Fix: RPL, corrected error handling: using an OBJECT expression for a TableSlot column index [5503]
Bug Number: 5503
Release notes (y/n): y
For Release Nums: 6.5

[Patrick assisted in the analysis and solution approach for this bug].

I wasn't able to reproduce the assertion failure and crash. But the problem 
was clear.  We somehow got to a place in code where we really didn't expect 
to get to; this case is prevented with higher-level checks.  But somehow, 
we did get there.  The error handling in that spot was mis-coded: returning 
a SUCCESS (false) indication and allowing the primary return result (an 
index integer) to remain un-initialized after detecting the error and 
forcing an assertion failure.  [This is in method ObjSlotLookupExpr::
getColumnIndex(...)].

I recoded the error case -- a RPL expression provided as a column index to 
a TableSlot being of an unsupported RPL expression type -- to generate an 
appropriate error message and return a FAILURE (true) indication.  (In the 
failure/crash scenario, an OBJECT type is encountered, which cannot be used 
as a TableSlot column index).

A trivial bug model I created to test this bug (bug5503.mdl.gz, RiverWare 
6.4.9) has an Initialization rule defined as:

  1:  FOR ( DATETIME t IN @"Start Timestep" TO @"Finish Timestep" ) DO
  2:     DataObj1.AggSeries [ t, 1 ] = DataObj1.Series [ t ]
  3:  END FOR
  4:  %"Reach1" . "Outflow" [ @"Start Timestep" ] = 100.000 "cfs"
  5:  %"Reach2" . "Outflow" [ @"Start Timestep" ] = 200.000 "cfs"


I attempted to create the error case by copying either the %"Reach1" or 
%"Reach2" expressions, and pasting that over the '1' column index in the 
assignment statement inside the FOR statement.  Doing so generates this 
warning: "Type mismatch between selection and paste buffer. The resulting 
Rule will be invalid. Are you sure you want to paste?" [Yes] [No].  
Clicking yes, and running the model catches the problem at a higher level 
-- i.e. it doesn't get to the failure/crash code indicated in the bug report.

I tested the message generation by (temporarily) causing the error to be 
(falsely) detected during a valid TableSlot lookup.  Here is the generated 
message:

   Context: RULE: Rule
   Message: Evaluation of the left-hand side of the Assignment statement 
     failed for the following reason(s): The specification's type {NUMERIC} 
     used as a column index for slot "DataObj1.AggSeries" is not valid for 
     TableSlots. This occurred at the following location within the 
     expression: $ "DataObj1.AggSeries" ["t", 1.00000000].

(This jury rigged message is reporting a valid state as an error -- 
just to test the error message generation code).

---