Re: [xsl] suppressing only the last PI

[Abel Braaksma <abel.online@xxxxxxxxx>]

Emily.Garrett@xxxxxxxxxxx wrote:
> However, the following instruction occurs prior to the content and
> should suppress only the last PI, but is suppressing all processing
> instructions named TL_XSL and whose value beings with 'PageEnd_' and not
> just the last one:
>
>    <xsl:apply-templates
> select="node()[not(self::processing-instruction('TL_XSL')[starts-with(.,
> 'PageEnd_')][position()=last()])]"/>
>         
> Can anyone tell what is wrong with it and why it is not just suppressing
> the last one? 

Hi Emily,

Instead of not(), use position != last(). Like this:

node()[self::processing-instruction('TL_XSL')
[starts-with(.,'PageEnd_')]][position()!=last()]

Using the function not() returns true if the so-called effective boolean 
value of the argument returns false. In your case, the instruction 
inside the brackets is an xpath expression that always returns something 
when at a processing instruction. The reverse being false. This has to 
do with shifting the square brackets. This becomes apparent when 
dissecting your expression:

node()
  [not
     (
        self::processing-instruction('TL_XSL')
        [starts-with(., 'PageEnd_')]
        [position()=last()]
     )
  ]

Remove  not() and proc-instr() for clarity leaves:

1)      node()[[starts-with(., 'PageEnd_')]
2)      [position()=last()]]

Look closely at the square brackets. It says: "In the current node, select all that starts with PageEnd, and if so, select the last one you find". Effectively, you select a "group of one node and take its last node", which is always true (for nodes that start with PageEnd that is).

Reverse that, and it is always false.

If you want to use not(), you can do so, but put the brackets in the right position, and reverse only the part you want to reverse (namely, the test for last()):

node()
  [self::processing-instruction('TL_XSL')]
  [starts-with(., 'PageEnd_')]
  [not(position()=last())]

But I'd find it more readable using != instead of not(), but that's a matter of taste.

HtH,

Cheers,
-- Abel Braaksma
  http://www.nuntia.nl