Using XPath, How do I select a node based on its text content and value of an attribute?

Question

Given this XML:

<DocText>
<WithQuads>
    <Page pageNumber="3">
        <Word>
            July
            <Quad>
                <P1 X="84" Y="711.25" />
                <P2 X="102.062" Y="711.25" />
                <P3 X="102.062" Y="723.658" />
                <P4 X="84.0" Y="723.658" />
            </Quad>
        </Word>
        <Word>
        </Word>
        <Word>
            30,
            <Quad>
                <P1 X="104.812" Y="711.25" />
                <P2 X="118.562" Y="711.25" />
                <P3 X="118.562" Y="723.658" />
                <P4 X="104.812" Y="723.658" />
            </Quad>
        </Word>
    </Page>
</WithQuads>

I'd like to find the nodes that have text of 'July' and a Quad/P1/X attribute Greater than 90. Thus, in this case, it should not return any matches. However, if I use GT (>) or LT (<), I get a match on the first Word element. If I use eq (=), I get no match.

So:

//Word[text()='July' and //P1[@X < 90]]

will return true, as will

//Word[text()='July' and //P1[@X > 90]]

How do I constrain this properly on the P1@X attribute?

In addition, imagine I have multiple Page elements, for different page numbers. How would I additionally constrain the above search to find Nodes with text()='July', P1@X < 90, and Page@pageNumber=3?

Solution 1

Generally I would consider the use of an unprefixed // as a bad smell in an XPath.

Try this:-

/DocText/WithQuads/Page/Word[text()='July' and Quad/P1/@X > 90]

Your problem is that you use the //P1[@X < 90] which starts back at the beginning of the document and starts hunting any P1 hence it will always be true. Similarly //P1[@X > 90] is always true.

Solution 2

Apart form the "//" issue, this XML is a very weird use of mixed content. The predicate text()='July' will match the element if any child text node is exactly equal to July, which isn't true in your example because of surrounding whitespace. Depending on the exact definition of the source XML, I would go for [text()[normalize-space(.)='July'] and Quad/P1/@X > 90]

Comments

[marc esher]

thank you, Michael. I was wondering about the whitespace.... I formatted the sample before pasting into stack overflow, but my source XML is all "tight". When I ran the xpath against the formatted version it did indeed fail to work correctly. I will try using normalize-space(.)

[Michael Sorens]

I am surprised that this, in fact, worked because of the whitespace issues addressed in Michael Kay's answer. I tried this answer in a couple different XPath evaluators and it failed to match with either. Once I switched to the predicate with 'normalize-space', then I made a successful match.

[Mark Jeronimus]

You could use .//P1 to start hunting at the current level instead of specifying a fixed path