Getting Started

We’re glad to see you here. Today we’re going to install SDF and create a hello world project. Let’s get started.

Creating a Workspace

To create a new SDF Workspace, run the following command:

sdf new --sample hello && cd hello

After running the command, you will see the following output:

    Created hello/.gitignore
    Created hello/models/main.sql
    Created hello/workspace.sdf.yml
 
        Welcome to your new SDF Workspace! To help you on your journey:
        
        💡 Join the SDF Community Slack -> https://sdf.com/join 
        
        📚 Read the Docs to Get Started -> https://docs.sdf.com/
        
   Finished new in 0.289 secs

You can view and modify the source code for all our examples directly in our GitHub repository.

This will create a new directory with the name of your workspace. In this case, that’s hello. The workspace will contain the following files and folder structure:

.
├── models
│   └── main.sql
└── workspace.sdf.yml
2 directories, 2 files

The directory will contain a workspace.sdf.yml file, which is the primary configuration file for your project. It contains the following YML:

workspace:
  name: hello
  edition: “1.3”
  description: “A minimal workspace”
  includes:
    - path: models

For more on this, see our workspaces guide and for a full reference to our YML schema and more see the reference section.

Lastly, our workspace contains a single model file, main.sql, which contains the following SQL:

select ‘Hello World!’ as message

Next, let’s take a deeper look at your project.

SDF uses a cache to fingerprint outputs and accelerate recomputation. This cache is by default located in the sdftarget/ directory. The cache is machine specific and should not be checked in to git. An appropriate .gitignore file is created as part of the sdf new command.

We refer to SQL statements in SDF as models. Models are SQL statements that will materialized in your data warehouse, or locally with the SDF DB. They differ from tables as they can be materialized as tables, views, and more based on the configuration. Furthermore, they can be templatized with jinja and SDF SQL variables. SDF recommends specifying one model per file, as each model receives a fully qualified name (database.schema.table) that can correspond nicely to a directory structure. See our indexing documentation for more.

Exploring Your Project

Let’s see just how easy it is to set up SDF and run your first query.

1. Static Analysis with `sdf compile`

   First, we’ll run the core command of SDF: sdf compile.

   The --show flag allows you to modulate SDF’s output and the all option indicates that we would like to see all schemas from all models referenced in the workspace.

   sdf compile --show all
   

   Your output should look like:

   Working set 1 model file, 1 .sdf file
     Compiling hello.pub.main (./models/main.sql)
      Finished 1 model [1 succeeded] in 0.871 secs
   Schema hello.pub.main
   ┌─────────────┬──────────────────┬────────────┬─────────────┐
   │ column_name ┆ data_type        ┆ classifier ┆ description │
   ╞═════════════╪══════════════════╪════════════╪═════════════╡
   │ message     ┆ varchar not null ┆            ┆             │
   └─────────────┴──────────────────┴────────────┴─────────────┘

   As you can see from the output, SDF has statically analyzed the query and determined there’s a single non-nullable column named column and it’s of type varchar. You’ll also see an empty classifier block in the output. This is for metadata we’ll attach to columns, but we’ll get to that later.

2. Create a New Model

   Now, create a new file in the source directory called main2.sql with the query:

   main2.sql

   SELECT * FROM main;
   
3. Lineage with `sdf lineage`

   SDF guarantees rich column level lineage. The command below specifies a particular column, in a particular table that we would like to inspect.

   sdf lineage main2 --column message
   

   Working set 2 model files, 1 .sdf file
     Compiling hello.pub.main (./models/main.sql)
     Compiling hello.pub.main2 (./models/main2.sql)
      Finished 2 models [2 succeeded] in 0.853 secs
   hello.pub.main2.message
   │
   │ copy
   └──────┐
          hello.pub.main.message

4. Execution with `sdf run`

   Next, let’s execute the query, using SDF as the database. We’ll execute the query with SDF’s integrated execution runtime, right on your machine.

   sdf run --show all
   

   Your output should look like:

   Working set 2 model files, 1 .sdf file
       Running hello.pub.main (./models/main.sql)
       Running hello.pub.main2 (./models/main2.sql)
      Finished 2 models [2 succeeded] in 0.854 secs
   Table hello.pub.main
   ┌──────────────┐
   │ message      │
   ╞══════════════╡
   │ Hello World! │
   └──────────────┘
   1 rows.
   Table hello.pub.main2
   ┌──────────────┐
   │ message      │
   ╞══════════════╡
   │ Hello World! │
   └──────────────┘
   1 rows.

In this guide we showed you just how easy it is to install SDF and run your first query.

If using VSCode, SDF’s YML schema is available for type and syntax checking via the Red HAT YAML. This will add auto-fill, type checking, and YML validation directly inline while editing sdf.yml files.

Quick Links

Tutorials

Try it yourself! Get to know the real value of SDF through our short series of tutorials

Setup SDF

Learn more about setting up SDF in your data stack