Subject: Re: [xsl] [XSLT2] setting context inside funcdefs From: Francis Norton <francis@xxxxxxxxxxx> Date: Sat, 15 Feb 2003 10:12:27 +0000 |
That's a persuasive agument but the idea of having to create an explicit context parameter for every function that needs context is still pretty ugly. Was any thought given to an alternative such as giving xsl:function a third attribute, copyContext="<boolean>", with a default of "false"? And would it be a technically satisfactory solution?You should pass all the information the function needs (unless it's available as a global variable) in the arguments to the function call.
We banned passing the context implicitly because it prevents many optimizations. For example, if you're doing lazy evaluation, then you typically have to save all aspects of the context that an expression depends on. Saving the context "just in case" the function uses it is very expensive. It also means you can't move function calls out of loops and predicates such as //a[f() = 2].
Current Thread |
---|
|
<- Previous | Index | Next -> |
---|---|---|
RE: [xsl] [XSLT2] setting context i, Michael Kay | Thread | Re: [xsl] [XSLT2] setting context i, Tobias Reif |
[xsl] Re: Reference to variable can, Dimitre Novatchev | Date | Re: [xsl] [XSLT2] setting context i, Tobias Reif |
Month |