GEOMESA-1111 Return null for non-existant field names

[ddseapy]

Signed-off-by: David Seapy ddseapy@ccri.com

[elahrvivaz]

i wonder if we could code this a little smarter - we're going to be calling ctx.indexOf for every eval if the idx is -1. Nothing is springing to mind though...

[elahrvivaz]

well actually we could just make it default to -2 - but we're still doing 2 checks every time. probably doesn't really matter...

[elahrvivaz]

we could wrap the get in a try/catch, which would be basically free for the normal case, but make the null case way more expensive...

[anthonyccri]

How about this?

     var doEval: EvaluationContext => Any = null
     override def eval(args: Array[Any])(implicit ctx: EvaluationContext): Any = {
       if (idx == -1) {
         idx = ctx.indexOf(n)
         doEval = 
           if(idx < 0)  _  => null
           else         ec => ec.get(idx)
       }
       doEval(ctx)
     }

[ddseapy]

I can see how this would work if default and the "idx ==" check where both -2 instead of -1. Then it would only do ctx.indexOf on the first eval() execution.

[ddseapy]

or if "idx == -1" was "doEval == null", which is a bit clearer, basically saying "if we don't have a doEval defined then define it")

[ddseapy]

would this work to avoid the check altogether?:

var doEval: EvaluationContext => Any = { ctx =>
      val idx = ctx.indexOf(n)
      doEval =
        if(idx < 0)  _  => null
        else         ec => ec.get(idx)
      doEval(ctx)
    }
    override def eval(args: Array[Any])(implicit ctx: EvaluationContext): Any = doEval(ctx)

[elahrvivaz]

looks reasonable

[elahrvivaz]

hehe - @ddseapy you used master for your working branch?
merged as e5ecf99