Two-column fixture phone (555-0142) not detected — 7-digit local format below PHONE_RE 10-digit floor

[s-annam]

Problem

The two-column sidebar fixture's phone number is present in the extracted plain text (555-0142) but the parser reports "Phone not detected" in the Contact card (3 of 5 detected). This drags Completeness scoring (the score's "Missing: phone, LinkedIn" line).

Surfaced during Playwright QA of the #60 two-column reading-order fix.

Observed (live UI, chromium-two-column-sidebar.pdf)

• Extracted plain text contains: Jamie Rivera Software Engineering Leader 555-0142 jamie.rivera@example.com …
• Contact card shows Phone — not detected (⚠).

Root cause

This is a fixture artifact, not a general parser gap. PHONE_RE in src/lib/heuristics/regex.ts:19:

export const PHONE_RE =
  /(?:(?:\+?\d{1,3})[\s.-]?)?\(?\d{3}\)?[\s.-]?\d{3}[\s.-]?\d{4}\b/g;

requires a full 10-digit number (\d{3} area + \d{3} + \d{4}, optional country code). The fixture's 555-0142 is only 7 digits (\d{3}-\d{4}) → no match. A real US résumé phone is 10 digits, so the regex is correct for production; the fixture simply carries an unmatchable 7-digit local number.

Broadening PHONE_RE to accept bare 7-digit runs is not advised — it would false-match zip+4, ID numbers, and stray NNN-NNNN runs (e.g. a 1985 - 1989-style range with a dash).

Proposed fix

Primary — fix the fixture to a realistic 10-digit synthetic number. Re-export chromium-two-column-sidebar.pdf with a full synthetic 555 phone that a real résumé would use and that PHONE_RE matches, e.g.:

• (512) 555-0142, or
• 512-555-0142, or
• 555-555-0142

(555-01XX is the reserved fictional block — keep the 555-0142 suffix, prepend a synthetic area+exchange so it's a valid 10-digit shape. Stays PII-free.)

Then re-bake the snapshot — Contact becomes 4/5 and Completeness improves accordingly.

Decision to confirm: should a 7-digit local number ever be surfaced? Recommendation: no — keep PHONE_RE 10-digit-only (production-correct, low false-positive), and fix the fixture. Document this in the fixture README so future synthetic personas use 10-digit numbers.

Affected files

• tests/fixtures/pdfs/unknown/chromium-two-column-sidebar.pdf — re-export with a 10-digit synthetic phone.
• tests/fixtures/pdfs/unknown/chromium-two-column-sidebar.expected.json — re-bake (npm run bake-fixtures); Contact field count + Completeness shift.
• tests/fixtures/pdfs/README.md — add a one-line note: synthetic personas use full 10-digit 555 phone numbers so PHONE_RE matches.
• src/lib/heuristics/regex.ts:19 — no change (documented as intentional 10-digit floor).

Acceptance criteria

• chromium-two-column-sidebar.pdf re-exported with a 10-digit synthetic 555 phone; pdftotext confirms the new number and that name/email stay synthetic (Jamie Rivera / @example.com).
• Live parse surfaces the phone — Contact card shows 4 of 5 detected; phone no longer in the score's missing list for this fixture.
• chromium-two-column-sidebar.expected.json re-baked and committed alongside the PDF.
• Fixture README documents the 10-digit synthetic-phone convention.
• PHONE_RE left at 10-digit (no broadening); npm run test green.

Notes

• Follow-up to Two-column resumes: sidebar interleaves into experience — detection not wired to text assembly #60. Low priority — cosmetic for one fixture, no production-code change.
• PII policy: verify the re-exported binary with pdftotext tests/fixtures/pdfs/unknown/chromium-two-column-sidebar.pdf - | head -40 before commit (binary is the exposure surface, not the PR prose).

[s-annam]

Implementation Complete — 12f23f8 (PR #100)

Acceptance Criteria

• 10-digit synthetic 555 phone — fixture now carries (512) 555-0142; pdftotext confirms synthetic persona (Jamie Rivera / jamie.rivera@example.com).
• Phone surfaced; Contact improves — phone added to fieldsPopulated; completeness 24→27; phone removed from the score's missing list (only LinkedIn remains).
• expected.json re-baked + committed alongside the PDF.
• README documents the 10-digit synthetic-phone convention.
• PHONE_RE left untouched (no 7-digit broadening); npm run test green — 407/407, no src/ change.

Note on approach + churn

In-place phone patching was infeasible — the fixture uses CID-encoded Type0 fonts, so 555-0142 has no ASCII byte run to swap. The PDF was regenerated from reconstructed HTML (no committed HTML source exists for these print-to-pdf fixtures). Consequence: skillsCount (5→10), experienceCount (7→5), and sectionSource (markdown→regex) shifted — layout-detection artifacts of regeneration, not parser regressions. The two_column trigger and the fixture's #60 purpose are preserved.

Root-cause note: the original issue body's "PHONE_RE 10-digit floor" framing is now stale — detection runs through libphonenumber (findFirstPhone), which also rejects the 7-digit number. The fix (valid 10-digit fixture, no parser change) is unchanged.

Closes on merge of #100.