Key-Value - Chained

In this example we will show how we can create a service whose source and sink are both key-value.

Prerequisites

This guide uses local Fluvio cluster. If you need to install it, please follow the instructions at here.

Dataflow

Overview

For this example, we will write a dataflow that tracks the manufacture location of a car based on the VIN number. The source is key-value whose key is the vin and the value is the car info. The sink's key is the manufacture location and the value is an object with the old key.

Define the types

Like the previous example, we will need to define our types.

types:
  car-info:
    type: object
    properties:
      year:
        type: u32 
      brand:
        type: string
      model:
        type: string
  car-short:
    type: object
    properties:
      vin:
        type: string
      desc:
        type: string

car-info

This object is for the source's value.

car-short

This object is for the sinks's value. The vin is the key from the sink.

Topic List

The following is our list of topics. Both are key-value.

topics:
  input:
    schema:
      key:
        type: string
      value:
        type: car-info
  output:
    schema:
      key:
        type: string
      value:
        type: car-short

Transform

We will apply two map tranforms.

transforms:
  - operator: map
    run: |
      fn vin_to_car(vin: Option<String>, car: CarInfo) -> Result<CarShort> {
        Ok(CarShort{
            vin: vin.unwrap(),
            desc: format!("{} {} {}",car.year,car.brand,car.model),
        })
      }

  - operator: map
    run: |
      fn car_to_manu(vin: Option<String>, car_short: CarShort) -> Result<(Option<String>, CarShort)> {
        let country = if let Some(first_char) = vin.unwrap().chars().next() {
            if first_char == '1' || first_char == '4' || first_char == '5' { "United States" } 
            else if first_char == '2' { "Canada" } 
            else if first_char == '3' { "Mexico" } 
            else if first_char == 'J' { "Japan" } 
            else if first_char == 'K' { "South Korea" } 
            else if first_char == 'L' { "China" } 
            else if first_char == 'S' { "United Kingdom" } 
            else if first_char == 'V' { "France" } 
            else if first_char == 'W' { "Germany" } 
            else if first_char == 'Z' { "Italy" } 
            else { "Unknown" }
            } 
        else { "Invalid VIN" };
        Ok((Some(country.to_string()), car_short))
      }

vin-to-car() creates the value for the sink and car_to_manu modifies the key for the sink.

Running the Example

Full Code

Copy and paste following config and save it as dataflow.yaml.

apiVersion: 0.5.0

meta:
  name: chained-key
  version: 0.1.0
  namespace: example

config:
  converter: json
  consumer:
    default_starting_offset:
      value: 0
      position: End

types:
  car-info:
    type: object
    properties:
      year:
        type: u32 
      brand:
        type: string
      model:
        type: string
  car-short:
    type: object
    properties:
      vin:
        type: string
      desc:
        type: string
#Both topics are key-values
topics:
  input:
    schema:
      key:
        type: string
      value:
        type: car-info
  output:
    schema:
      key:
        type: string
      value:
        type: car-short

services:
  kv-to-kv:
    sources:
      - type: topic
        id: input

    transforms:
      - operator: map
        run: |
          fn vin_to_car(vin: Option<String>, car: CarInfo) -> Result<CarShort> {
            Ok(CarShort{
                vin: vin.unwrap(),
                desc: format!("{} {} {}",car.year,car.brand,car.model),
            })
          }

      - operator: map
        run: |
          fn car_to_manu(vin: Option<String>, car_short: CarShort) -> Result<(Option<String>, CarShort)> {
            let country = if let Some(first_char) = vin.unwrap().chars().next() {
                if first_char == '1' || first_char == '4' || first_char == '5' { "United States" } 
                else if first_char == '2' { "Canada" } 
                else if first_char == '3' { "Mexico" } 
                else if first_char == 'J' { "Japan" } 
                else if first_char == 'K' { "South Korea" } 
                else if first_char == 'L' { "China" } 
                else if first_char == 'S' { "United Kingdom" } 
                else if first_char == 'V' { "France" } 
                else if first_char == 'W' { "Germany" } 
                else if first_char == 'Z' { "Italy" } 
                else { "Unknown" }
                } 
            else { "Invalid VIN" };
            Ok((Some(country.to_string()), car_short))
          }

    sinks:
      - type: topic
        id: output

Running SDF

To run example:

$ sdf run

Produce data

We will produce some data by writing it into a file name cars.txt.

WBA7B41090G123456>{"year":1995,"brand":"BMW","model":"740i"}
1HGCB765XMA000001>{"year":1991,"brand":"Honda","model":"Accord"}
JM1FD3319P0112345>{"year":1991,"brand":"Mazda","model":"RX-7"}
KMHDH41D48U123456>{"year":2008,"brand":"Hyundai","model":"Elantra"}
JHMCB7658KC123456>{"year":1989,"brand":"Honda","model":"Accord"}
ZFA3120000S123456>{"year":2015,"brand":"Fiat","model": "500"}

We can produce data via

$ fluvio produce input --key-separator ">" -f cars.txt
$ fluvio consume input -Bdk

Consume data

To consume the data

$ fluvio consume output -Bdk
[Germany] {"desc":"1995 BMW 740i","vin":"WBA7B41090G123456"}
[United States] {"desc":"1991 Honda Accord","vin":"1HGCB765XMA000001"}
[Japan] {"desc":"1991 Mazda RX-7","vin":"JM1FD3319P0112345"}
[South Korea] {"desc":"2008 Hyundai Elantra","vin":"KMHDH41D48U123456"}
[Japan] {"desc":"1989 Honda Accord","vin":"JHMCB7658KC123456"}
[Italy] {"desc":"2015 Fiat 500","vin":"ZFA3120000S123456"}

The output's key is the country of the manufacture for the car and the value is the new object generated from the vin_to_car() function.

Cleanup

Exit sdf terminal and clean-up. The --force flag removes the topics:

$ sdf clean --force

Conclusion

This how-to focused on using key-values as output as well as the input. The following pages contains another example of key-value as chained.

1. Key Value Chained