So I have a basic table structured:

<tbody>
  <tr>
    <td></td>
    <td></td>
  </tr>
  <tr>
    <td></td>
    <td></td>
  </tr>
</tbody> etc....

I’m trying to target a link <a> element in one cell only if the cell above it does or does not contain a word. For example:

<tbody>
  <tr>
    <td></td>
    <td><b>Fire Sale!</b></td>
  </tr>
  <tr>
    <td></td>
    <td><a href="something">linktext</a></td>
  </tr>
/tbody>

So I’d want to target the <a> only if the cell above it contains “Fire Sale!” or perhaps does not contain a word, I have scenarios both ways.

The problem is no matter what I do I can’t keep the conditional axes to find the cell right above, another <a> next to (<td> immediately above) the <a> we want is also targeted e.g.

<tbody>
  <tr>
    <td></td>
    <td><b>Fire Sale!</b></td>
  </tr>
  <tr>
    <td><a href="somethingelse">link I don't want</a></td>
    <td><a href="something">linktext</a></td>
  </tr>
/tbody>

I’ve tried stuff like:

//tr/td/b/a[@href]/ancestor::tbody/tr/td/b[contains(text(),'Fire Sale!')]

But no matter what, because of the odd relationship between tr and td I always end up getting an affirmative conditional. That is, they share the same ancestor tree structure for the most part and targeting back down to the <td> above my main target seems impossible. Is there some way to use variables or I feel count() might help but I’m just not sure of the syntax for the whole thing.

Any ideas?