How to query XML like SQL?

XML is a hierarchy of nodes with types of Text Node, Element Node, and therefore... Table 1: Documents [DocumentId | DocumentName | FileName] 1 | My xml document | index.xml] Table 2: Element Nodes [DocumentId | NodeId | ParentNodeId | Namespace | ElementName | SiblingNodeOrder] 1 | 1 | null | Element | http://w3.org/... | root | 0 1 | 2 | null | Element | null | paragraph | 0 1 | 3 | null | Element | null | paragraph | 1 Table 3: Attributes [NodeId | Namespace | Name | Value] 3 | null | style | text-decoration: blink Table 4: Text Nodes [NodeId | Value | SiblingNodeOrder] 2 | This is stupid 3 | THE ONLY PEOPLE WHO WOULD TRY TO DO THIS ARE IDIOTS WHO HAVE BEEN RESTING ON THEIR LAURELS AND POSTING ON SLASHDOT FOR THE LAST FIVE YEARS WITHOUT LEARNING WHAT THE XML IS BUT WHY REASON WITH THOSE THICK HEADS THEY PROBABLY THINK VIRTUAL MACHINES ARE OVERKILL AND UNICODE IS A FAD | 0

I can't think of ANY realworld xml that doesn't map meaningfully to the relational model. Wow. That's just... wow. How about every XHTML file on earth?

ook - is there a way to test CONTAINS or BEGINS-WITH style clauses? IE, counting foo nodes where value begins with 'baz'?

I have to agree that you want XPath, and if you want to do queries that XPath won't support, you want XQuery. http://www.sleepycat.com/products/bdbxml.html supports XQuery and XPath. I'm not sure how well it works; the only XQuery database I've used is a huge commercial product.

nomisxid: XPath has a contains() and a starts-with(). See the http://www.w3.org/TR/xpath.

The XPath equivalents are starts-with() and contains(). The fact that you need those, however, implies that your XML schema is badly in need of a redesign. Which, given that your dev group's suggestion was to use grep to solve this problem, is perhaps not too surprising :)

ook, isn't xhtml the PRESENTATION of data, not storage of data? It is related to XML, but they are not the same thing, nor used for the same purposes. That being said, I don't buy it, even for xhtml. Show me your xhtml that isn't a structured document. The whole point of a markup language is to present your data in a non-random order so that it has meaning. Every example I see of xhtml is just reading data from a datasource and presenting it on a webpage.

thanx nev, I was looking at this stupid page, http://www.w3schools.com/xpath/xpath_operators.asp which made it seem like the handful of operators they listed was the definitive list.

I'm sure there are a lot of XML documents which could very easily be translated into an SQL database. Take this example from W3Schools introduction to XSLT: http://www.w3schools.com/xsl/cdcatalog.xml Excerpt: <catalog> <cd> <title>Empire Burlesque</title> <artist>Bob Dylan</artist> <country>USA</country> <company>Columbia</company> <price>10.90</price> <year>1985</year> </cd> <cd> <title>Hide your heart</title> <artist>Bonnie Tyler</artist> <country>UK</country> <company>CBS Records</company> <price>9.90</price> <year>1988</year> </cd> [etc] <catalog> It's relationships are so simple that of course you could transfer it into a database in five minutes. You'd have a "titles" table, a "titles_artists" linking table, a "titles_companies" linking table, a couple more and you're done. Do a join to get all artists on Columbia, or all albums by Bob Dylan, or all albums by Bob Dylan on Columbia and so on. But that's a deceptively simple example. A CD has a title, an artist and a couple of other things. It's nested one level deep and that's it. Because of the arbritrary levels of complexity which are possible in XML, you have no way of knowing if any given XML file could be easily translated. What about a file with nesting seventeen levels deep, with attributes at each level? Do you end up with 172 tables to describe all the relationships? Maybe it's 1717 but whatever, it's too many. You'd either have to use human judgement to decide which relationships were important and discard the rest or end up creating thousands of tables to describe the relationships between different bits of data. XSLT and XPath are obviously much better ways to get at the data matching any given relationship.