Documentation
¶
Overview ¶
Package mailparse produces the "parsed view" of a raw RFC 5322 message — inbound mail and the retained copy of sent outbound mail (decision 9 / Slice 4b-3). It derives two representations server-side: the injection-reduced text an agent feeds to a model by default (prefer the text/plain part, else HTML→text, strip quoted reply chains / forwarded headers, cap the length as a token-stuffing guard), and the decoded text/html part for human display.
It is a CONVENIENCE, not a security control: the raw message is always available, and the security decision is made on structured metadata (the auth verdict + sender provenance, which survive as fields), never on this stripped text. So over-stripping is acceptable — the caller can fall back to raw. Best-effort throughout: a parse failure yields the best text available (possibly empty), never an error.
Index ¶
Constants ¶
const DefaultMaxBytes = 16 * 1024
DefaultMaxBytes caps the parsed text. Generous enough for real messages, small enough to blunt token-stuffing. Callers may override.
const MaxHTMLBytes = 1 << 20 // 1 MiB
MaxHTMLBytes caps the display HTML (ParsedView.HTML). Unlike the text view — which is capped tight as a token-stuffing guard for model input — the HTML is for human display, so the bound is generous: just a backstop against a pathological multi-MB part bloating every detail response. raw_message stays the full-fidelity escape hatch.
Variables ¶
This section is empty.
Functions ¶
func ParsedBody ¶
ParsedBody extracts the injection-reduced text (see ParsedView.Text). truncated is true when the cap cut content. maxBytes <= 0 uses DefaultMaxBytes. Retained for callers that only need the text; callers wanting HTML use Parse.
Types ¶
type Attachment ¶
type Attachment struct {
Filename string // decoded filename (RFC 2047/2231), "" if the part has none
ContentType string // media type only (e.g. "application/pdf"), params stripped
// ContentID is the part's Content-ID with the surrounding angle brackets
// stripped (e.g. "ii_abc@mail.gmail.com"), or "" if the part carries none.
// This is the key an HTML body's `cid:` reference resolves against, so a
// renderer can map `<img src="cid:ii_abc@mail.gmail.com">` to this
// attachment. Present mainly on inline images embedded by mail clients.
ContentID string
Data []byte // decoded bytes (Content-Transfer-Encoding applied)
}
Attachment is one decoded attachment part of a MIME message.
func AttachmentAt ¶
func AttachmentAt(raw []byte, index int) (Attachment, bool)
AttachmentAt returns the attachment at the 0-based index, or false if out of range. It re-walks the message (the caller holds the raw bytes); attachments are typically few, so a per-call walk is fine.
func Attachments ¶
func Attachments(raw []byte) []Attachment
Attachments walks the MIME tree of a raw RFC 5322 message and returns its attachment parts in stable document order (depth-first, as they appear in the wire bytes). Index N from this function is the SAME index used by AttachmentAt and by the message read view — it is the authoritative attachment index.
An "attachment" is a leaf part that carries a filename (in Content-Disposition or the Content-Type name param) OR an explicit `Content-Disposition: attachment`. Body parts (text/plain, text/html with no filename) and multipart containers are NOT attachments. Named inline parts (e.g. cid: images) ARE included — they are real fetchable bytes.
On a parse failure or a non-multipart message with no attachment part, returns an empty slice (never nil-panics).
type ParsedView ¶
type ParsedView struct {
// Text is the injection-reduced plain text: text/plain preferred (else
// HTML→text), quoted reply/forward chains stripped, capped at maxBytes.
Text string
// Truncated is true when the cap cut Text.
Truncated bool
// HTML is the decoded text/html part for display, or "" when the message
// carries no HTML part. Full fidelity (NOT quote-stripped, unlike Text) and
// capped at MaxHTMLBytes; raw_message stays the canonical copy.
HTML string
}
ParsedView is the derived view of a raw RFC 5322 message: the injection-reduced Text an agent feeds a model by default, plus the decoded HTML part for human display.
Source Files