Benoit, the code that makes an http call then uses the result in an emit() call, which only makes sense as part of a map function.
B. On 13 November 2013 15:57, Benoit Chesneau <[email protected]> wrote: > On Wed, Nov 13, 2013 at 4:50 PM, Robert Newson <[email protected]> wrote: > >> what Jens said. We prevent this deliberately. Crowbar it back in at >> your own risk. >> >> B. >> > > The code pasted in the initial mail isn't about map/reduce but a show. Imo > if the user want to take more risk I don't see why not if it's feasible > technically without introducing any changes to the defaults ;) Especially > in shows and list. There are a lot of good reason to do that. we could even > provides some limitation on what could be cal there if i red the action.js > from 2008 well. > > - benoit > > > >> >> >> On 13 November 2013 15:45, Jens Alfke <[email protected]> wrote: >> > >> > On Nov 13, 2013, at 4:48 AM, Hank Knight <[email protected]> wrote: >> > >> >> function(doc, req) { >> >> var http = new XMLHttpRequest(); >> >> http.open( "GET", "http://www.w3schools.com/xml/note.xml";, false ); >> >> http.send( null ); >> >> emit("xyz",http.responseText); >> >> } >> > >> > Wait, is that supposed to be a map function? Whether or not this is >> technically feasible, it’s an incorrect use of map/reduce. >> > >> > It’s critical that a map function be a ‘pure' function with no access to >> outside state. That is, every time the function is called with the same >> input, it must produce the same output. Otherwise you won’t get a >> self-consistent index. >> > >> > —Jens >>
