XPath with dittoed fields?

Question

In this document if the second column is blank it means use the previous row's value.

<doc>
<table>
<tr><td>ASU</td><td>CS</td><td>3</td></tr>
<tr><td>ASU</td><td>English</td><td>3</td></tr>
<tr><td>ASU</td><td></td><td>4</td></tr>
<tr><td>ASU</td><td>French</td><td>3</td></tr>
</table>
<table>
<tr><td>CMU</td><td>CS</td><td>4</td></tr>
<tr><td>CMU</td><td>English</td><td>3</td></tr>
<tr><td>CMU</td><td>French</td><td>3</td></tr>
<tr><td>CMU</td><td></td><td>4</td></tr>
</table>
<table>
<tr><td>SDSU</td><td>English</td><td>3</td></tr>
<tr><td>SDSU</td><td></td><td>4</td></tr>
<tr><td>SDSU</td><td></td><td>5</td></tr>
<tr><td>SDSU</td><td>French</td><td>4</td></tr>
</table>
</doc>

I want rows were the second columns are English so these would be the rows:

<tr><td>ASU</td><td>English</td><td>3</td></tr>
<tr><td>ASU</td><td></td><td>4</td></tr>
<tr><td>CMU</td><td>English</td><td>3</td></tr>
<tr><td>SDSU</td><td>English</td><td>3</td></tr>
<tr><td>SDSU</td><td></td><td>4</td></tr>
<tr><td>SDSU</td><td></td><td>5</td></tr>

What would the XPath be for this?

Answer 1

(This is using XPath 1.0, there may be better solutions with more recent XPath versions).

First, you want trs, so that’s straightforward:

/doc/table/tr[...some predicate...]

The rows you want are either:

1. Those with where the second tr just contains “English”

   tr[2] = 'English'
   

2. Or those where the second tr is empty...

   tr[2] = ''
   

   and, looking at the previous sibling rows which don’t have an empty second tr...

   preceding-sibling::tr[td[2] != '']
   

   the first one ([1]) has a second tr that contains “English”

   /td[2] = 'English'
   

So combining all that, a query that gives you the desired rows is:

/doc/table/tr[td[2] = 'English'
  or (td[2] = ''
    and preceding-sibling::tr[td[2] != ''][1]/td[2] = 'English')]