I have a Golang program which makes real time predictions on a machine learning model built using TensorFlow. The data for the prediction needs to be read line by line from Stdin and the prediction has to be performed on each line of data. The flow of data is not constant. I need a system that ensures that every time there is data to be read from Stdin, the prediction method is invoked and if there is no data in Stdin, the program waits for new data and does not terminate.
I tried achieving this using channels and the select, but if there is no data in the Stdin, the program terminates. Below is the code snippet:
func run_the_model(in <-chan string) {
go func(){
...
...
...
//Fetch the model
//Run the prediction
//print the result on StdOut
}()
}
func main() {
data := make(chan string)
// read data from Stdin
go func() {
scan := bufio.NewScanner(os.Stdin)
for scan.Scan() {
data <- scan.Text()
}
}()
time.Sleep(time.Second * 5)
select{
case <-data:
run_the_model(data)
time.Sleep(time.Second * 5)
default:
println("Waiting for data")
time.Sleep(time.Duration(math.MaxInt64))
}
}
When there is no new data in Stdin the Select's default case must be executed and when there is new data in the data
channel, the run_the_model
must be executed. How can this be achieved?