Files
libredesk/internal/inbox/channel/email/imap.go

569 lines
18 KiB
Go

package email
import (
"context"
"crypto/tls"
"encoding/json"
"fmt"
"strings"
"time"
"github.com/abhinavxd/libredesk/internal/attachment"
"github.com/abhinavxd/libredesk/internal/conversation/models"
"github.com/abhinavxd/libredesk/internal/envelope"
"github.com/abhinavxd/libredesk/internal/stringutil"
umodels "github.com/abhinavxd/libredesk/internal/user/models"
"github.com/emersion/go-imap/v2"
"github.com/emersion/go-imap/v2/imapclient"
"github.com/jhillyerd/enmime"
"github.com/volatiletech/null/v9"
)
const (
defaultReadInterval = time.Duration(5 * time.Minute)
defaultScanInboxSince = time.Duration(48 * time.Hour)
)
// ReadIncomingMessages reads and processes incoming messages from an IMAP server based on the provided configuration.
func (e *Email) ReadIncomingMessages(ctx context.Context, cfg IMAPConfig) error {
readInterval, err := time.ParseDuration(cfg.ReadInterval)
if err != nil {
e.lo.Warn("could not parse IMAP read interval, using the default read interval of 5 minutes", "interval", cfg.ReadInterval, "inbox_id", e.Identifier(), "error", err)
readInterval = defaultReadInterval
}
scanInboxSince, err := time.ParseDuration(cfg.ScanInboxSince)
if err != nil {
e.lo.Warn("could not parse IMAP scan inbox since duration, using the default value of 48 hours", "interval", cfg.ScanInboxSince, "inbox_id", e.Identifier(), "error", err)
scanInboxSince = defaultScanInboxSince
}
readTicker := time.NewTicker(readInterval)
defer readTicker.Stop()
for {
select {
case <-ctx.Done():
return nil
case <-readTicker.C:
// If the ticker interval is too short, it may trigger while the previous `processMailbox` call is still running,
// leading to overlapping executions or delays in handling context cancellation, check if the context is already done.
if ctx.Err() != nil {
return nil
}
if err := e.processMailbox(ctx, scanInboxSince, cfg); err != nil && err != context.Canceled {
e.lo.Error("error searching emails", "error", err)
}
e.lo.Info("email search complete", "mailbox", cfg.Mailbox, "inbox_id", e.Identifier())
}
}
}
// processMailbox processes emails in the specified mailbox.
func (e *Email) processMailbox(ctx context.Context, scanInboxSince time.Duration, cfg IMAPConfig) error {
var (
client *imapclient.Client
err error
)
address := fmt.Sprintf("%s:%d", cfg.Host, cfg.Port)
imapOptions := &imapclient.Options{
TLSConfig: &tls.Config{
InsecureSkipVerify: cfg.TLSSkipVerify,
},
}
switch cfg.TLSType {
case "none":
client, err = imapclient.DialInsecure(address, imapOptions)
case "starttls":
client, err = imapclient.DialStartTLS(address, imapOptions)
case "tls":
client, err = imapclient.DialTLS(address, imapOptions)
default:
return fmt.Errorf("unknown IMAP TLS type: %q", cfg.TLSType)
}
if err != nil {
return fmt.Errorf("failed to connect to IMAP server: %w", err)
}
defer client.Logout()
if err := client.Login(cfg.Username, cfg.Password).Wait(); err != nil {
return fmt.Errorf("error logging in to the IMAP server: %w", err)
}
if _, err := client.Select(cfg.Mailbox, &imap.SelectOptions{ReadOnly: true}).Wait(); err != nil {
return fmt.Errorf("error selecting mailbox: %w", err)
}
// Scan emails since the specified duration.
since := time.Now().Add(-scanInboxSince)
e.lo.Info("searching emails", "since", since, "mailbox", cfg.Mailbox, "inbox_id", e.Identifier())
// Search for messages in the mailbox.
searchResults, err := e.searchMessages(client, since)
if err != nil {
return fmt.Errorf("error searching messages: %w", err)
}
return e.fetchAndProcessMessages(ctx, client, searchResults, e.Identifier())
}
// searchMessages searches for messages in the specified time range.
func (e *Email) searchMessages(client *imapclient.Client, since time.Time) (*imap.SearchData, error) {
searchCMD := client.Search(&imap.SearchCriteria{
Since: since,
},
&imap.SearchOptions{
ReturnMin: true,
ReturnMax: true,
ReturnAll: true,
ReturnCount: true,
},
)
return searchCMD.Wait()
}
// fetchAndProcessMessages fetches and processes messages based on the search results.
func (e *Email) fetchAndProcessMessages(ctx context.Context, client *imapclient.Client, searchResults *imap.SearchData, inboxID int) error {
seqSet := imap.SeqSet{}
seqSet.AddRange(searchResults.Min, searchResults.Max)
// Fetch envelope and headers needed for auto-reply detection.
fetchOptions := &imap.FetchOptions{
Envelope: true,
BodySection: []*imap.FetchItemBodySection{
{
Specifier: imap.PartSpecifierHeader,
HeaderFields: []string{
headerAutoSubmitted,
headerAutoreply,
headerLibredeskLoopPrevention,
headerMessageID,
},
},
},
}
// Collect messages to process later.
type msgData struct {
env *imap.Envelope
seqNum uint32
autoReply bool
isLoop bool
extractedMessageID string
}
var messages []msgData
fetchCmd := client.Fetch(seqSet, fetchOptions)
// Extract the inbox email address.
inboxEmail, err := stringutil.ExtractEmail(e.FromAddress())
if err != nil {
e.lo.Error("failed to extract email address from the 'From' header", "error", err)
return fmt.Errorf("failed to extract email address from 'From' header: %w", err)
}
if inboxEmail == "" {
e.lo.Error("inbox email address is empty, cannot process messages", "inbox_id", e.Identifier())
return fmt.Errorf("inbox (%d) email address is empty, cannot process messages", e.Identifier())
}
for {
// Check for context cancellation before fetching the next message.
select {
case <-ctx.Done():
return ctx.Err()
default:
}
// Fetch the next message.
msg := fetchCmd.Next()
if msg == nil {
// No more messages to process.
break
}
var (
env *imap.Envelope
autoReply bool
isLoop bool
extractedMessageID string
)
// Process all fetch items for the current message.
for {
// Check for context cancellation before processing the next item.
select {
case <-ctx.Done():
return ctx.Err()
default:
}
// Fetch the next item in the message.
item := msg.Next()
if item == nil {
// No message items left to process.
break
}
// Body section.
if bs, ok := item.(imapclient.FetchItemDataBodySection); ok && bs.Literal != nil {
envelope, err := enmime.ReadEnvelope(bs.Literal)
if err != nil {
e.lo.Error("error reading envelope", "error", err)
continue
}
if isAutoReply(envelope) {
autoReply = true
}
if isLoopMessage(envelope, inboxEmail) {
isLoop = true
}
// Extract Message-Id from raw headers as fallback for problematic Message IDs
extractedMessageID = extractMessageIDFromHeaders(envelope)
}
// Envelope.
if ed, ok := item.(imapclient.FetchItemDataEnvelope); ok {
env = ed.Envelope
}
}
// Skip if we couldn't get the envelope.
if env == nil {
e.lo.Warn("skipping message without envelope", "seq_num", msg.SeqNum, "inbox_id", e.Identifier())
continue
}
messages = append(messages, msgData{env: env, seqNum: msg.SeqNum, autoReply: autoReply, isLoop: isLoop, extractedMessageID: extractedMessageID})
}
// Now process each collected message.
for _, msgData := range messages {
// Check for context cancellation before processing each message.
select {
case <-ctx.Done():
return ctx.Err()
default:
}
// Skip if this is an auto-reply message.
if msgData.autoReply {
e.lo.Info("skipping auto-reply message", "subject", msgData.env.Subject, "message_id", msgData.env.MessageID)
continue
}
// Skip if this message is a loop prevention message.
if msgData.isLoop {
e.lo.Info("skipping message with loop prevention header", "subject", msgData.env.Subject, "message_id", msgData.env.MessageID)
continue
}
// Process the envelope.
if err := e.processEnvelope(ctx, client, msgData.env, msgData.seqNum, inboxID, msgData.extractedMessageID); err != nil && err != context.Canceled {
e.lo.Error("error processing envelope", "error", err)
}
}
return nil
}
// processEnvelope processes a single email envelope.
func (e *Email) processEnvelope(ctx context.Context, client *imapclient.Client, env *imap.Envelope, seqNum uint32, inboxID int, extractedMessageID string) error {
if len(env.From) == 0 {
e.lo.Warn("no sender received for email", "message_id", env.MessageID)
return nil
}
var fromAddress = strings.ToLower(env.From[0].Addr())
// Determine final Message ID - prefer IMAP-parsed, fallback to raw header extraction
messageID := env.MessageID
if messageID == "" {
messageID = extractedMessageID
if messageID != "" {
e.lo.Debug("using raw header Message-ID as fallback for malformed ID", "message_id", messageID, "subject", env.Subject, "from", fromAddress)
}
}
// Drop message if we still don't have a valid Message ID
if messageID == "" {
e.lo.Error("dropping message: no valid Message-ID found in IMAP parsing or raw headers", "subject", env.Subject, "from", fromAddress)
return nil
}
// Check if the message already exists in the database; if it does, ignore it.
exists, err := e.messageStore.MessageExists(messageID)
if err != nil {
e.lo.Error("error checking if message exists", "message_id", messageID)
return fmt.Errorf("checking if message exists in DB: %w", err)
}
if exists {
return nil
}
// Check if contact with this email is blocked / disabed, if so, ignore the message.
if contact, err := e.userStore.GetContact(0, fromAddress); err != nil {
envErr, ok := err.(envelope.Error)
if !ok || envErr.ErrorType != envelope.NotFoundError {
e.lo.Error("error checking if user is blocked", "email", fromAddress, "error", err)
return fmt.Errorf("checking if user is blocked: %w", err)
}
} else if !contact.Enabled {
e.lo.Debug("contact is blocked, ignoring message", "email", fromAddress)
return nil
}
e.lo.Debug("processing new incoming message", "message_id", messageID, "subject", env.Subject, "from", fromAddress, "inbox_id", inboxID)
// Make contact.
firstName, lastName := getContactName(env.From[0])
var contact = umodels.User{
InboxID: inboxID,
FirstName: firstName,
LastName: lastName,
SourceChannel: null.NewString(e.Channel(), true),
SourceChannelID: null.NewString(fromAddress, true),
Email: null.NewString(fromAddress, true),
Type: umodels.UserTypeContact,
}
// Lowercase and set the `to`, `cc`, `from` and `bcc` addresses in message meta.
var ccAddr = make([]string, 0, len(env.Cc))
var toAddr = make([]string, 0, len(env.To))
var bccAddr = make([]string, 0, len(env.Bcc))
var fromAddr = make([]string, 0, len(env.From))
for _, cc := range env.Cc {
if cc.Addr() != "" {
ccAddr = append(ccAddr, strings.ToLower(cc.Addr()))
}
}
for _, to := range env.To {
if to.Addr() != "" {
toAddr = append(toAddr, strings.ToLower(to.Addr()))
}
}
for _, bcc := range env.Bcc {
if bcc.Addr() != "" {
bccAddr = append(bccAddr, strings.ToLower(bcc.Addr()))
}
}
for _, from := range env.From {
if from.Addr() != "" {
fromAddr = append(fromAddr, strings.ToLower(from.Addr()))
}
}
meta, err := json.Marshal(map[string]interface{}{
"from": fromAddr,
"cc": ccAddr,
"bcc": bccAddr,
"to": toAddr,
"subject": env.Subject,
})
if err != nil {
e.lo.Error("error marshalling meta", "error", err)
return fmt.Errorf("marshalling meta: %w", err)
}
incomingMsg := models.IncomingMessage{
Message: models.Message{
Channel: e.Channel(),
SenderType: models.SenderTypeContact,
Type: models.MessageIncoming,
InboxID: inboxID,
Status: models.MessageStatusReceived,
Subject: env.Subject,
SourceID: null.StringFrom(messageID),
Meta: meta,
},
Contact: contact,
InboxID: inboxID,
}
// Fetch full message body.
fetchOptions := &imap.FetchOptions{
BodySection: []*imap.FetchItemBodySection{{}},
}
seqSet := imap.SeqSet{}
seqSet.AddNum(seqNum)
fullFetchCmd := client.Fetch(seqSet, fetchOptions)
fullMsg := fullFetchCmd.Next()
if fullMsg == nil {
return nil
}
// Fetch full message.
for {
// Check for context cancellation before processing the next item.
select {
case <-ctx.Done():
return ctx.Err()
default:
}
fullFetchItem := fullMsg.Next()
if fullFetchItem == nil {
return nil
}
if fullItem, ok := fullFetchItem.(imapclient.FetchItemDataBodySection); ok {
e.lo.Debug("fetching full message body", "message_id", messageID)
return e.processFullMessage(fullItem, incomingMsg)
}
}
}
// processFullMessage processes the full message and enqueues it for inserting into the database.
func (e *Email) processFullMessage(item imapclient.FetchItemDataBodySection, incomingMsg models.IncomingMessage) error {
envelope, err := enmime.ReadEnvelope(item.Literal)
if err != nil {
e.lo.Error("error parsing email envelope", "error", err, "message_id", incomingMsg.Message.SourceID.String)
for _, err := range envelope.Errors {
e.lo.Error("error parsing email envelope. envelope_error: ", "error", err.Error(), "message_id", incomingMsg.Message.SourceID.String)
}
return fmt.Errorf("parsing email envelope: %w", err)
}
// Log any envelope errors.
for _, err := range envelope.Errors {
e.lo.Error("error parsing email envelope", "error", err.Error(), "message_id", incomingMsg.Message.SourceID.String)
}
// Extract all HTML content by traversing the tree
var allHTML strings.Builder
if envelope.Root != nil {
htmlParts := extractAllHTMLParts(envelope.Root)
if len(htmlParts) > 0 {
allHTML.WriteString("<div>")
for _, part := range htmlParts {
allHTML.WriteString(part)
}
allHTML.WriteString("</div>")
}
}
// Set message content - prioritize combined HTML
if allHTML.Len() > 0 {
incomingMsg.Message.Content = allHTML.String()
incomingMsg.Message.ContentType = models.ContentTypeHTML
e.lo.Debug("extracted HTML content from parts", "message_id", incomingMsg.Message.SourceID.String, "content", incomingMsg.Message.Content)
} else if len(envelope.HTML) > 0 {
incomingMsg.Message.Content = envelope.HTML
incomingMsg.Message.ContentType = models.ContentTypeHTML
} else if len(envelope.Text) > 0 {
incomingMsg.Message.Content = envelope.Text
incomingMsg.Message.ContentType = models.ContentTypeText
}
e.lo.Debug("envelope HTML content", "message_id", incomingMsg.Message.SourceID.String, "content", incomingMsg.Message.Content)
e.lo.Debug("envelope text content", "message_id", incomingMsg.Message.SourceID.String, "content", envelope.Text)
// Clean headers
inReplyTo := strings.ReplaceAll(strings.ReplaceAll(envelope.GetHeader("In-Reply-To"), "<", ""), ">", "")
references := strings.Fields(envelope.GetHeader("References"))
for i, ref := range references {
references[i] = strings.Trim(strings.TrimSpace(ref), " <>")
}
incomingMsg.Message.InReplyTo = inReplyTo
incomingMsg.Message.References = references
// Process attachments
for _, att := range envelope.Attachments {
incomingMsg.Message.Attachments = append(incomingMsg.Message.Attachments, attachment.Attachment{
Name: att.FileName,
Content: att.Content,
ContentType: att.ContentType,
ContentID: att.ContentID,
Size: len(att.Content),
Disposition: attachment.DispositionAttachment,
})
}
// Process inlines - treat ones without ContentID as regular attachments
for _, inline := range envelope.Inlines {
disposition := attachment.DispositionInline
if inline.ContentID == "" {
disposition = attachment.DispositionAttachment
}
incomingMsg.Message.Attachments = append(incomingMsg.Message.Attachments, attachment.Attachment{
Name: inline.FileName,
Content: inline.Content,
ContentType: inline.ContentType,
ContentID: inline.ContentID,
Size: len(inline.Content),
Disposition: disposition,
})
}
e.lo.Debug("enqueuing incoming email message", "message_id", incomingMsg.Message.SourceID.String,
"attachments", len(envelope.Attachments), "inline_attachments", len(envelope.Inlines))
if err := e.messageStore.EnqueueIncoming(incomingMsg); err != nil {
return err
}
return nil
}
// getContactName extracts the contact's first and last name from the IMAP address.
func getContactName(imapAddr imap.Address) (string, string) {
from := strings.TrimSpace(imapAddr.Name)
names := strings.Fields(from)
if len(names) == 0 {
return imapAddr.Host, ""
}
if len(names) == 1 {
return names[0], ""
}
return names[0], names[1]
}
// isAutoReply checks if a given email envelope indicates an auto-reply message.
func isAutoReply(envelope *enmime.Envelope) bool {
if as := strings.ToLower(strings.TrimSpace(envelope.GetHeader("Auto-Submitted"))); as != "" && as != "no" {
return true
}
if strings.TrimSpace(envelope.GetHeader("X-Autoreply")) != "" {
return true
}
return false
}
// isLoopMessage returns true if the email is a loop prevention message. i.e., it has the `X-Libredesk-Loop-Prevention` header with the inbox email address.
func isLoopMessage(envelope *enmime.Envelope, inboxEmailaddress string) bool {
loopHeader := envelope.GetHeader(headerLibredeskLoopPrevention)
if loopHeader == "" {
return false
}
return strings.EqualFold(loopHeader, inboxEmailaddress)
}
// extractAllHTMLParts extracts all HTML parts from the given enmime part by traversing the tree.
func extractAllHTMLParts(part *enmime.Part) []string {
var htmlParts []string
// Check current part
if strings.HasPrefix(part.ContentType, "text/html") && len(part.Content) > 0 {
htmlParts = append(htmlParts, string(part.Content))
}
// Process children recursively
for child := part.FirstChild; child != nil; child = child.NextSibling {
childParts := extractAllHTMLParts(child)
htmlParts = append(htmlParts, childParts...)
}
return htmlParts
}
// extractMessageIDFromHeaders extracts and cleans the Message-ID from email headers.
// This function handles problematic Message IDs by extracting them from raw headers
// and cleaning them of angle brackets and whitespace.
func extractMessageIDFromHeaders(envelope *enmime.Envelope) string {
if rawMessageID := envelope.GetHeader(headerMessageID); rawMessageID != "" {
return strings.TrimSpace(strings.Trim(rawMessageID, "<>"))
}
return ""
}