Identifying nodes on position relative to grandparent - html

Question

I have the following HTML table:
tab2 <- '<table>
<thead>
<tr>
<th rowspan="2">a</th>
<th>b</th>
<th colspan="2" rowspan="2">c</th>
</tr>
<tr>
<td></td>
</tr>
</thead>
<tbody>
<tr>
<td>1</td>
<td>2</td>
<td>3</td>
<td></td>
</tr>
</tbody>
</table>'
It has three rows, the first two are header information, the last one is the body. The goal is to extract the header information using only the header row position relative to the table node (1+2), i.e. without having to pay attention to whether the header nodes have a thead parent or not.
I tried
//tr[position() < 3]
doesn't work because position() works relative to parent node thead and tbody.
I am using R with the XML package (which uses XPath 1.0). This is what I get when I use above XPath
xpathSApply(tab2, "//tr[position() < 3]")
[[1]]
<tr><th rowspan="2">a</th>
<th>b</th>
<th colspan="2" rowspan="2">c</th>
</tr>
[[2]]
<tr><td/>
</tr>
[[3]]
<tr><td>1</td>
<td>2</td>
<td>3</td>
<td/>
</tr>
I get all three rows. Which makes sense according to how I understand position(). It works relative to its parent.
Context
I am writing a function that allows users to parse HTML tables with the R programming language and assemble an R data structure from it. The function allows users to pass a numeric value for which rows provide header information and which body information. For the above table, users should be able to say row 1 and row 2 (in the entire table) provide header information. I need to process this input so it works on HTML tables unconditional on whether this table makes use of thead and tbody elements, or not. The problem with
//tr[position() < 3]
is that it also returns the body row (third row). Hope this makes it clear(er).

Answer 1

Use the following XPath expression:
/table//tr[count(preceding::tr) < 2]
It does not care whether a certain tr is inside thead or not. It just considers tr element that are preceded by zero or one other tr element. The result is the following:
<tr>
<th rowspan="2">a</th>
<th>b</th>
<th colspan="2" rowspan="2">c</th>
</tr>
-----------------------
<tr>
<td/>
</tr>
Caveat: This simple approach only works if there is only one table in the HTML document. But as long as you are working with exactly this HTML snippet , it suffices.

Answer 2

This expression will work for a document with any number of tables.
//table/descendant::tr[position() < 3]
By using the descendant forward axis, the [position() < 3] subscript will select the first and second node in the set of tr descendants of the table (rather than finding their position relative to their parent node, as with //tr in your question).
http://jsfiddle.net/uutavwvk/1/