Aggregation is optimized incorrectly

[tomjaguarpaw]

If I have a table A, and I sum a numeric column while grouping by a string column (say) the string column in the resulting table B has the property that no string appears twice. If I then project out the string column to get C then C contains all the distinct strings from A.

However, HaskellDB's query optimizer seems to get rid of the aggregation entirely without noticing the that the strings in C should be distinct.

import Database.HaskellDB
import Database.HaskellDB.PrimQuery
import Database.HaskellDB.Query
import Database.HaskellDB.DBLayout

data Name' = Name'
instance FieldTag Name' where
  fieldName _ = "name"

data Apples = Apples
instance FieldTag Apples where
  fieldName _ = "age"

name :: Attr Name' String
name = mkAttr Name'

apples :: Attr Apples Int
apples = mkAttr Apples

applesTable :: Table (RecCons Name' (Expr String) (RecCons Apples (Expr Int) RecNil))
applesTable = Table "apples" [ ("name", AttrExpr "namecol")
                             , ("apples", AttrExpr "applecol")]

totalApples :: Query (Rel (RecCons Name' (Expr String) (RecCons Apples (Expr Int) RecNil)))
totalApples = do
  t <- table applesTable
  project (name << t!name #
           apples << _sum(t!apples))

justNames :: Query (Rel (RecCons Name' (Expr String) RecNil))
justNames = do
  t <- totalApples
  project (name << t!name)


*Main> putStrLn $ showSql justNames
SELECT namecol as name
FROM apples as T1

[tomjaguarpaw]

The culprit is here, on deleted lines 85 to 103. (I think the change to removeD of a Group is also broken.)

09791ae#diff-ec78e70ce0ca00fde0c3c7c90b7c74c1L85