Question

<Root>
  <P1 Text ="A" >
    <P2 Text = "AA">
      <P3 Text = "AAA">
        <L Text = "l_A"/>
        <L Text = "l_B"/>
        <L Text = "l_C"/>
      </P3>
      <P3 Text = "BBB">
        <L Text = "l_D"/>
        <L Text = "l_E"/>
        <L Text = "l_F"/>
      </P3>
    </P2>
    <P2 Text = "BB">
       <L Text = "l_G"/>
       <L Text = "l_H"/>
       <L Text = "l_I"/>
    </P2>
  </P1>
</Root>  

From an XML document containing thousands of variably nested nodes up to 10 levels deep, I
would like to retrieve programmatically only the leaves belonging to any of the “P” parents
as follows: for instance, in the example above, selecting P2 “AA” would yield l_A to l_F and P3 “BBB” would give l_D to l_F.

Answer 1

Something like this (returns a list of strings):

    XDocument doc = XDocument.Load(@"test.xml");

    string level = "P3";
    string levelAttr = "AAA";

    var list = (from d in doc.Descendants(level)
                let xAttribute = d.Attribute("Text")
                where xAttribute != null && xAttribute.Value == levelAttr
                from l in d.Descendants("L")
                let lAttribute = l.Attribute("Text")
                where lAttribute != null
                select lAttribute.Value);

You may remove the attribute null checks if the Text attribute is always there…