[decius7bc]

#168 added whitespace stripping between adjacent RFC 2047 encoded-words, but the regex wasn't anchored. It removed whitespace before every =? in the rest of the value, so a plain word followed by an encoded-word lost its separating space:

raw =
  "Subject: =?utf-8?Q?Gr=C3=BC=C3=9Fe?= from the\r\n" <>
    " =?utf-8?Q?M=C3=BCller?= family\r\n" <>
    "From: a@example.com\r\n\r\nbody\r\n"

Mail.Parsers.RFC2822.parse(raw).headers["subject"]
# before: "Grüße from theMüller family"
# after:  "Grüße from the Müller family"

Folding isn't the cause. A one-line value like =?utf-8?Q?a?= x =?utf-8?Q?b?= decoded to "a xb" too.

RFC 2047 §6.2 says only linear white space between two adjacent encoded-words is ignored. Whitespace next to ordinary text must stay.

Changes proposed in this pull request

The regex in parse_encoded_word/2 now:

• only matches at the start of the remainder (\A), i.e. directly after the encoded-word that was just decoded;
• only matches linear white space, 1*([CRLF] (SP / HTAB)), instead of \s. \s also matches a bare CR, LF, VT or FF;
• only strips it when a complete encoded-word (=?charset?[BbQq]?text?=) follows, so a stray =? that isn't an encoded-word keeps the space before it.

Tests

• Adjacent encoded-words still decode to "ab": on one line, and folded with a space or a tab (the Ignore space between two encoded words (as per RFC 2047) #168 behaviour).
• Bare CR, LF, VT and FF between encoded-words are kept.
• Whitespace between an encoded-word and plain text is kept, including the folded Subject above and a stray =?.

🤖 Generated with Claude Code

[bcardarella]

@andrewtimberlake thoughts?