# &&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&
  # && NEXT TODO:   
  #   (0) CROP the dates according to what we are predicting
  #   (1) Reshape and Transform
  #   (2) Then run models
  #   Find the best model for iTunes Specifically
  #   Forecast iTunes 
  #   Keeping in mind that goal is to forecast actual dollar ammount ... 
  #   ... Test robustness against forecating actual dollar ammount
  #   If good, save it, move on to spotify
  # &&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&&


setScienceIfNot("Acc_vs_Anal_2015", load=FALSE)
setSubProj("iTunesForecast")

## create suffix, 
create_objects_for_quick_col_grab(verbose=FALSE)
kCols.datestore <- c("date", "storeid")

is_prediction_run <- FALSE

if (.Pfm == "Linux") {
  ## CREATE THE TABLE
  DT.iTunes.acc_v_anal <- 
      create_DT.iTunes.acc_v_anal(
        assignTo = "DT.iTunes.acc_v_anal"
        , re_pull=TRUE
        , include_country = FALSE
        , include_trans = TRUE
        , include_label = FALSE
        , include_sc_group = FALSE
        , add_asr
        , dont_crop_at_maxDate = is_prediction_run
        , confirm = !is_prediction_run
        , expand_all_dates = TRUE
        )
  addRevenueFromWorksheet_(DT.iTunes.acc_v_anal)
} else {
  if (!exists("DT.iTunes.acc_v_anal"))
    loadFromJesus("DT.iTunes.acc_v_anal")
}

## Starting Dates
## Idenitfy the first and last non-na date for each column.  
{
  message ("These are the max/min dates per column: \n")
  print(
  rbind (
    DT.iTunes.acc_v_anal[, c(type="Starting Date", lapply(.SD[, revAndUnitsAndASRCols, with=FALSE], function(x) {date[locateFirstNonNA(x)]})), by=storeid, .SDcols=c("date", revAndUnitsAndASRCols)]
    , DT.iTunes.acc_v_anal[, c(type="Last Date", lapply(.SD[, revAndUnitsAndASRCols, with=FALSE], function(x) {date[locateLastNonNA(x)]})), by=storeid, .SDcols=c("date", revAndUnitsAndASRCols)]
    )
  )
}


# ## Make sure that each group has the full range of dates
# setkeyIfNot(DT.iTunes.acc_v_anal, "date", superset.ok=TRUE)
# DT.iTunes.acc_v_anal <- DT.iTunes.acc_v_anal[CJ_allDatesByCols(DT.iTunes.acc_v_anal, dateCol="date", universal.range=TRUE)]

## Clean up the 'months_back' column
if (all(DT.iTunes.acc_v_anal[, lunique(removeNA(months_back_TRANSFORMED)), by=date][, V1]))
    DT.iTunes.acc_v_anal[, "months_back_TRANSFORMED" := unique(removeNA(months_back_TRANSFORMED)), by=date]


## < see where the missing dates where >
cat("The missing dates for iTunes where here: \n")
print(DT.iTunes.acc_v_anal[is.na(transac_type_abbr)])

## Make Wide
DT.iTunes.acc_v_anal.wide <- make_DT.transform_wide(DT.iTunes.acc_v_anal)
warning("2015-02-10 - REMINDER TO RICK:   Remember to add the columns for Rasrp increase on itself", call.=FALSE)


stopifnot(exists("DT.iTunes.acc_v_anal.pre_transform"))
stopifnot(exists("DT.iTunes.acc_v_anal.transform_wide"))

## MODEL IT