Subject: Re: [xsl] When to use conditional constructions?|
From: Michael Kay <mike@xxxxxxxxxxxx>
Date: Sun, 30 Mar 2014 18:28:32 +0100
This raises the question as to how much of the logic to put in XSLT instructions and how much in XPath expressions. Dimitre, I think, uses a style of coding in which a great deal of logic goes at the XPath level, leading to XPath expressions of 20 lines or more. Most other XSLT users I've come across seem to try and avoid this. I suspect the reasons are mainly to do with the difficulty of formatting it nicely, and poor diagnostics if you get it wrong (the XSLT processor will typically only give you the line number of the XSLT instruction). Personally I'm happy to go to three or four lines for an XPath expression, but beyond that I start to feel uncomfortable with it; can't really explain why. So I'm comfortable using XPath conditionals in roughly the same contexts as I would use the ternary conditional of C or Java: maximum of one line per operand. I do have a strong preferance for multiple template rules over xsl:choose. I think this is mainly because it makes it easier to evolve the code as it develops over time. Though a strong downside is that it can be very hard for the reader of the code to work out which templates are going to fire under which circumstances. At the heart of this is that XSLT is a two-language system; it isn't fully composable, in that XSLT constructs can't be called from XPath constructs. This creates a strong incentive to break your code up into functions, which can be called easily enough from either context. Michael Kay Saxonica On 30 Mar 2014, at 13:25, David Rudel <fwqhgads@xxxxxxxxx> wrote: > On Sun, Mar 30, 2014 at 1:07 AM, Dimitre Novatchev <dnovatchev@xxxxxxxxx> wrote: > >> This is the full proof that XSLT conditional instructions can be >> eliminated in any version of XSLT. >> >> BTW, I have quite a lot of experience writing complex transformations >> without any XSLT conditional instructions. :) >> >> Cheers, >> Dimitre Novatchev >> > > Dimitre, as typical, your posts are very insightful. I'm now wondering > how much I should consider avoiding conditionals in my own work. > > This provoked a general question I have for experiences XSLT > programmers who frequently use <xsl:apply-templates> or some other > method to avoid what would normally (in a more imperative regime) be > done using conditional constructions. > > Have you found that, in general, it is best to avoid conditional > constructs? If so, could you share what advantages you have found by > doing so? > > Does this behavior extend to XPath's "if" statement as well, or do you > see that as a different beast? (In other words, do you find that the > advantages you obtain by avoiding XSL's conditionals do not apply as > much when considering cases that can be addressed using XPath > instead?) > > Are there specific cases where conditional constructs should be > favored? (E.g., in <xsl:iterate> to allow Tail Call optimization?) > > As I have confessed in the past, for my work the thing I value most > about XSLT is its support of XPath. I still tend to use imperative > programming, so my scripts tend to look like the XSLT equivalent of a > Java program that uses a single class. > > -David > > -- > > "A false conclusion, once arrived at and widely accepted is not > dislodged easily, and the less it is understood, the more tenaciously > it is held." - Cantor's Law of Preservation of Ignorance.