return (NA)
                        }
 
-                       # Determine indices of no-NAs days preceded by no-NAs yerstedays
-                       tdays = .getNoNA2(data, max(today-memory,2), today-1)
-
                        # Get optional args
                        local = ifelse(hasArg("local"), list(...)$local, TRUE) #same level + season?
                        simtype = ifelse(hasArg("simtype"), list(...)$simtype, "none") #or "endo", or "exo"
+                       opera = ifelse(hasArg("opera"), list(...)$opera, FALSE) #operational mode?
+
+                       # Determine indices of no-NAs days preceded by no-NAs yerstedays
+                       tdays = .getNoNA2(data, max(today-memory,2), ifelse(opera,today-1,data$getSize()))
+                       if (!opera)
+                               tdays = setdiff(tdays, today) #always exclude current day
+
+                       # Shortcut if window is known
                        if (hasArg("window"))
                        {
-                               return ( private$.predictShapeAux(data,
-                                       tdays, today, predict_from, horizon, local, list(...)$window, simtype, TRUE) )
+                               return ( private$.predictShapeAux(data, tdays, today, predict_from, horizon,
+                                       local, list(...)$window, simtype, opera, TRUE) )
                        }
 
                        # Indices of similar days for cross-validation; TODO: 20 = magic number
                        cv_days = getSimilarDaysIndices(today, data, limit=20, same_season=FALSE,
-                               days_in=tdays)
+                               days_in=tdays, operational=opera)
 
                        # Optimize h : h |--> sum of prediction errors on last N "similar" days
                        errorOnLastNdays = function(window, simtype)
                                {
                                        # mix_strategy is never used here (simtype != "mix"), therefore left blank
                                        prediction = private$.predictShapeAux(data, tdays, cv_days[i], predict_from,
-                                               horizon, local, window, simtype, FALSE)
+                                               horizon, local, window, simtype, opera, FALSE)
                                        if (!is.na(prediction[1]))
                                        {
                                                nb_jours = nb_jours + 1
                                        1
 
                        return( private$.predictShapeAux(data, tdays, today, predict_from, horizon, local,
-                               best_window, simtype, TRUE) )
+                               best_window, simtype, opera, TRUE) )
                }
        ),
        private = list(
-               # Precondition: "today" is full (no NAs)
+               # Precondition: "yersteday until predict_from-1" is full (no NAs)
                .predictShapeAux = function(data, tdays, today, predict_from, horizon, local, window,
-                       simtype, final_call)
+                       simtype, opera, final_call)
                {
-                       tdays_cut = tdays[ tdays <= today-1 ]
-                       if (length(tdays_cut) <= 1)
+                       tdays_cut = tdays[ tdays != today ]
+                       if (length(tdays_cut) == 0)
                                return (NA)
 
                        if (local)
                        {
                                # TODO: 60 == magic number
                                tdays = getSimilarDaysIndices(today, data, limit=60, same_season=TRUE,
-                                       days_in=tdays_cut)
-                               if (length(tdays) <= 1)
-                                       return (NA)
-                               # TODO: 10, 12 == magic numbers
-                               tdays = .getConstrainedNeighbs(today,data,tdays,min_neighbs=10,max_neighbs=12)
+                                       days_in=tdays_cut, operational=opera)
+#                              if (length(tdays) <= 1)
+#                                      return (NA)
+                               # TODO: 10 == magic number
+                               tdays = .getConstrainedNeighbs(today, data, tdays, min_neighbs=10)
                                if (length(tdays) == 1)
                                {
                                        if (final_call)
                                # Compute endogen similarities using given window
                                window_endo = ifelse(simtype=="mix", window[1], window)
 
-                               # Distances from last observed day to days in the past
-                               lastSerie = c( data$getSerie(today-1),
-                                       data$getSerie(today)[if (predict_from>=2) 1:(predict_from-1) else c()] )
-                               distances2 = sapply(tdays, function(i) {
-                                       delta = lastSerie - c(data$getSerie(i-1),
-                                               data$getSerie(i)[if (predict_from>=2) 1:(predict_from-1) else c()])
-                                       sqrt(mean(delta^2))
-                               })
+                               # Distances from last observed day to selected days in the past
+                               distances2 <- .computeDistsEndo(data, today, tdays, predict_from)
+
+                               if (local)
+                               {
+                                       max_neighbs = 12 #TODO: 12 = arbitrary number
+                                       if (length(distances2) > max_neighbs)
+                                       {
+                                               ordering <- order(distances2)
+                                               tdays <- tdays[ ordering[1:max_neighbs] ]
+                                               distances2 <- distances2[ ordering[1:max_neighbs] ]
+                                       }
+                               }
 
                                simils_endo <- .computeSimils(distances2, window_endo)
                        }
                                # Compute exogen similarities using given window
                                window_exo = ifelse(simtype=="mix", window[2], window)
 
-                               M = matrix( ncol=1+length(tdays), nrow=1+length(data$getExo(1)) )
-                               M[,1] = c( data$getLevelHat(today), as.double(data$getExoHat(today)) )
-                               for (i in seq_along(tdays))
-                                       M[,i+1] = c( data$getLevel(tdays[i]), as.double(data$getExo(tdays[i])) )
-
-                               sigma = cov(t(M)) #NOTE: robust covariance is way too slow
-                               # TODO: 10 == magic number; more robust way == det, or always ginv()
-                               sigma_inv =
-                                       if (length(tdays) > 10)
-                                               solve(sigma)
-                                       else
-                                               MASS::ginv(sigma)
-
-                               # Distances from last observed day to days in the past
-                               distances2 = sapply(seq_along(tdays), function(i) {
-                                       delta = M[,1] - M[,i+1]
-                                       delta %*% sigma_inv %*% delta
-                               })
+                               distances2 <- .computeDistsExo(data, today, tdays)
 
                                simils_exo <- .computeSimils(distances2, window_exo)
                        }
 # @param min_neighbs Minimum number of points in a neighborhood
 # @param max_neighbs Maximum number of points in a neighborhood
 #
-.getConstrainedNeighbs = function(today, data, tdays, min_neighbs=10, max_neighbs=12)
+.getConstrainedNeighbs = function(today, data, tdays, min_neighbs=10)
 {
        levelToday = data$getLevelHat(today)
-       levelYersteday = data$getLevel(today-1)
+#      levelYersteday = data$getLevel(today-1)
        distances = sapply(tdays, function(i) {
-               sqrt((data$getLevel(i-1)-levelYersteday)^2 + (data$getLevel(i)-levelToday)^2)
+#              sqrt((data$getLevel(i-1)-levelYersteday)^2 + (data$getLevel(i)-levelToday)^2)
+               abs(data$getLevel(i)-levelToday)
        })
        #TODO: 1, +1, +3 : magic numbers
        dist_thresh = 1
                dist_thresh = dist_thresh + ifelse(dist_thresh>1,3,1)
        }
        tdays = tdays[same_pollution]
-       max_neighbs = 12
-       if (nb_neighbs > max_neighbs)
-       {
-               # Keep only max_neighbs closest neighbors
-               tdays = tdays[ order(distances[same_pollution])[1:max_neighbs] ]
-       }
+#      max_neighbs = 12
+#      if (nb_neighbs > max_neighbs)
+#      {
+#              # Keep only max_neighbs closest neighbors
+#              tdays = tdays[ order(distances[same_pollution])[1:max_neighbs] ]
+#      }
        tdays
 }
 
        }
        exp(-distances2/(sd_dist*window^2))
 }
+
+.computeDistsEndo <- function(data, today, tdays, predict_from)
+{
+       lastSerie = c( data$getSerie(today-1),
+               data$getSerie(today)[if (predict_from>=2) 1:(predict_from-1) else c()] )
+       sapply(tdays, function(i) {
+               delta = lastSerie - c(data$getSerie(i-1),
+                       data$getSerie(i)[if (predict_from>=2) 1:(predict_from-1) else c()])
+               sqrt(mean(delta^2))
+       })
+}
+
+.computeDistsExo <- function(data, today, tdays)
+{
+       M = matrix( ncol=1+length(tdays), nrow=1+length(data$getExo(1)) )
+       M[,1] = c( data$getLevelHat(today), as.double(data$getExoHat(today)) )
+       for (i in seq_along(tdays))
+               M[,i+1] = c( data$getLevel(tdays[i]), as.double(data$getExo(tdays[i])) )
+
+       sigma = cov(t(M)) #NOTE: robust covariance is way too slow
+       # TODO: 10 == magic number; more robust way == det, or always ginv()
+       sigma_inv =
+               if (length(tdays) > 10)
+                       solve(sigma)
+               else
+                       MASS::ginv(sigma)
+
+       # Distances from last observed day to days in the past
+       sapply(seq_along(tdays), function(i) {
+               delta = M[,1] - M[,i+1]
+               delta %*% sigma_inv %*% delta
+       })
+}
 
 {
        data = getDataTest(150)
 
-
-
-
-#TODO: debug
-
-
-
-       # index 144-1 == 143 : serie type 2
-       pred = computeForecast(data, 143, "Neighbors", "Zero", predict_from=8,
-               horizon=length(data$getSerie(1)), simtype="endo", local=FALSE, h_window=1)
+       # index 144 : serie type 3, yersteday type 2
+       pred = computeForecast(data, 144, "Neighbors", "Zero", predict_from=1,
+               horizon=length(data$getSerie(1)), simtype="endo", local=FALSE, window=1, opera=TRUE)
        f = computeFilaments(data, pred, 1, 8, limit=60, plot=FALSE)
 
-       # Expected output: 50-3-10 series of type 2, then 23 series of type 3 (closest next)
+       # Expected output: 50-3-10 series of type 2+1 = 3,
+       # then 23 series of type 3+1 %% 3 = 1 (3 = closest next)
        expect_identical(length(f$neighb_indices), as.integer(60))
        expect_identical(length(f$colors), as.integer(60))
-       expect_equal(f$index, 143)
-       expect_true(all(I(f$neighb_indices) >= 2))
+       expect_equal(f$index, 144)
+       expect_true(all(I(f$neighb_indices) != 2))
        for (i in 1:37)
        {
-               expect_equal(I(f$neighb_indices[i]), 2)
+               expect_equal(I(f$neighb_indices[i]), 3)
                expect_match(f$colors[i], f$colors[1])
        }
        for (i in 38:60)
        {
-               expect_equal(I(f$neighb_indices[i]), 3)
+               expect_equal(I(f$neighb_indices[i]), 1)
                expect_match(f$colors[i], f$colors[38])
        }
        expect_match(f$colors[1], "#1*")
        expect_match(f$colors[38], "#E*")
 
-       # index 143-1 == 142 : serie type 1
-       pred = computeForecast(data, 143, "Neighbors", "Zero", predict_from=8,
-               horizon=length(data$getSerie(1)), simtype="endo", local=FALSE, h_window=1)
+       # index 143 : serie type 2
+       pred = computeForecast(data, 143, "Neighbors", "Zero", predict_from=1,
+               horizon=length(data$getSerie(1)), simtype="endo", local=FALSE, window=1, opera=TRUE)
        f = computeFilaments(data, pred, 1, 8, limit=50, plot=FALSE)
 
-       # Expected output: 50-10-3 series of type 1, then 13 series of type 3 (closest next)
-       # NOTE: -10 because only past days with no-NAs tomorrow => exclude type 1 in [60,90[
+       # Expected output: 50-10-3 series of type 1+1=2,
+       # then 13 series of type 3+1 %% 3 = 1 (closest next)
+       # NOTE: -10 because only past tomorrows with no-NAs yerstedays
+       #        => exclude type 2 in [60,90[
        expect_identical(length(f$neighb_indices), as.integer(50))
        expect_identical(length(f$colors), as.integer(50))
        expect_equal(f$index, 143)
-       expect_true(all(I(f$neighb_indices) != 2))
+       expect_true(all(I(f$neighb_indices) <= 2))
        for (i in 1:37)
        {
-               expect_equal(I(f$neighb_indices[i]), 1)
+               expect_equal(I(f$neighb_indices[i]), 2)
                expect_match(f$colors[i], f$colors[1])
        }
        for (i in 38:50)
        {
-               expect_equal(I(f$neighb_indices[i]), 3)
+               expect_equal(I(f$neighb_indices[i]), 1)
                expect_match(f$colors[i], f$colors[38])
        }
        expect_match(f$colors[1], "#1*")