10072 2010-07-03 03:04:48 +0000 math:sqrt incorrectly listed in section 4.7 "Trigonometric functions" 2010-07-13 22:20:38 +0000 1 1 1 Unclassified XPath / XQuery / XSLT Functions and Operators 3.0 Working drafts All All RESOLVED FIXED http://www.w3.org/TR/xpath-functions-11/#func-sqrt P2 trivial --- 1 dnovatchev mike davidc public-qt-comments oldest_to_newest 36553 0 dnovatchev 2010-07-03 03:04:48 +0000 math:sqrt is incorrectly listed in section 4.7 "Trigonometric functions" http://www.w3.org/TR/xpath-functions-11/#func-sqrt. It is well-known that the sqrt() function is in the clas of exponential functions -- x^a, where a = 1/2. Therefore, this function should not be listed under "trigonometric functions. Dimitre Novatchev 36556 1 mike 2010-07-03 11:29:30 +0000 Thanks for the observation. I may have been influenced by the fact that the categorization of the functions in the 2.0/1.0 spec is hopelessly illogical, and serves very little useful purpose. So at least I'm maintaining a tradition. But I'll try to do better. 36784 2 mike 2010-07-13 11:31:02 +0000 I changed the section heading to "trigonometric and other mathematical". The WG agreed this was editorial, and it has been reclassified accordingly. 36785 3 dnovatchev 2010-07-13 12:54:46 +0000 (In reply to comment #2) > I changed the section heading to "trigonometric and other mathematical". > The WG agreed this was editorial, and it has been reclassified accordingly. While this is a step in the right direction, the new heading sounds ... strange. The right solution is to have: Math Functions Exponential Trigonometric Or, even better: Math Functions and Constants Constants Functions Exponential Trigonometric While it is possible, presenting pi as a function seems weird and unnatural. Following this path we should have: math:one() math:zero() math:two() math:e() math:pi() math:ten() Then, the next step would be to forbid writing:1, 0, 2, 10, or considering this unnatural. Going in this direction would be considered wrong by most programmers and may hurt the serious adoption of XPath, XQuery and XSLT. Dimitre Novatchev 36787 4 davidc 2010-07-13 13:12:26 +0000 (In reply to comment #3) [note this is not a WG comment, just an interested observer] > While it is possible, presenting pi as a function seems weird and unnatural. Technically it's no different from having true() and false() as functions is it? 36789 5 dnovatchev 2010-07-13 13:39:32 +0000 (In reply to comment #4) > (In reply to comment #3) > [note this is not a WG comment, just an interested observer] > > While it is possible, presenting pi as a function seems weird and unnatural. > Technically it's no different from having true() and false() as functions is > it? In fact it *is* different. We do not have two literal boolean values but we do have numeric literals. While the constructors true() and false() give us a *unique* representation of all possible boolean values, introducing functions to represent numerical constants results in a *redundant* representation for some numerical values. Maybe we should be aware that there are some issues arising from redundancy? 36816 6 mike 2010-07-13 22:20:38 +0000 >The right solution is to have: Math Functions Exponential Trigonometric The section structure of this document was designed, I think, to help readers find their way around the spec, not to provide a rigorous taxonomy of the functions on offer. As a taxonomy, the hierarchy considered as a whole is indefensible; if we had a section headed "Math Functions" then it would immediately invite comment as to why a number of other functions were not present in this section. But I'm going to have to reconsider the section headings anyway as the WG has today decided to add a number of other functions such as exp, log, and pow. > Presenting pi as a function seems weird and unnatural Interesting reaction: I'm surprised by it. I have to say that to me presenting a constant as a zero-argument function seems entirely natural. The only other way of doing it that comes to mind is a built-in global variable, and that seems less consistent with other design decisions made in the language, notably the presentation of true() and false(). >While the constructors true() and false() give us a *unique* representation of all possible boolean values, introducing functions to represent numerical constants results in a *redundant* representation for some numerical values. I'm sorry, I don't follow that argument: why is it bad to have a redundant representation of some numbers? Is it bad that in "2" and "+2" we have two redundant representations of the number 2? If you don't like presenting pi() as a function, what would you prefer?