Fixed issue #132 and #133

[fedelemantuano]

Summary

• Address headers where the display name is an email address (e.g. From: alice@example.com <bob@example.com>) were silently dropped, returning [["", ""]] instead of the real values
• Absent address headers (bcc, cc, reply-to, delivered-to) appeared in the output as [["", ""]] instead of being omitted

Root cause

Both bugs trace back to Python's email.utils.getaddresses, which returns [('', '')] — a non-empty list with one empty tuple — for input it cannot parse. This includes both absent headers (empty string input) and headers with RFC-non-compliant display names.

The CVE-2023-27043 security hardening (backported to Python 3.9+) made strict=True the default. This correctly rejects unquoted @ in display names per RFC 5322 §3.4. The right call for an MTA — but mail-parser is a security/forensics tool: hiding an address because its display name looks like an email address (a common pattern in phishing and impersonation mail) defeats the purpose of the tool.

Changes

Bug 1 — Email address as display name is silently dropped (fixes #132)

New get_addresses() helper in utils.py:

1. Tries strict=True first — CVE hardening preserved for well-formed input
2. If every result is ('', '') on a non-empty header, falls back to a regex that extracts angle-bracket addresses and display names directly from the raw header value
3. The function is fully documented with the RFC/CVE rationale

Before:

From: alice@example.com <bob@example.com>

"from": [["", ""]]

After:

"from": [["alice@example.com", "bob@example.com"]]

Bug 2 — Absent address headers appear in output as [["", ""]]

One-line guard — if email_addr — added to the list comprehension in __getattr__. Any tuple with an empty address is filtered out, so absent headers produce [], which is falsy and excluded from the output.

Before:

"bcc":      [["", ""]],
"cc":       [["", ""]],
"reply-to": [["", ""]]

After: keys are absent from the output entirely.

Test plan

• tests/mails/mail_test_19 — new synthetic email with email-as-display-name in From, CC, Reply-To, and mixed To
• TestEmailAsDisplayName — 14 tests covering both bugs and edge cases:
  • alice@example.com <bob@example.com> → [("alice@example.com", "bob@example.com")]
  • bob@example.com <bob@example.com> → [("", "bob@example.com")] (name == addr, suppressed)
  • "alice@example.com" <bob@example.com> → strict path, no change
  • Alice Smith <alice@example.com> → strict path, no change
  • alice@example.com (bare) → no change
  • Header absent → [], key omitted from output
  • Multiple addresses all using email-as-name → all recovered
• 187 passed | coverage 99% | pre-commit all hooks passed

[coveralls]

coverage: 99.662% (-0.1%) from 99.797%
when pulling 6c2a5e0 on issue_133
into f6df398 on develop.