I did a little research into this at MarkLogic World (where research is defined as having drinks with the engineer responsible) and found out that he changed the behavior in MarkLogic 6 so that now the comma sequence operator is no longer lazy and this expression no longer short circuits.
He made the change because he found it sped up other expression evaluation because he could use more efficient C++ internal data structures. I suggested an optimization to recognize [1] as a special case, where maybe lazy evaluation would be good to preserve because internal data structures wouldn't be needed to store the sequence. Anyway, there you have it. -jh- On Apr 10, 2013, at 8:39 PM, Tim Finney <[email protected]> wrote: > Hi Jason, > > Thanks for your reply. I have a profiler set up and its timings suggest > that the (foo(), bar(args))[1] is not short-circuiting when foo() > exists. Different timings occur when I substitute different bar(args) > functions and foo() exists. > > Best, > > Tim Finney > > On 04/11/2013 11:08 AM, Jason Hunter wrote: >> It shortcuts out like you want. >> >> Sent from my iPhone >> >> On Apr 10, 2013, at 7:53 PM, Tim Finney <[email protected]> wrote: >> >>> Hi All, >>> >>> (foo(), bar())[1] >>> >>> can be used instead of >>> >>> if (exists(foo())) then foo() else bar(). >>> >>> But does MarkLogic evaluate bar() if I use the (foo(), bar())[1] >>> construct and foo() exists? I'm interested to know because bar() could >>> take a long time to evaluate. >>> >>> Best, >>> >>> Tim Finney >>> _______________________________________________ >>> 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 _______________________________________________ General mailing list [email protected] http://developer.marklogic.com/mailman/listinfo/general
