Looking at a Mail Message
Take a minute to look at the XML format returned by the above query. It's the genuine internal document format used by MarkMail to represent emails. You may be thinking, hey, emails aren't usually written in the XML format. That's right, they're usually sent over the wire in a format known as RFC 822, and persisted on disk in a variety of formats like mbox, maildir, or PST. MarkMail reads email in these various formats and converts them into this singular XML model.
The <message> root element has several attributes that provide metadata about
the email: its list, date, MarkMail id, mail message id, etc. Below that are
all the various mail headers are gathered under a <headers> element. A
handful of important headers (from, to, cc, subject) have attributes on them
which hold strings extracted from the raw header value.
The <body> element holds the actual text. Each paragraph gets its own <para>
element, including a quote depth. The initial greeting and footer (if any)
are detected during the load and identified with corresponding elements, which
makes it easy to exclude them from searches. Any email addresses and URLs in
the body are tagged. That way, during display, the emails can be obfuscated
and URLs rendered as hyperlinks. (You can see we obfuscated the email
addresses in the source here so spammers can't use this site to gather
addresses. In the real MarkMail database the emails are unobfuscated in the
source but obfuscated on display and revealed only after the user solves a
CAPTCHA. That's the only change between this data and the live site.)
What's most interesting about this data model is how simple it is. It enables a robust, feature-rich site but the model fits on the back of a napkin. An email is one logical item, represented by one physical item in the database. Try to imagine how you'd model, retrieve, and query this kind of textual, hierarchical, irregular data using the relational model.
In the above query we gave you a document identifier to use. If we weren't around giving you ids, what would you do? Well, you might just ask for a message at random. Here's how you do that:
This query returns the first document in the database (order is essentially random, though it is stable and doesn't change between executions). The first message happens to be written in Russian. That's fine; MarkLogic supports storage and searching of all languages whose text can be represented in XML, and provides advanced support (tokenization, stemming, and collation) for more than a dozen languages, including Russian.
If you look at the message's id attribute and do some copy/paste work you
can derive a URL to see this mail on MarkMail. In fact, here's a query to
generate that link for you:
Pro tip: You can drop the leading doc() and you get the same result. Any
rooted XPath is assumed to query across all documents. Let's use that
shortcut and ask for ten random emails to explore:
Lay of the Land
Drilling in with XPath
Stack Overflow: Get the most useful answers to questions from the MarkLogic community, or ask your own question.