[Python] Is there a way to make this R function complete its task faster?

I have a function that reads supermarket product allocation data, uses data based on his horizontal position (x), shelf level, aisle, and item no. I want to determine the upwards/downwards neighbor products by checking products placed on the upper/lower shelf that reduce the x difference. For that I have the following code:


# Directorios de entrada y salida
input_dir <- "J:/Zippedi/Data/Stage 3/Product_Data/Simplified_by_section/Data_with_neighbors_UD_and_CPs"
output_dir <- "J:/Zippedi/Data/Stage 3/Product_Data/Simplified_by_section/Data_with_neighbors_UD_and_CPs_ver_2"

# Crear directorio de salida si no existe
if (!dir.exists(output_dir)) {
print("Directory does not exist. Creating...")
dir.create(output_dir, recursive = TRUE)
print("Directory successfully created.")
} else {
print("Directory already exists")
}

# Verificar archivos disponibles
print("Checking facings files available...")
files_list <- list.files(input_dir, pattern = "\\.csv$", full.names = TRUE)

for (file in files_list) {

# Leer archivo CSV
setwd(input_dir)
df <- fread(file) # fread es más eficiente para archivos grandes

# DataFrame de salida
output_df <- data.table()

# Bucle por cada 'aisle'
for (aisle_index in unique(df$aisle)) {
print(aisle_index)

df_aisle <- df[df$aisle == aisle_index, ]

for (aisle_part_index in unique(df_aisle$aisle_part)) {
print(aisle_part_index)

df_aisle_part <- df_aisle[df_aisle$aisle_part == aisle_part_index, ]

for (orient_index in unique(df_aisle_part$orient)) {
print(orient_index)

df_orient <- df_aisle_part[df_aisle_part$orient == orient_index, ]

for (item_index in unique(df_orient$item)) {
print(item_index)

item_data <- df_orient[df_orient$item == item_index, ]

for (date in unique(item_data$date_PD)) {
print(date)

data_for_date <- df_orient[df_orient$date_PD == date,]
item_data <- df_orient[df_orient$item == item_index & df_orient$date_PD == date, ]

current_shelf_level <- as.numeric(item_data$No_of_cluster[1])
print(current_shelf_level)
current_loc_x <- as.numeric(item_data$Location_x_mean[1])
print(current_loc_x)

uw_shelf_data <- data_for_date[data_for_date$No_of_cluster == current_shelf_level + 1, ]
print(paste("No of rows in uw_shelf_data:", nrow(uw_shelf_data)))
dw_shelf_data <- data_for_date[data_for_date$No_of_cluster == current_shelf_level - 1, ]
print(paste("No of rows in dw_shelf_data:", nrow(dw_shelf_data)))

if (nrow(uw_shelf_data) == 0) {
# Procesar `neighbor_dw`
dw_shelf_data$dist <- abs(dw_shelf_data$Location_x_mean - current_loc_x)
closest_dw_neighbor <- dw_shelf_data[which.min(dw_shelf_data$dist), ]
item_data$neighbor_dw <- as.character(closest_dw_neighbor$item[1])
item_data$neighbor_uw <- NA_character_

} else if (nrow(dw_shelf_data) == 0) {
# Procesar `neighbor_uw`
uw_shelf_data$dist <- abs(uw_shelf_data$Location_x_mean - current_loc_x)
closest_uw_neighbor <- uw_shelf_data[which.min(uw_shelf_data$dist), ]
item_data$neighbor_uw <- as.character(closest_uw_neighbor$item[1])
item_data$neighbor_dw <- NA_character_

} else if (nrow(dw_shelf_data) == 0 & nrow(uw_shelf_data) == 0) {
item_data$neighbor_uw <- NA_character_
item_data$neighbor_dw <- NA_character_
} else {
# Procesar ambos
uw_shelf_data$dist <- abs(uw_shelf_data$Location_x_mean - current_loc_x)
dw_shelf_data$dist <- abs(dw_shelf_data$Location_x_mean - current_loc_x)
closest_uw_neighbor <- uw_shelf_data[which.min(uw_shelf_data$dist), ]
closest_dw_neighbor <- dw_shelf_data[which.min(dw_shelf_data$dist), ]
item_data$neighbor_uw <- as.character(closest_uw_neighbor$item[1])
item_data$neighbor_dw <- as.character(closest_dw_neighbor$item[1])
}

# Crear columnas faltantes en output_df
#print("Adding missing columns to output_df")
#missing_columns <- setdiff(names(item_data), names(output_df))
#for (col in missing_columns) {
# output_df[, (col) := NA]
#}
#print("success.")

# Añadir item_data a output_df
print("Binding item_data to output_df")
output_df <- rbind(output_df, item_data, fill = TRUE)
print("success.")

# Limpiar memoria
rm(item_data, closest_uw_neighbor, closest_dw_neighbor)
gc()
}
}

# Limpiar memoria
rm(df_orient)
gc()
}

rm(df_aisle_part)
gc()
}

rm(df_aisle)
gc()
}

# Guardar archivo
output_filename <- paste0(output_dir, "/", basename(file))
fwrite(output_df, output_filename)

# Limpiar memoria
rm(output_df)
gc()
}
}


But it solves the task to slowly. It reads 5 products per second. I need to find a way to solve the task faster, either in R or Python.

I tried using ChatGPT but it won't help.

Continue reading...