“Tricky XSLT recursive tree transformation” – Forum Thread – Discuss

- qnn
- 23 Sep 11, 12:05 pm
- Comment #1

Hi!

I have a loosely structured XHTML data and I need to convert it to properly structured XML to import to a Symphony section.

Here's the example:

<tbody>
<tr>
    <td class="header"><img src="http://www.abc.com/images/icon_apples.gif"/><img src="http://www.abc.com/images/flag/portugal.gif" alt="Portugal"/> First Grade</td>
</tr>
<tr>
    <td>Green</td>
    <td>Round shaped</td>
    <td>Tasty</td>
</tr>
<tr>
    <td>Red</td>
    <td>Round shaped</td>
    <td>Bitter</td>
</tr>
<tr>
    <td>Pink</td>
    <td>Round shaped</td>
    <td>Tasty</td>
</tr>
<tr>
    <td class="header"><img src="http://www.abc.com/images/icon_strawberries.gif"/><img src="http://www.abc.com/images/flag/usa.gif" alt="USA"/> Fifth Grade</td>
</tr>
<tr>
    <td>Red</td>
    <td>Heart shaped</td>
    <td>Super tasty</td>
</tr>
<tr>
    <td class="header"><img src="http://www.abc.com/images/icon_bananas.gif"/><img src="http://www.abc.com/images/flag/congo.gif" alt="Congo"/> Third Grade</td>
</tr>
<tr>
    <td>Yellow</td>
    <td>Smile shaped</td>
    <td>Fairly tasty</td>
</tr>
<tr>
    <td>Brown</td>
    <td>Smile shaped</td>
    <td>Too sweet</td>
</tr>

I am trying to achieve following structure:

    <data>
    <entry>
        <type>Apples</type>
        <country>Portugal</country>
        <rank>First Grade</rank>
        <color>Green</color>
        <shape>Round shaped</shape>
        <taste>Tasty</taste>
    </entry>
    <entry>
        <type>Apples</type>
        <country>Portugal</country>
        <rank>First Grade</rank>
        <color>Red</color>
        <shape>Round shaped</shape>
        <taste>Bitter</taste>
    </entry>
    <entry>
        <type>Apples</type>
        <country>Portugal</country>
        <rank>First Grade</rank>
        <color>Pink</color>
        <shape>Round shaped</shape>
        <taste>Tasty</taste>
    </entry>
    <entry>
        <type>Strawberries</type>
        <country>USA</country>
        <rank>Fifth Grade</rank>
        <color>Red</color>
        <shape>Heart shaped</shape>
        <taste>Super</taste>
    </entry>
    <entry>
        <type>Bananas</type>
        <country>Congo</country>
        <rank>Third Grade</rank>
        <color>Yellow</color>
        <shape>Smile shaped</shape>
        <taste>Fairly tasty</taste>
    </entry>
    <entry>
        <type>Bananas</type>
        <country>Congo</country>
        <rank>Third Grade</rank>
        <color>Brown</color>
        <shape>Smile shaped</shape>
        <taste>Too sweet</taste>
    </entry>
</data>

Firstly I need to extract the fruit type from the tbody/tr/td/img[1]/@src, secondly the country from tbody/tr/td/img[2]/@alt attribute and finally the grade from tbody/tr/td itself.

Next I need to populate all the entries under each category while including those values (like shown above).

But... As you can see, the the data I was given is very loosely structured. A category is simply a td and after that come all the items in that category. To make the things worse, in my datasets, the number of items under each category varies between 1 and 100...

I've tried a few approaches but just can't seem to get it. Any help is greatly appreciated. I know that XSLT 2.0 introduces xsl:for-each-group, but I am limited to XSLT 1.0.

- michael-e
- 23 Sep 11, 6:14 pm
- Comment #2

One way to achieve this is: Count the elements up to the next header line.

<xsl:template match="//tbody">
    <xsl:for-each select="tr[td/@class='header']">
        <xsl:variable name="elements-number" select="count(following-sibling::*[not(self::tr[td/@class='header'])]) - count(following-sibling::tr/following-sibling::*[not(self::tr[td/@class='header'])])"/>
        <xsl:apply-templates select="following-sibling::*[position() &lt;= $elements-number]" mode="element">
            <xsl:with-param name="rank" select="."/>
        </xsl:apply-templates>
    </xsl:for-each>
</xsl:template>

<xsl:template match="tr" mode="element">
    <xsl:param name="rank"/>
    <entry>
        <color>
            <xsl:value-of select="td[1]"/>
        </color>
        <shape>
            <xsl:value-of select="td[2]"/>
        </shape>
        <taste>
            <xsl:value-of select="td[3]"/>
        </taste>
        <rank>
            <xsl:value-of select="rank"/>
        </rank>
    </entry>
</xsl:template>

I have not included the type or country, but that shouldn't be too hard.

- michael-e
- 23 Sep 11, 6:16 pm
- Comment #3

Sorry, it's buggy. It only outputs 3 elements. Wait a minute...

- michael-e
- 23 Sep 11, 6:18 pm
- Comment #4

I was counting wrong. Here you are:

<xsl:template match="//tbody">
    <xsl:for-each select="tr[td/@class='header']">
        <xsl:variable name="elements-number" select="
            count(following-sibling::*[not(self::tr[td/@class='header'])]) 
            - count(following-sibling::tr[td/@class='header']/following-sibling::*[not(self::tr[td/@class='header'])])"/>
        <xsl:apply-templates select="following-sibling::*[position() &lt;= $elements-number]" mode="element">
            <xsl:with-param name="rank" select="normalize-space(td/text())"/>
        </xsl:apply-templates>
    </xsl:for-each>
</xsl:template>

<xsl:template match="tr" mode="element">
    <xsl:param name="rank"/>
    <entry>
        <color>
            <xsl:value-of select="td[1]"/>
        </color>
        <shape>
            <xsl:value-of select="td[2]"/>
        </shape>
        <taste>
            <xsl:value-of select="td[3]"/>
        </taste>
        <rank>
            <xsl:value-of select="$rank"/>
        </rank>
    </entry>
</xsl:template>

[EDIT]: Added normalize-space()...

- michael-e
- 23 Sep 11, 6:24 pm
- Comment #5

Just a short explanation:

For each tr element with a class 'header', count all following tr elements which don't have the class 'header'
Subtract the count of elements following the next tr element with a class 'header' (and not having a class 'header')
Result: number of elements up to the next "header"

I am sure there are simpler ways, but this technique saved my day once. :-)

- michael-e
- 23 Sep 11, 7:07 pm
- Comment #6

If you want a more "expert-like" solution, you may use IDs:

<xsl:template match="//tbody">
    <data>
        <xsl:apply-templates select="tr[td/@class='header']" mode="header"/>
    </data>
</xsl:template>

<xsl:template match="tr" mode="header">
    <xsl:variable name="next-header" select="generate-id(following-sibling::tr[td/@class='header'])"/>
    <xsl:apply-templates select="following-sibling::tr[not(generate-id(.) = $next-header or preceding-sibling::tr[generate-id(.) = $next-header])]" mode="element">
        <xsl:with-param name="rank" select="normalize-space(td/text())"/>
    </xsl:apply-templates>
</xsl:template>

<xsl:template match="tr" mode="element">
    <xsl:param name="rank"/>
    <entry>
        <color>
            <xsl:value-of select="td[1]"/>
        </color>
        <shape>
            <xsl:value-of select="td[2]"/>
        </shape>
        <taste>
            <xsl:value-of select="td[3]"/>
        </taste>
        <rank>
            <xsl:value-of select="$rank"/>
        </rank>
    </entry>
</xsl:template>

[EDIT]: Added initial apply-templates.

- michael-e
- 23 Sep 11, 7:09 pm
- Comment #7

The above only selects elements which are not the "next header" or following the "next header". Same result.

Symphony.

Tricky XSLT recursive tree transformation

Search

Server Requirements

Symphony.

Tricky XSLT recursive tree transformation

Search

You are looking at page 1 of 1

Server Requirements

Sign in