WaveRecipe Metadata Type Guide

Architectural Overview#

The WaveRecipe metadata type represents a visual Data Prep 3.0 transformation pipeline within CRM Analytics. As Salesforce’s flagship data integration framework for analytics, Data Prep recipes enable data engineers to build self-documenting transformation directed acyclic graphs (DAGs). These pipelines clean, aggregate, join, window, predict, and output large data volumes directly to CRM Analytics datasets, Salesforce core objects, or external data lakes.

Architecturally, WaveRecipe metadata decouples the pipeline container settings from the underlying processing graph. The XML descriptor file defines metadata ownership, labels, and target specifications, while the companion .wdpr file holds the rich JSON graph detailing transformation nodes, connection edges, visual canvas coordinates, and field-level mappings.

Because WaveRecipe replaces legacy WaveDataflow implementations for modern enterprise designs, establishing automated CI/CD validation for recipes is vital to preventing pipeline regressions and schema mismatches.

Directory and File Suffix#

In Salesforce DX source format, WaveRecipe artifacts are stored in the wave/ directory as a multi-file bundle:

• Data Prep JSON Graph: force-app/main/default/wave/<RecipeName>.wdpr
• Metadata Manifest File: force-app/main/default/wave/<RecipeName>.wdpr-meta.xml

force-app/
└── main/
    └── default/
        └── wave/
            ├── Global_Pipeline_Transformation.wdpr
            └── Global_Pipeline_Transformation.wdpr-meta.xml

Key Metadata Fields#

The .wdpr-meta.xml XML manifest exposes the following administrative fields:

The companion .wdpr file encapsulates:

• nodes: Dictionary of transformation nodes (LOAD_DATASET, JOIN, TRANSFORM, AGGREGATE, FILTER, OUTPUT).
• ui: Canvas layout information, coordinates, and visual node grouping.
• parameters: Execution parameters and writeback targets.

Declarative XML and JSON Examples#

Metadata XML Descriptor (wave/Global_Pipeline_Transformation.wdpr-meta.xml)#

<?xml version="1.0" encoding="UTF-8"?>
<WaveRecipe xmlns="http://soap.sforce.com/2006/04/metadata">
    <dataflowType>Recipe</dataflowType>
    <description>Modern Data Prep recipe combining opportunities and account tiers into high-performance executive datasets.</description>
    <format>R3</format>
    <masterLabel>Global Pipeline Transformation</masterLabel>
</WaveRecipe>

Accompanying Data Prep JSON Graph (wave/Global_Pipeline_Transformation.wdpr)#

{
  "version": "68.0",
  "nodes": {
    "LOAD_OPP": {
      "action": "load",
      "sources": [],
      "parameters": {
        "dataset": {
          "type": "analyticsDataset",
          "sourceObjectName": "Opportunity_Analytics_Dataset"
        },
        "fields": ["Id", "AccountId", "Amount", "StageName", "CloseDate"]
      }
    },
    "FILTER_WON": {
      "action": "filter",
      "sources": ["LOAD_OPP"],
      "parameters": {
        "filterExpressions": [
          {
            "field": "StageName",
            "operands": ["Closed Won"],
            "operator": "EQUAL"
          }
        ]
      }
    },
    "SAVE_METRICS": {
      "action": "save",
      "sources": ["FILTER_WON"],
      "parameters": {
        "dataset": {
          "type": "analyticsDataset",
          "label": "Closed Won Revenue Metrics",
          "name": "Closed_Won_Revenue_Metrics"
        }
      }
    }
  },
  "ui": {
    "nodes": {
      "LOAD_OPP": { "label": "Load Opportunities", "type": "LOAD_DATASET", "top": 100, "left": 100 },
      "FILTER_WON": { "label": "Filter Closed Won", "type": "FILTER", "top": 100, "left": 250 },
      "SAVE_METRICS": { "label": "Save Revenue Dataset", "type": "OUTPUT", "top": 100, "left": 400 }
    }
  }
}

Manifest Definition (package.xml)#

<?xml version="1.0" encoding="UTF-8"?>
<Package xmlns="http://soap.sforce.com/2006/04/metadata">
    <types>
        <members>Global_Pipeline_Transformation</members>
        <name>WaveRecipe</name>
    </types>
    <version>68.0</version>
</Package>

Retrieval and Deployment#

Retrieving Recipes with Salesforce CLI#

Retrieve the recipe metadata and its JSON graph together into your workspace:

# Retrieve a specific recipe
sf project retrieve start --metadata "WaveRecipe:Global_Pipeline_Transformation"

# Retrieve all recipes in the connected org
sf project retrieve start --metadata "WaveRecipe"

Deploying Recipes across Environments#

Validate and deploy your recipes across development, testing, and production orgs:

# Validate recipe syntax against target sandbox
sf project deploy validate \
    --metadata "WaveRecipe:Global_Pipeline_Transformation" \
    --target-org uat-sandbox

# Deploy the recipe to production
sf project deploy start \
    --metadata "WaveRecipe:Global_Pipeline_Transformation" \
    --target-org production-org

Inspecting Recipes via Tooling API#

Enterprise pipelines can verify deployed recipe versions and runtime IDs using the Tooling API:

sf data query --use-tooling-api --query \
    "SELECT Id, DeveloperName, MasterLabel, DataflowType, Format, CreatedDate, LastModifiedDate FROM Recipe WHERE DeveloperName = 'Global_Pipeline_Transformation'"

Architect Deployment Best Practices & Gotchas#

Frequently asked questions

What format does WaveRecipe use for modern data preparation?

Modern Data Prep recipes use the R3 format, storing visual graph nodes, input sources, transformations, and output targets inside the wdpr JSON file.

What happens when a WaveRecipe references non-existent fields during deployment?

The deployment fails with schema validation errors. All source fields referenced in recipe nodes must exist in the target org with read access granted to Analytics Cloud users.

How do WaveRecipe components differ from WaveDataflow components?

WaveRecipe provides modern visual Data Prep graphs with advanced capabilities like machine learning transforms and multi-value lookups, superseding legacy WaveDataflow pipelines.

Can a WaveRecipe publish output to an external system?

Yes, WaveRecipe output nodes can write to internal CRM Analytics datasets, Salesforce custom objects, or external cloud targets such as Amazon S3, Snowflake, or BigQuery.