Subject Grouping

When a message arrives with no usable references (Message-ID, In-Reply-To, and References all missing or unmatched), jwz-go falls back to grouping by canonical subject. This recovers threads where intermediate MTAs or mailing-list software stripped headers.

Canonical form

CanonicalSubject lowercases the subject and repeatedly strips a small set of locale-aware reply/forward prefixes until none remain:

jwz.CanonicalSubject("Re: AW: SV: Release plan") // "release plan"
jwz.CanonicalSubject("Fwd: FW: Re: report")      // "report"
jwz.CanonicalSubject("  Hello  ")                // "hello"

Recognized prefixes

Prefix	Language
`Re:`	English (reply)
`Fwd:`, `Fw:`	English (forward)
`AW:`, `WG:`	German (Antwort, Weitergeleitet)
`Tr:`	French (Transféré)
`Reé:`, `Resp:`	Portuguese / Spanish (Resposta / Respuesta)
`SV:`, `VS:`	Swedish / Norwegian / Danish (Svar / Videresendt)
`RV:`	Spanish (Reenvío)
`ENC:`	Portuguese (Encaminhada)
`Antw:`	Dutch (Antwoord)
`Odp:`	Polish (Odpowiedź)
`R:`, `I:`	Italian (Risposta / Inoltro)

Tip

Need a missing locale? Open an issue with the prefix, the language, and a sample subject — adding one is a one-line change to subject.go.

Grouping rules

After all reference-based linking is done, the remaining roots are walked in input order. Each root's first message contributes a canonical subject. If a prior root already used that subject, the new root is linked under it. Otherwise, the new root is kept and registered.

Empty subjects are not grouped — they stay as independent threads.

Why grouping happens after references

Reference-based linking is high-confidence: those headers are explicit about the parent. Subject grouping is best-effort and produces false positives whenever two unrelated messages share a generic subject ("Question", "Quick update"). Running reference linking first means most real chains are already collapsed before subject grouping runs, so the fallback only kicks in for the genuinely orphan tail.