HTML parsing with XPath: flattened hierarchical data - html

Question

My target HTML is a flattened table of elements with 2 levels of data defined by class attribute:
<tr>
<td class="type">Type 1</td>
</tr>
<tr>
<td class="name">name1</td>
<td class="year">1970</td>
<td class="rank">1</td>
</tr>
<tr>
<td class="name">name2</td>
<td class="year">1982</td>
<td class="rank">3</td>
</tr>
Goal is parse out list of name, year, rank elements, which I accomplish with these xpath expressions:
//td[#class = 'name']/text()
//td[#class = 'year']/text()
//td[#class = 'rank']/text()
Each element is under immediately preceding
<tr>
<td class="type">Type 1</td>
</tr>
I would like to have "Type 1" assigned to each element parsed above. It could be separate list of the same length. Of course, my target HTML contains many such elements within the same 2-level hierarchy: type - element (name, year, rank).

Answer 1

The following rather clumsy xpath concatenates the closest, previous #type td to the name td matched above.
concat(//td[#class = 'name']/preceding::td[#class='type'][1]/text(), '-',
//td[#class = 'name']/text())
This probably makes more sense when shown in the following xsl
<xsl:for-each select="//td[#class='name']">
<Name>
<xsl:value-of select="concat(preceding::td[#class='type'][1]/text(),
'-', ./text())" />
</Name>
</xsl:for-each>
Applied to the following xml
<xml>
<tr>
<td class="type">Type 1</td>
</tr>
<tr>
<td class="name">name1</td>
<td class="year">1970</td>
<td class="rank">1</td>
</tr>
<tr>
<td class="name">name2</td>
<td class="year">1982</td>
<td class="rank">3</td>
</tr>
<tr>
<td class="type">Type 2</td>
</tr>
<tr>
<td class="name">name3</td>
<td class="year">1971</td>
<td class="rank">2</td>
</tr>
<tr>
<td class="name">name4</td>
<td class="year">1983</td>
<td class="rank">4</td>
</tr>
</xml>
With the result
<Name>Type 1-name1</Name>
<Name>Type 1-name2</Name>
<Name>Type 2-name3</Name>
<Name>Type 2-name4</Name>

Answer 2

Solution 1
First, find the td elements of interest. For example, the name tds with the following pseudo-code:
name_tds = doc.evalXPath("//td[#class = 'name']")
Then you can find the corresponding type td using a name td as context node like this:
type_td = name_td.evalXPath("../preceding-sibling::tr[td[#class = 'type']][1]/td")
Solution 2
Simply iterate all the tds and remember the last type you found. Pseudo-code:
foreach (td in doc.evalXPath("//td") {
class = td.getAttribute("class");
if (class == "type") {
type = td.textContent();
}
else if (class == "name") {
name = td.textContent();
println("type: " + type + ", name: " + name);
}
// Same for year and rank.
}