Yeah, map:map is also a mutator extension and is another way to create side effects.
One thing to keep in mind is that causing side effects can defeat optimization (in theory, I can't speak to any specific optimizations ML does in specific cases). For example, a simple FLWOR that iterates over a sequence must run each iteration sequentially if a variable is modified within the body. But if no variables are modified, other than the "let" bindings, then the XQuery processor may be able to parallelize execution of the FLWOR. Say you're on a 16 core machine and are iterating over a sequence of length 10. It's possible that all 10 tuples could be processed concurrently. XQuery just needs to make sure to assemble the resulting values in the correct order - it's not necessary to *process* in sequence if there are no side effects. Using an xdmp:set in such a FLWOR could (theoretically) cause your code to take ten times longer to run. On Sep 15, 2011, at 4:36 PM, Colleen Whitney wrote: > Yes, it's an extension, but so are try/catch blocks. > > You can bind a map to a global variable (in which case it can be referenced > and modified by any function), or create and pass maps through functions as a > local variable just like any other XQuery primitive type, as Erik suggests. > > You can use xdmp:set to modify a global variable, but as Ron noted it does > sidestep the functional paradigm and is a very sharp tool, it's very easy to > shoot yourself in the foot if you've got various functions setting it > concurrently. > > --Colleen > > ________________________________ > From: [email protected] > [[email protected]] On Behalf Of Tim Meagher > [[email protected]] > Sent: Thursday, September 15, 2011 8:03 AM > To: 'General MarkLogic Developer Discussion' > Subject: Re: [MarkLogic Dev General] How to effectively pass variables > by reference to an xquery function? > > Hi Erik, > > It certainly appears that using a map would effectively allow me to create > global variables in the map hash that can be accessed within functions, but > if I understand Ron's objections, they have to do with using an xquery > extension in the form of xdmp:set. However if I understand correctly, map > is also a MarkLogic extension, in which case I would merely be replacing the > use of one MarkLogic extension with another. > > I confess that I regularly use xdmp:set as a simple way of modifying local > variables, but the question I have is if using xdmp:set to modify global > variables in a function is a supported feature with deterministic results. > > BTW, isn't the map global in scope? Why or how would you pass it into a > function rather than just directly accessing it in the function? > > Thank you! > > Tim > > -----Original Message----- > From: [email protected] > [mailto:[email protected]] On Behalf Of Erik Hennum > Sent: Thursday, September 15, 2011 10:33 AM > To: General MarkLogic Developer Discussion > Subject: Re: [MarkLogic Dev General] How to effectively pass variables by > reference to an xquery function? > > Hi, Tim: > > Distinguishing the function output from function errors seems like a good > idea (witness stdout and stderr). > > One possibility is to create a map at the top of the call stack and pass the > map through to called functions. Called functions can append to an error > key in the map. > > The benefit is having strongly typed function returns and recording errors > or warnings without the global side effects that Ron sensibly cautions > against. > The cost is the map creation, but that's minimal if you create one map per > request. > > > Hoping that helps, > > > Erik Hennum > ________________________________________ > From: [email protected] > [[email protected]] On Behalf Of Ron Hitchens > [[email protected]] > Sent: Thursday, September 15, 2011 7:12 AM > To: General MarkLogic Developer Discussion > Subject: Re: [MarkLogic Dev General] How to effectively pass variables by > reference to an xquery function? > > XQuery is designed as a functional language. One of the design > principles of a functional language is that there should not be side > effects. One way of insuring no side effects is to not let variables change > their values. > You never "set" a variable, you bind a value to it that never changes > thereafter. > > So that means you can't modify a variable through a reference because you > can't modify variables. > > You can in fact mutate variables with xdmp:set. But that's a MarkLogic > extension that is not part of XQuery proper. There are uses for xdmp:set, > but you should always think twice before using it. Inducing side effects in > a side-effect free language is definitely a sharp tool. > > I would recommend returning an XML fragment with all the values you want > your function to produce. > > On Sep 15, 2011, at 2:45 PM, Tim Meagher wrote: > >> Hi Folks, >> >> I am calling a function that returns a value but I also want the >> function to set a status as well. I'm wondering what is the best way >> to pass by reference? I can return an xml structure that contains the >> return value and a status, but I'm also wondering about just using >> xdmp:set() to set a global variable in a function. > _______________________________________________ > General mailing list > [email protected] > http://developer.marklogic.com/mailman/listinfo/general > > _______________________________________________ > General mailing list > [email protected] > http://developer.marklogic.com/mailman/listinfo/general > _______________________________________________ > General mailing list > [email protected] > http://developer.marklogic.com/mailman/listinfo/general --- Ron Hitchens {mailto:[email protected]} Ronsoft Technologies +44 7879 358 212 (voice) http://www.ronsoft.com +1 707 924 3878 (fax) Bit Twiddling At Its Finest "No amount of belief establishes any fact." -Unknown _______________________________________________ General mailing list [email protected] http://developer.marklogic.com/mailman/listinfo/general
