Part 4: Random Data Generation

This is the final part of the tutorial where we will

Learn to connect Feldera pipelines to a random row generator for testing, benchmarking and debugging purposes.
Give a glimpse of the HTTP API to programmatically interact with Feldera.

You can skip this part if you are working with pre-existing data sources.

Why random data?

When creating a new pipeline, you might find yourself writing some SQL, without having any data to test it readily available. In this case, you can use the random data generator to create test data on the fly. This is especially useful when you want to test with large volumes of data.

Step 1. Create datagen connectors

You already learned how to create connectors and connect them to your pipeline in the previous parts of the tutorial. The datagen connector is just another connector that generates random rows for a table with some constraints on what gets generated based on the configuration you provide.

Let's configure a datagen connector for the VENDOR table to generate the following contents:

ID	NAME	ADDRESS
0	Gravitech Dynamics	222 Graviton Lane
1	HyperDrive Innovations	456 Warp Way
2	DarkMatter Devices	333 Singularity Street

create table VENDOR (
    id bigint not null primary key,
    name varchar,
    address varchar
) with (
  'materialized' = 'true',
  'connectors' = '[{
    "transport": {
      "name": "datagen",
      "config": {
        "plan": [
          { "limit": 3,
            "fields": {
              "name": { "values": ["Gravitech Dynamics", "HyperDrive Innovations", "DarkMatter Devices"] },
              "address": { "values": ["222 Graviton Lane", "456 Warp Way", "333 Singularity Street"] } } }
        ]
      }
    }
  }]'
);

First, we specify datagen as the transport. In the config section, we define a plan that describes how the rows are generated. You can add multiple plans to this list, and they will be executed sequentially, but for now we only need one.

In the plan we set the limit parameter, it specifies how many rows should be generated. In fields, we describe how the values for each column should be generated: For the name and address column, we give the list of the three names and addresses from the table above. We don't need to configure anything for the id column because the default strategy for generating integer values is to generate an incrementing sequence of numbers starting from 0.

We'll cover these "increment" generation strategies in more detail for the next table, PART:

create table PART (
    id bigint not null primary key,
    name varchar
) WITH (
  'connectors' = '[{
    "transport": {
      "name": "datagen",
      "config": {
        "plan": [
          { "limit": 3,
            "fields": {
              "id": { "strategy": "increment", "range": [1, 4] },
              "name": { "strategy": "increment", "values": ["Flux Capacitor", "Warp Core", "Kyber Crystal"] } } }
        ]
      }
    }
  }]'
);

This will fill the table with the following contents:

ID	NAME
1	Flux Capacitor
2	Warp Core
3	Kyber Crystal

Each member of the fields section can set a strategy that defines how the values are generated. The increment strategy is the default, so we could've omitted it like in the previous table. What's new is that we added the range parameter for the id column. That means we narrow the range of values generated for this field. Instead of starting from 0 as we did in the previous table, the id rows now have values 1, 2, 3.

For the name column, we also use the increment strategy. Again, we specify a fixed set of values. As previously, the increment strategy will select the values from the list one-by-one.

For the last table, PRICE, we insert some static contents to the table as we did before, but then we add a second plan to the connector that dynamically updates the prices to make it more interesting.

create table PRICE (
  part bigint not null,
  vendor bigint not null,
  price decimal,
  -- Make sure that new updates overwrite existing entries in PRICE for the same part and vendor ids.
  PRIMARY KEY (part, vendor)
) with (
  'materialized' = 'true',
  'connectors' = '[
  {
    "transport": {
      "name": "datagen",
      "config": {
        "plan": [
          { "limit": 3,
            "fields": {
              "part": { "range": [1, 4] },
              "vendor": { "values": [1, 0, 2] },
              "price": { "values": [10000, 15000, 9000] } } },
           { "rate": 1,
             "fields": {
              "vendor": { "values": [1, 0, 2] },
              "part": { "strategy": "zipf", "range": [1, 4] },
              "price": { "strategy": "uniform", "range": [9000, 15000] } }
           }
        ]
      }
    }
  }
  ]'
);

The first plan is similar to what we saw in the previous table (except we omit specifying the default increment strategy). The second plan has some new settings. We add a rate: 1 to tell the connector to emit one record every second, and we omit limit so this plan will continuously generate records until the pipeline is stopped. We keep the vendor column fixed, so every time we emit a record it will affect a different vendor. Next we use a new strategy, a Zipf distribution for the part column. This means that the connector will generate a random part ID, but the distribution of the IDs will be skewed towards the first ID in the range. Finally, we set the price column to be generated with a uniform strategy, and a range, which means that the price will be a random number between 9000 and 15000.

note

The data generator currently only generates insertions and does not delete previously added records; therefore we added a PRIMARY KEY constraint to the table, making sure that new updates overwrite existing entries in PRICE for the same part and vendor ids.

Let's start the pipeline and inspect its output in the Change Stream tab in the WebConsole. You should see changes in PREFERRED_VENDOR view approximately every second.

To summarize Part 4 of the tutorial, we can attach a random generator to Feldera tables to simulate different scenarios such as backfill, continuous evaluation or a combination of the two. You'll find a complete datagen reference in the connectors section.

Part 4: Random Data Generation

Why random data?​

Step 1. Create datagen connectors​

Why random data?

Step 1. Create datagen connectors