf = form.getTextField(field.getName());
raw = tf.getText() ?? undefined;
} catch {
continue;
}
if (!raw || raw.trim().length === 0) continue;
const parsed = coerceFieldValue(target, raw.trim());
if (parsed === null) continue;
out[target] = { value: parsed, confidence: 1, engine: 'acroform' };
}
if (Object.keys(out).length === 0) return null;
return {
engine: 'acroform',
fields: out,
meanConfidence: 1,
warnings: [],
};
}
// ─── tier 2: OCR via Tesseract ───────────────────────────────────────────────
/**
* Runs Tesseract against a PDF rasterized to one image per page. Tesseract.js
* accepts image inputs; we use a lazy `pdfjs-dist`-style rasterization fallback
* via dynamic import. To keep the parser unit-testable without a WASM bundle,
* the actual recognize() call is encapsulated in the `runOcr` adapter that
* production wires to tesseract.js and tests can stub.
*/
export interface OcrAdapter {
/** Returns plain text + a 0..100 mean confidence score. */
recognize(buffer: Buffer): Promise<{ text: string; confidence: number }>;
}
/** Default adapter — dynamically imports tesseract.js so the WASM bundle isn't
* pulled into client builds. */
async function defaultOcrAdapter(): Promise<OcrAdapter> {
return {
recognize: async (buffer: Buffer) => {
const tesseract = await import('tesseract.js');
// Tesseract handles PDF inputs by rasterizing the first page; for our
// single-page spec sheets that's sufficient.
const result = await tesseract.recognize(buffer, 'eng');
return {
text: result.data.text ?? '',
confidence: typeof result.data.confidence === 'number' ? result.data.confidence : 0,
};
},
};
}
/**
* Heuristic extraction from OCR text. The patterns mirror the layout
* documented in plan §9.2:
*
* - "Length: 206' 8" / 63m"
* - "Mooring: A12" or large "A1" near "BERTH NUMBER"
* - "WEEK HIGH / LOW" and "DAY HIGH / LOW" pricing blocks
* - "ALL PRICES ABOVE ARE CONFIRMED THROUGH UNTIL <date>"
*/
export function extractFromOcrText(rawText: string): {
fields: Partial<Record<keyof ExtractedBerthFields, ParsedField>>;
warnings: string[];
} {
const warnings: string[] = [];
const out: Partial<Record<keyof ExtractedBerthFields, ParsedField>> = {};
// Normalize whitespace for line-based regexes but keep structure.
const text = rawText.replace(/ /g, ' ');
// Mooring number: BERTH NUMBER block. We try a couple of layouts.
const mooringMatch =
text.match(/BERTH\s+NUMBER[\s\S]{0,80}?\b([A-Z]\d{1,3})\b/i) ??
text.match(/^\s*([A-Z]\d{1,3})\s*$/m) ??
text.match(/Mooring(?:\s+Number)?\s*[:#]?\s*([A-Z]\d{1,3})/i);
if (mooringMatch) {
out.mooringNumber = { value: mooringMatch[1]!.toUpperCase(), confidence: 0.85, engine: 'ocr' };
}
// Length / Width / Water Depth — `Label: <imperial> / <metric>` form.
// Imperial may be `206' 8"` style; we capture the numeric prefix in feet
// and parse the metric independently because they're rarely lossless.
const dimensional = (
label: string,
ftKey: keyof ExtractedBerthFields,
mKey: keyof ExtractedBerthFields,
) => {
const re = new RegExp(
`${label}\\s*[:.]?\\s*([0-9]+(?:'\\s*[0-9]+\")?(?:\\.[0-9]+)?)\\s*(?:ft)?\\s*\\/\\s*([0-9]+(?:\\.[0-9]+)?)\\s*m`,
'i',
);
const m = text.match(re);
if (!m) return;
const ft = parseFeetInches(m[1]!);
const meters = Number(m[2]);
if (ft != null && Number.isFinite(ft)) {
out[ftKey] = { value: ft, confidence: 0.8, engine: 'ocr' } as ParsedField;
}
if (Number.isFinite(meters)) {
out[mKey] = { value: meters, confidence: 0.85, engine: 'ocr' } as ParsedField;
}
if (ft != null && Number.isFinite(meters) && Math.abs(ft * 0.3048 - meters) / meters > 0.01) {
warnings.push(
`${label}: imperial/metric mismatch — ${ft}ft vs ${meters}m differ >1% (using imperial as source of truth).`,
);
}
};
dimensional('Length', 'lengthFt', 'lengthM');
dimensional('Width', 'widthFt', 'widthM');
dimensional('Water\\s+Depth', 'waterDepth', 'waterDepthM');
// Max draft of vessel maps to the berth's draft column.
dimensional('Max\\.?\\s*draught(?:\\s+of\\s+vessel)?', 'draftFt', 'draftM');
// Singular labels (`Bow Facing: East`, `Pontoon: QUAY PT`).
const labelToKey: Array<[RegExp, keyof ExtractedBerthFields]> = [
[/Bow\s+Facing\s*[:.]?\s*([A-Za-z .]+?)(?:\n|$)/i, 'bowFacing'],
[/Pontoon\s*[:.]?\s*([A-Za-z0-9 .\-]+?)(?:\n|$)/i, 'sidePontoon'],
[/Mooring\s+Type\s*[:.]?\s*([A-Za-z0-9 \-\/]+?)(?:\n|$)/i, 'mooringType'],
[/Cleat\s+Type\s*[:.]?\s*([A-Za-z0-9 \-]+?)(?:\n|$)/i, 'cleatType'],
[/Cleat\s+Capacity\s*[:.]?\s*([A-Za-z0-9 \-]+?)(?:\n|$)/i, 'cleatCapacity'],
[/Bollard\s+Type\s*[:.]?\s*([A-Za-z0-9 \-]+?)(?:\n|$)/i, 'bollardType'],
[/Bollard\s+Capacity\s*[:.]?\s*([A-Za-z0-9 \-]+?)(?:\n|$)/i, 'bollardCapacity'],
[/Access\s*[:.]?\s*([A-Za-z0-9 .,()\-]+?)(?:\n|$)/i, 'access'],
];
for (const [re, key] of labelToKey) {
const m = text.match(re);
if (m && m[1]) {
out[key] = { value: m[1].trim(), confidence: 0.75, engine: 'ocr' } as ParsedField;
}
}
// Power Capacity (kW) and Voltage at 60Hz.
const powerMatch = text.match(/Power\s+Capacity\s*[:.]?\s*([0-9]+(?:\.[0-9]+)?)\s*kW/i);
if (powerMatch) {
out.powerCapacity = { value: Number(powerMatch[1]), confidence: 0.85, engine: 'ocr' };
}
const voltageMatch = text.match(/Voltage(?:\s+at\s+60\s*Hz)?\s*[:.]?\s*([0-9]+)\s*V/i);
if (voltageMatch) {
out.voltage = { value: Number(voltageMatch[1]), confidence: 0.85, engine: 'ocr' };
}
// Pricing: "WEEK HIGH / LOW: 11,341 USD / 8,100 USD"
const weekMatch = text.match(
/WEEK\s+HIGH\s*\/\s*LOW[:.\s]*([0-9,]+)\s*USD\s*\/\s*([0-9,]+)\s*USD/i,
);
if (weekMatch) {
out.weeklyRateHighUsd = {
value: Number(weekMatch[1]!.replace(/,/g, '')),
confidence: 0.8,
engine: 'ocr',
};
out.weeklyRateLowUsd = {
value: Number(weekMatch[2]!.replace(/,/g, '')),
confidence: 0.8,
engine: 'ocr',
};
}
const dayMatch = text.match(
/DAY\s+HIGH\s*\/\s*LOW[:.\s]*([0-9,]+)\s*USD\s*\/\s*([0-9,]+)\s*USD/i,
);
if (dayMatch) {
out.dailyRateHighUsd = {
value: Number(dayMatch[1]!.replace(/,/g, '')),
confidence: 0.8,
engine: 'ocr',
};
out.dailyRateLowUsd = {
value: Number(dayMatch[2]!.replace(/,/g, '')),
confidence: 0.8,
engine: 'ocr',
};
}
// Purchase price: "PURCHASE PRICE:\nFEE SIMPLE OR STRATA LOT\n3,880,800 USD"
const priceMatch = text.match(/PURCHASE\s+PRICE[\s\S]{0,80}?([0-9][0-9,]+)\s*USD/i);
if (priceMatch) {
out.price = { value: Number(priceMatch[1]!.replace(/,/g, '')), confidence: 0.7, engine: 'ocr' };
}
// Pricing validity: "ALL PRICES ABOVE ARE CONFIRMED THROUGH UNTIL SEPTEMBER 15TH, 2025"
const validityMatch = text.match(
/CONFIRMED\s+THROUGH\s+UNTIL\s+([A-Za-z]+\s+[0-9]{1,2})(?:[A-Z]{2})?,?\s+([0-9]{4})/i,
);
if (validityMatch) {
const iso = parseHumanDate(`${validityMatch[1]} ${validityMatch[2]}`);
if (iso) {
out.pricingValidUntil = { value: iso, confidence: 0.75, engine: 'ocr' };
} else {
warnings.push(
'Could not normalize "CONFIRMED THROUGH UNTIL" date; pricing_valid_until skipped.',
);
}
}
return { fields: out, warnings };
}
async function tryOcr(buffer: Buffer, adapter?: OcrAdapter): Promise<ParseResult | null> {
const ocr = adapter ?? (await defaultOcrAdapter());
const result = await ocr.recognize(buffer);
if (!result.text || result.text.length === 0) {
return {
engine: 'ocr',
fields: {},
meanConfidence: 0,
rawText: '',
warnings: ['OCR produced no text.'],
};
}
const { fields, warnings } = extractFromOcrText(result.text);
// Tesseract gives 0..100; normalize to 0..1 and use it as a global floor —
// per-field confidence is set by the regex tier above.
const floor = Math.max(0, Math.min(result.confidence, 100)) / 100;
for (const key of Object.keys(fields) as Array<keyof ExtractedBerthFields>) {
const f = fields[key];
if (f) f.confidence = Math.min(f.confidence, Math.max(floor, 0.5));
}
const values = Object.values(fields);
const meanConfidence =
values.length === 0
? 0
: values.reduce((sum, v) => sum + (v?.confidence ?? 0), 0) / values.length;
return {
engine: 'ocr',
fields,
meanConfidence,
rawText: result.text,
warnings,
};
}
// ─── tier 3: AI fallback ─────────────────────────────────────────────────────
/** Confidence floor below which we recommend the AI tier in the diff dialog. */
export const OCR_LOW_CONFIDENCE_THRESHOLD = 0.55;
/** True when the rep should be offered an "AI parse" button. */
export function shouldOfferAiTier(parse: ParseResult): boolean {
if (parse.engine !== 'ocr') return false;
if (Object.keys(parse.fields).length === 0) return true;
return parse.meanConfidence < OCR_LOW_CONFIDENCE_THRESHOLD;
}
// ─── public entry point ──────────────────────────────────────────────────────
export interface ParseBerthPdfOptions {
/** Override Tesseract for testing. Production flows resolve the default. */
ocrAdapter?: OcrAdapter;
/** Skip the OCR tier when only AcroForm is wanted (e.g. unit tests). */
skipOcr?: boolean;
}
/**
* Parse a per-berth PDF buffer. Each tier falls back to the next; the
* returned result's `engine` field tells callers which tier produced the
* fields (used by the reconcile-diff dialog to colour confidence chips).
*
* The AI tier is never invoked from this entry point — that's a separate
* deliberate action triggered from the diff dialog so OPENAI_API_KEY isn't
* spent on every upload.
*/
export async function parseBerthPdf(
buffer: Buffer,
opts: ParseBerthPdfOptions = {},
): Promise<ParseResult> {
if (!isPdfMagic(buffer)) {
throw new Error('PDF magic-byte check failed: file does not begin with %PDF-');
}
const acro = await tryAcroForm(buffer);
if (acro && Object.keys(acro.fields).length > 0) return acro;
if (opts.skipOcr) {
return {
engine: 'ocr',
fields: {},
meanConfidence: 0,
warnings: ['skipOcr=true; no AcroForm fields found.'],
};
}
const ocr = await tryOcr(buffer, opts.ocrAdapter);
return (
ocr ?? {
engine: 'ocr',
fields: {},
meanConfidence: 0,
warnings: ['OCR adapter returned null.'],
}
);
}
// ─── helpers ─────────────────────────────────────────────────────────────────
/** Coerce an AcroForm raw value to the right scalar for the target column. */
function coerceFieldValue(key: keyof ExtractedBerthFields, raw: string): string | number | null {
// String columns
const stringKeys: Array<keyof ExtractedBerthFields> = [
'mooringNumber',
'bowFacing',
'sidePontoon',
'mooringType',
'cleatType',
'cleatCapacity',
'bollardType',
'bollardCapacity',
'access',
'pricingValidUntil',
];
if (stringKeys.includes(key)) {
if (key === 'pricingValidUntil') {
// Accept ISO YYYY-MM-DD as-is; otherwise try a humane parse.
if (/^\d{4}-\d{2}-\d{2}$/.test(raw)) return raw;
return parseHumanDate(raw);
}
return raw;
}
// Numeric columns: strip currency / unit suffixes and commas.
const numeric = Number(raw.replace(/[^0-9.\-]/g, ''));
return Number.isFinite(numeric) ? numeric : null;
}
/** Parse a human date like "September 15 2025" → "2025-09-15". */
export function parseHumanDate(raw: string): string | null {
const cleaned = raw.replace(/(\d+)(st|nd|rd|th)/i, '$1').trim();
// Force UTC interpretation by appending a Z; otherwise dates without an
// explicit zone get parsed in the runner's local TZ and `toISOString()`
// shifts the day by ±1 (caught a -0700 -> 09-14 regression locally).
const d = new Date(cleaned + ' UTC');
if (Number.isNaN(d.getTime())) return null;
return d.toISOString().slice(0, 10);
}
/** Convert "206' 8\"" or "82" → 206.667 / 82. Returns null on parse failure. */
export function parseFeetInches(raw: string): number | null {
const trimmed = raw.trim();
const ftIn = trimmed.match(/^([0-9]+)\s*'\s*([0-9]+)\s*"$/);
if (ftIn) {
return Number(ftIn[1]) + Number(ftIn[2]) / 12;
}
const ftOnly = trimmed.match(/^([0-9]+(?:\.[0-9]+)?)/);
if (ftOnly) return Number(ftOnly[1]);
return null;
}