Question

I got html code like the following:

<p style="margin:0 0 0.5em 0;"><b>Blablub</b></p>
<table> ... </table>

Now I want to query the content of the <b> right above the table but only if the table does not have any attributes. I tried the following query:

//table[not(@*)]/preceding-sibling::p/b

If I remove the preceding-sibling::p/b part entirely it works. It gives me exactly the tables I need. However, if I use this query it gives me content of an <b> tag which precedes a table WITH attributes.

Answer 1

Use:

//table[not(@*)]/preceding-sibling::*[1][self::p]/b

This means: Select all b elements that are children of all p elements that are the first preceding sibling of a table that has no attributes.

This is quite different from the problematic expression cited in the question:

//table[not(@*)]/preceding-sibling::p[1]/b

The latter selects the b children of the first p following sibling — there is no guarantee that the first p following sibling is also the first element sibling.