Is there an "if -then - else " statement in XPath?

Question

It seems with all the rich amount of function in xpath that you could do an "if" . However , my engine keeps insisting "there is no such function" , and I hardly find any documentation on the web (I found some dubious sources , but the syntax they had didn't work)

I need to remove ':' from the end of a string (if exist), so I wanted to do this:

if (fn:ends-with(//div [@id='head']/text(),': '))
            then (fn:substring-before(//div [@id='head']/text(),': ') )
            else (//div [@id='head']/text())

Any advice?

Is there more context to the problem? XPath is usually used to query information, not as a manipulation tool. Manipulation is usually left to something like an XSL template or a higher level language — James Conigliaro, Jun 09 '09 at 16:19
I find the w3schools tutorials easy to work with. You might start there. http://www.w3schools.com/Xpath/ — Michael Petrotta, Jun 09 '09 at 16:23
@James: This is actually a query. Don't let the if-then-else syntax fool you - it's technically a "conditional expression" and not an "if statement" - more akin to the ternary conditional operator in many programming languages. — Noldorin, Jun 09 '09 at 16:32
Similar question: http://stackoverflow.com/questions/8833225/xslt-1-0-idiom-for-ternary-if — Vadzim, Sep 11 '14 at 07:08

Tomalak · Accepted Answer · 2021-08-06T10:35:30.093

66

Yes, there is a way to do it in XPath 1.0:

concat(
  substring($s1, 1, number($condition)      * string-length($s1)),
  substring($s2, 1, number(not($condition)) * string-length($s2))
)

This relies on the concatenation of two mutually exclusive strings, the first one being empty if the condition is false (0 * string-length(...)), the second one being empty if the condition is true. This is called "Becker's method", attributed to Oliver Becker (original link is now dead, the web archive has a copy).

In your case:

concat(
  substring(
    substring-before(//div[@id='head']/text(), ': '),
    1, 
    number(
      ends-with(//div[@id='head']/text(), ': ')
    )
    * string-length(substring-before(//div [@id='head']/text(), ': '))
  ),
  substring(
    //div[@id='head']/text(), 
    1, 
    number(not(
      ends-with(//div[@id='head']/text(), ': ')
    ))
    * string-length(//div[@id='head']/text())
  )
)

Though I would try to get rid of all the "//" before.

Also, there is the possibility that //div[@id='head'] returns more than one node.
Just be aware of that — using //div[@id='head'][1] is more defensive.

edited Aug 06 '21 at 10:35

answered Jun 09 '09 at 18:14

Tomalak

332,285
67
532
628

I'll probably won't use this (I'll just upgrade the xpath engine) but this is a great trick and a great tip :) – Yossale Jun 10 '09 at 07:57
1

With numeric expressions (it you don't need to select any string) it's simpler since you can do X * number(condition) + Y * number(condition) etc., where number(condition) gives you 0 for false and 1 for true ("Boolean True is converted to 1; Boolean False is converted to 0." http://msdn.microsoft.com/en-us/library/aa926043.aspx) – George Birbilis Sep 26 '14 at 02:11
1

maybe a +1 is needed in those expressions, I had to add such at:
...

George Birbilis

Sep 26 '14 at 09:37

the becker method. I was looking for the `?` operator of c#. – LuckyLikey Feb 03 '16 at 08:36

@LuckyLikey Well, we can't all get what we want, can we? There is a nicer way to do it in XPath 2.0+. – Tomalak Feb 03 '16 at 10:39

@Tomalak Be sure, im glad I found this solution. Thanks for the hint with XPath 2.0+ – LuckyLikey Feb 03 '16 at 10:46

1

@Tomalak The link is broken, here's another link to a page discussing the method: http://blog.alessio.marchetti.name/post/2011/02/12/the-Oliver-Becker-s-XPath-method – ADJenks Mar 29 '19 at 18:40

Noldorin · Answer 2 · 2009-06-09T17:46:17.360

The official language specification for XPath 2.0 on W3.org details that the language does indeed support if statements. See Section 3.8 Conditional Expressions, in particular. Along with the syntax format and explanation, it gives the following example:

if ($widget1/unit-cost < $widget2/unit-cost) 
  then $widget1
  else $widget2

This would suggest that you shouldn't have brackets surrounding your expressions (otherwise the syntax looks correct). I'm not wholly confident, but it's surely worth a try. So you'll want to change your query to look like this:

if (fn:ends-with(//div [@id='head']/text(),': '))
  then fn:substring-before(//div [@id='head']/text(),': ')
  else //div [@id='head']/text()

I do strongly suspect this may fix it however, as the fact that your XPath engine seems to be trying to interpret if as a function, where it is in fact a special construct of the language.

Finally, to point out the obvious, insure that your XPath engine does in fact support XPath 2.0 (as opposed to an earlier version)! I don't believe conditional expressions are part of previous versions of XPath.

I don't think it suggests you *shouldn't* use parentheses, only that you don't *need* to use them. If it's complaining about "no such function," then I suspect he's not using XPath 2. The spec requires parentheses around the conditional. — Rob Kennedy, Jun 09 '09 at 17:36
@Rob: It's quite possible, but then I do state that it may not necessarily be the fix. Always worth following the example if you're trying to get something working, however. :) — Noldorin, Jun 09 '09 at 17:45
But yeah, I missed the brackets around the condition somehow. Fixed now. — Noldorin, Jun 09 '09 at 17:46

score 4 · Answer 3 · edited Sep 09 '11 at 12:59

according to pkarat's, law you can achieve conditional XPath in version 1.0.

For your case, follow the concept:

concat(substring-before(your-xpath[contains(.,':')],':'),your-xpath[not(contains(.,':'))])

This will definitely work. See how it works. Give two inputs

praba:
karan

For 1st input: it contains : so condition true, string before : will be the output, say praba is your output. 2nd condition will be false so no problems.

For 2nd input: it does not contain : so condition fails, coming to 2nd condition the string doesn't contain : so condition true... therefore output karan will be thrown.

Finally your output would be praba,karan.

score 4 · Answer 4 · answered Jun 09 '09 at 16:24

How about using fn:replace(string,pattern,replace) instead?

XPATH is very often used in XSLTs and if you are in that situation and does not have XPATH 2.0 you could use:

  <xsl:choose>
    <xsl:when test="condition1">
      condition1-statements
    </xsl:when>
    <xsl:when test="condition2">
      condition2-statements
    </xsl:when>
    <xsl:otherwise>
      otherwise-statements
    </xsl:otherwise>
  </xsl:choose>

score 3 · Answer 5 · answered Feb 18 '11 at 20:02

Personally, I would use XSLT to transform the XML and remove the trailing colons. For example, suppose I have this input:

<?xml version="1.0" encoding="UTF-8"?>
<Document>
    <Paragraph>This paragraph ends in a period.</Paragraph>
    <Paragraph>This one ends in a colon:</Paragraph>
    <Paragraph>This one has a : in the middle.</Paragraph>
</Document>

If I wanted to strip out trailing colons in my paragraphs, I would use this XSLT:

<?xml version="1.0" encoding="UTF-8"?>
<xsl:stylesheet 
    xmlns:xsl="http://www.w3.org/1999/XSL/Transform" 
    xmlns:fn="http://www.w3.org/2005/xpath-functions"
    version="2.0">
    <!-- identity -->
    <xsl:template match="/|@*|node()">
        <xsl:copy>
            <xsl:apply-templates select="@*|node()"/>
        </xsl:copy>
    </xsl:template>
    <!-- strip out colons at the end of paragraphs -->
    <xsl:template match="Paragraph">
        <xsl:choose>
            <!-- if it ends with a : -->
            <xsl:when test="fn:ends-with(.,':')">
                <xsl:copy>
                    <!-- copy everything but the last character -->
                    <xsl:value-of select="substring(., 1, string-length(.)-1)"></xsl:value-of>
                </xsl:copy>
            </xsl:when>
            <xsl:otherwise>
                <xsl:copy>
                    <xsl:apply-templates/>
                </xsl:copy>
            </xsl:otherwise>
        </xsl:choose>
    </xsl:template>
</xsl:stylesheet>

score 2 · Answer 6 · answered Mar 15 '13 at 13:41

Unfortunately the previous answers were no option for me so i researched for a while and found this solution:

http://blog.alessio.marchetti.name/post/2011/02/12/the-Oliver-Becker-s-XPath-method

I use it to output text if a certain Node exists. 4 is the length of the text foo. So i guess a more elegant solution would be the use of a variable.

substring('foo',number(not(normalize-space(/elements/the/element/)))*4)

tiritea · Answer 7 · 2017-10-10T01:07:36.847

0

Somewhat simpler XPath 1.0 solution, adapted from Tomalek's (posted here) and Dimitre's (here):

concat(substring($s1, 1 div number($cond)), substring($s2, 1 div number(not($cond))))

Note: I found an explicit number() was required to convert the bool to an int otherwise some XPath evaluators threw a type mismatch error. Depending on how strict your XPath processor is type-matching you may not need it.

edited Oct 10 '17 at 01:07

answered Oct 06 '17 at 02:08

tiritea

1,229
13
18

Is there an "if -then - else " statement in XPath?

7 Answers7

Linked