Hello all,

I have another question related to the new statistics abilities when querying features through the REST API in 10.1. It seems that if I use a field to group the resulting statistics values, and if any records in the feature layer or table contain null values in the specified field, I get a valid-looking response, but with an empty 'features' value. When no values are null, here's an example of what the response looks like when I ask for the stats to be grouped by a column named "integer_attribute":

Code:
{
 "displayFieldName": "",
 "fieldAliases": {
  "integer_attribute": "integer_attribute",
  "total": "total"
 },
 "fields": [
  {
   "name": "integer_attribute",
   "type": "esriFieldTypeOID",
   "alias": "integer_attribute"
  },
  {
   "name": "total",
   "type": "esriFieldTypeDouble",
   "alias": "total"
  }
 ],
 "features": [
  {
   "attributes": {
    "integer_attribute": 0,
    "total": 8659
   }
  },
  {
   "attributes": {
    "integer_attribute": 1,
    "total": 7639
   }
  }
 ]
}
However, if any of the records contain a null value for the "integer_attribute", I get the following response:

Code:
{
 "displayFieldName": "",
 "fieldAliases": {
  "integer_attribute": "integer_attribute",
  "total": "total"
 },
 "fields": [
  {
   "name": "integer_attribute",
   "type": "esriFieldTypeOID",
   "alias": "integer_attribute"
  },
  {
   "name": "total",
   "type": "esriFieldTypeDouble",
   "alias": "total"
  }
 ],
 "features": [
 
 ]
}
Is this an outcome I should have expected? I was hoping that the result would look more like the first response, but with a third entry that contains the count of records where the integer_attribute is null. However, if this operation is not allowed when the group-by field contains null values, it would be helpful to get some kind of error that indicates what went wrong.

If I add a where clause to the query that excludes null values, I'll get results again...so I can still get results for non-null records as long as I do that. However, it would be helpful if I actually could get the total null values, because no data is also a valid type of information (i.e., it helps to reveal information that is not yet known or incomplete). The only workaround I can think of is to request the total statistics (un-grouped) separately, then get the grouped stats for the non-null records...then I'd have to loop through the results and subtract the totals from the un-grouped total...with the remainder being the number of null results. Can anyone suggest a better approach?