13745 2011-08-10 09:57:19 +0000 [XQuery 3.0] "deterministic" and "nondeterministic" 2011-09-06 19:15:56 +0000 1 1 1 Unclassified XPath / XQuery / XSLT XQuery 3.0 Working drafts PC Windows NT RESOLVED FIXED P2 normal --- 1 mike jonathan.robie public-qt-comments oldest_to_newest 54587 0 mike 2011-08-10 09:57:19 +0000 The XQuery specification often seems to use the phrase "nondeterministic" where "implementation-dependent" would be more appropriate. For example: 2.4.3: "The effective boolean value of a sequence that contains at least one node and at least one atomic value may be nondeterministic in regions of a query where ordering mode is unordered." 3.2.1: "In a region of a query where ordering mode is unordered, the result of a numeric predicate is nondeterministic, as explained in 3.10 Ordered and Unordered Expressions." I note also that "nondeterministic" is sometimes hyphenated, sometimes not; and that we sometimes say "is nondeterministic", sometimes "may be nondeterministic", with no apparent difference in intended meaning. I suspect our intent is that wherever the result of an expression is implementation-defined or implementation-dependent (i.e. varies from one implementation to another) then it is also implementation-defined or -dependent whether it is deterministic (i.e. varies from one execution to another). 56433 1 jonathan.robie 2011-09-06 19:15:37 +0000 (In reply to comment #0) > I suspect our intent is that wherever the result of an expression is > implementation-defined or implementation-dependent (i.e. varies from one > implementation to another) then it is also implementation-defined or -dependent > whether it is deterministic (i.e. varies from one execution to another). This is not a general principle. For instance, documentation order is stable even when it is implementation-defined. > The XQuery specification often seems to use the phrase "nondeterministic" where > "implementation-dependent" would be more appropriate. For example: > > 2.4.3: "The effective boolean value of a sequence that contains at least one > node and at least one atomic value may be nondeterministic in regions of a > query where ordering mode is unordered." We should either eliminate the term or define it. It looks like most users are in non-normative notes. The following use is normative: <quote> In addition to ordered and unordered expressions, XQuery provides a function named fn:unordered that operates on any sequence of items and returns the same sequence in a nondeterministic order. </quote> The other uses all seem to refer to this one, directly or indirectly. I would say that in this case the result should be implementation-defined, and is not required to be stable. In the non-normative notes, I would change "nondeterministic" to "implementation-defined". I do not believe this changes any requirements on processors, so it's probably an editorial change. I'll go ahead and make it, please reopen if I got this wrong. > I note also that "nondeterministic" is sometimes hyphenated, sometimes not; and > that we sometimes say "is nondeterministic", sometimes "may be > nondeterministic", with no apparent difference in intended meaning. Yes, it looks like the hyphenated spelling is used only when discussing annotations. I just eliminated these hyphens. And I changed "may be nondeterministic" to "is implementation-dependent" in the two places where I saw it.