How to use the "translate" Xpath function on a node-set

I have an XML document that contains items with dashes I'd like to strip

e.g.

<xmlDoc>
   <items>
      <item>a-b-c</item>
      <item>c-d-e</item>
   <items>
</xmlDoc>

I know I can find-replace a single item using this xpath

/xmldoc/items/item[1]/translate(text(),'-','')

Which will 开发者_如何转开发return

"abc"

however, how do I do this for the entire set?

This doesn't work

/xmldoc/items/item/translate(text(),'-','')

Nor this

translate(/xmldoc/items/item/text(),'-','')

Is there a way at all to achieve that?

I know I can find-replace a single item using this xpath

/xmldoc/items/item[1]/translate(text(),'-','')

Which will return

"abc"

however, how do I do this for the entire set?

This cannot be done with a single XPath 1.0 expression.

Use the following XPath 2.0 expression to produce a sequence of strings, each being the result of the application of the translate() function on the string value of the corresponding node:

/xmlDoc/items/item/translate(.,'-', '')

The translate function accepts in input a string and not a node-set. This means that writing something like:

"translate(/xmlDoc/items/item/text(),'-','')"

or

"translate(/xmlDoc/items/item,'-','')"

will result in a function call on the first node only (item[1]).

In XPath 1.0 I think you have no other chances than doing something ugly like:

"concat(translate(/xmlDoc/items/item,'-',''),
 translate(/xmlDoc/items/item[2],'-',''))"

Which is privative for a huge list of items and returns just a string.

In XPath 2.0 this can be solved nicely using for expressions:

  "for $item in /xmlDoc/items/item  
    return replace($item,'-','')"

Which returns a sequence type:

abc cde

PS Do no confuse function calls with location paths. They are different kind of expressions, and in XPath 1.0 can not be mixed.

here is yet anther example, running it against chrome developer tools, in prepartion for a selenium test.

$x("//table[@id='sometable_table']//tr[1=1 and ./td[2=2 and position()=2 and .//*[translate(text(), ',', '') ='1001']  ]  ]/td[position()=2]")

Essentially the the data sometable_table has a column containing numbers that appear localized. For example 1001 would appear as 1,001. With the above you have somewhat nasty xpath expression.

Where first you select all table rows. Then you focus on the data of the position 2 table data for the row. Then you go deeper into the contents of the position=2 table data expand the data on the cell until you find any node whose text after string replacement is 1001. Finally you ask for the table at position 2 to be returned.

But since all your main filters are at the table row level, you could be doing additional filters at table data columns at other positions as well, if you need to find the appropriate table row that has content (A) on a cell column and content (B) on a different column.

NOTE: It was actually quite nasty to write this, because intuitively, we all google for XPATH replace string. So I was getting furstrated trying to use xpath replace until i realized chrome supports XPATH 1.0. In xpath 1.0 the string functions that exist are different from xpath 2.0, you need to use this translate function.

See reference: http://www.edankert.com/xpathfunctions.html