PHP解析Gmail邮件回复

Question

I am trying to parse GMail emails, but have one problem: how do I know which message a reply corresponds to?

I tried sorting email by subject. For example, if a message has the subject "hi Jack", then all messages with subject "Re: hi Jack" are a reply to this mail.

But what do I do if I have many emails with the same subject? How do I know which email they are replies to?

Do emails perhaps have a unique code for what the reply goes to? Maybe there is an ID or something like that to know what the children of a message are(?).

Answer 1

Threading by subject is not a good idea because there may be as you noticed several different threads based on identical subjects.

You need to examine 3 headers in the message to make threading (or other kind of grouping) possible:

Message-ID: contains unique message identifier (what you call "unique code") in a string surrounded by < and > characters eg <123456@User1PC> Most MUAs will create identifiers in above form or something similar to that. This header should be generated when a new message is sent.

In-Reply-To: contains a message this particular reply is related to eg <789abcd@User2PC> . This header should be copied from Message-ID it replies to.

References: contains list of recent references to messages in this "thread". The format is similar to above except they are separated eg <123456@User1PC> <789abcd@User2PC> It is there so that you can use it to locate message in the thread.

If one message has been replied or posted a few days later it might be hard to locate it without list of references. Usually list of references is trimmed by mail clients to a reasonable size. By reasonable, I mean, trimming it enough to be able to locate message in a thread but keep the header under reasonable size (not having too many references). For example it may contain 5-10 references which is more than enough usually to connect it to other messages. References: are also useful in case if original message (first one) has been deleted so even without it, you can still utilize References: list to build a threaded (grouped) messages.

So, in order to thread messages, you would need to read all of them, and then sort threads based on the information you can extract from above headers.

If references or message ids are not in form you can recognize (eg <example@something> you can bail out by not threading these messages and displaying them as unthreaded. So generic algorithm for threading/locating might look something like this:

1. Take first message ID
2. Examine nearby (by date) messages to see if one of them contains message ID in its references list or in-reply-to - if there are none - you can't group it so keep it as standalone message.
3. Group messages somehow, perhaps based on Date: , or Received: header
4. Place this message into "Done" list so you don't need to examine it further (or related references)
5. Continue until you can't find any more references and then move to next message which is not already in "Done" list and repeat steps until you process entire message list.

It will probably take you a while to get this done properly but now at least you have a starting point to look into.