etielle 3.5.0

A declarative, type-safe Python DSL for mapping complex nested JSON to relational database schemas

Quickstart: Declarative JSON-to-Relational Mapping in Python

etielle is a simple, powerful Python library for reshaping nested JSON data, typically from an API, into relational tables that fit your database schema. Think of etielle as a “JSON extractor” that you program with clear instructions: “Go here in the JSON, pull this data, and put it in that table.” The library’s name is a play on ETL (“Extract, Transform, Load”), which is the technical term for this set of operations.

• Repository: Promptly-Technologies-LLC/etielle
• PyPI: etielle
• Python: >= 3.13

Why Use etielle? (For Beginners)

JSON data from APIs is often deeply nested and requires complicated parsing. etielle helps by:

• Traversing nested structures: Walk through arrays-within-dictionaries-within-arrays to any arbitrary depth.
• Performing arbitrary transformations: Use built-in functions or define your own custom transforms.
• Building relationships: Link records across your different output tables automatically.
• Emitting to arbitrary formats: Emit data to Pydantic models, TypedDicts, or ORM objects directly.
• Optionally loading data into a database: Load data into a database using SQLAlchemy or SQLModel.

Learning Path

1. Quickstart: Quick and dirty introduction to etielle and how to use it.
2. Introduction to ETL: The problem etielle is solving: JSON data ETL.
3. Navigation: How to navigate through your JSON data with goto() and each().
4. Transforms: Getting and altering values from the JSON data.
5. Mapping Tables: Outputting data with Field and TempField.
6. Relationships: Linking tables together with link_to().
7. Database Loading: Persisting data with load() and run().

Installation

We recommend using uv for faster installs, but pip works too.

uv add etielle
# or
pip install etielle

Optional: Install with ORM adapters

If you plan to bind relationships and flush to your database via SQLAlchemy or SQLModel, install with the optional extra for your ORM:

uv add "etielle[sqlalchemy]"
# or
uv add "etielle[sqlmodel]"

Quick Start: Your First Mapping

Let’s start with a simple example. Suppose you have this JSON:

import json

data = {
  "users": [
    {"id": "u1", "name": "Alice", "posts": [{"id": "p1", "title": "Hello"}, {"id": "p2", "title": "World"}]},
    {"id": "u2", "name": "Bob", "posts": []}
  ]
}

We want two tables: “users” (id, name) and “posts” (id, user_id, title).

Here’s the code using the fluent API:

from etielle import etl, Field, TempField, get, get_from_parent

# Build and run the pipeline
result = (
    etl(data)
    # Extract users
    .goto("users").each()
    .map_to(table="users", fields=[
        Field("id", get("id")),
        Field("name", get("name")),
    ])
    # Extract posts (nested under each user)
    .goto("posts").each()
    .map_to(table="posts", fields=[
        Field("id", get("id")),
        Field("user_id", get_from_parent("id")),  # Link to parent user
        Field("title", get("title")),
    ])
    .run()
)

# result.tables gives you dict access by table name
out = {table: list(rows.values()) for table, rows in result.tables.items()}
print(json.dumps(out, indent=2))

{
  "users": [
    {
      "id": "u1",
      "name": "Alice"
    },
    {
      "id": "u2",
      "name": "Bob"
    }
  ],
  "posts": [
    {
      "id": "p1",
      "user_id": "u1",
      "title": "Hello"
    },
    {
      "id": "p2",
      "user_id": "u1",
      "title": "World"
    }
  ]
}

Congrats! You’ve mapped your first JSON using the fluent E->T->L pattern.

Core Concepts: Breaking It Down

1. The etl() Entry Point

Everything starts with etl(data). This creates a pipeline builder that you chain methods onto:

from etielle import etl

result = (
    etl(data)           # Start with your JSON
    .goto("users")      # Navigate to a path
    .each()             # Iterate over items
    .map_to(...)        # Emit table rows
    .run()              # Execute and get results
)

2. Navigation: goto() and each()

Navigation tells etielle where to find your data:

• goto(path): Navigate to a nested location (supports dot notation: "data.users")
• each(): Iterate over items in a list or dict

# Navigate to data["response"]["users"] and iterate
.goto("response.users").each()

# Or use list syntax
.goto(["response", "users"]).each()

3. Fields and TempFields

When you call map_to(), you define what data to extract:

• Field(name, transform): A column that appears in your output
• TempField(name, transform): Used for joins/relationships, but NOT in output

.map_to(table="users", fields=[
    Field("id", get("id")),          # Output column (also available for joins)
    Field("name", get("name")),      # Output column
])

4. Transforms: Smart Data Extractors

Transforms are functions that pull values from the current context:

Transform	Purpose
get("name")	Get field from current node
get_from_parent("id")	Get field from parent context
get_from_root("config.version")	Get field from JSON root
literal(42)	Constant value
key()	Current dict key
index()	Current list index
concat(a, b, c)	Join strings

# Combine transforms
Field("full_id", concat(literal("user_"), get("id")))  # "user_u1"

5. Running the Pipeline

Call .run() to execute and get results:

result = pipeline.run()

# Access tables by name
users = result.tables["users"]    # Dict[tuple, dict]

# Or by model class (if using typed models)
users = result.tables[User]       # Dict[tuple, User]

# Check for errors
if result.errors:
    for table, errs in result.errors.items():
        print(f"{table}: {errs}")

Detailed Examples

Example 1: Merging Data from Multiple Paths

Merge user info from two parts of JSON using goto_root():

from etielle import etl, Field, TempField, get

data = {
    "users": [{"id": "u1", "name": "Alice"}],
    "profiles": [{"user_id": "u1", "email": "[e-mail]"}]
}

result = (
    etl(data)
    # First path: basic user data
    .goto("users").each()
    .map_to(table="users", join_on=["id"], fields=[
        Field("id", get("id")),
        Field("name", get("name")),
    ])
    # Second path: profile data (same table, merged by id)
    .goto_root()
    .goto("profiles").each()
    .map_to(table="users", join_on=["id"], fields=[
        Field("email", get("email")),
        TempField("id", get("user_id"))  # Join key for merging
    ])
    .run()
)

user = list(result.tables["users"].values())[0]
print(user)  # Has id, name, AND email merged together

{'id': 'u1', 'name': 'Alice', 'email': '[e-mail]'}

Example 2: Deep Nesting

Handle deeply nested structures with chained navigation:

# servers -> channels -> messages -> reactions (3 levels deep)
result = (
    etl(data)
    .goto("servers").each()
    .map_to(table="servers", fields=[...])

    .goto("channels").each()
    .map_to(table="channels", fields=[
        Field("server_id", get_from_parent("id", depth=1)),
        ...
    ])

    .goto("messages").each()
    .map_to(table="messages", fields=[
        Field("channel_id", get_from_parent("id", depth=1)),
        Field("server_id", get_from_parent("id", depth=2)),
        ...
    ])
    .run()
)

Example 3: Typed Output with Pydantic

Use model classes for validated, typed output:

from pydantic import BaseModel
from etielle import etl, Field, TempField, get

class User(BaseModel):
    id: str
    name: str
    email: str | None = None

data = {"users": [{"id": "u1", "name": "Alice"}]}

result = (
    etl(data)
    .goto("users").each()
    .map_to(table=User, fields=[  # Pass model class, not string
        Field("id", get("id")),
        Field("name", get("name")),
    ])
    .run()
)

user = list(result.tables[User].values())[0]
print(f"Type: {type(user).__name__}, name: {user.name}")

Type: User, name: Alice

Transform Cheatsheet

Transform	Purpose	Example
get(path)	From current node	get("user.name")
get_from_parent(path, depth=1)	From ancestor	get_from_parent("id")
get_from_root(path)	From JSON root	get_from_root("version")
key()	Current dict key	When iterating {"a": 1, "b": 2}
index()	Current list index	0, 1, 2, …
parent_key()	Parent’s dict key	Access parent iteration key
parent_index()	Parent’s list index	Access parent iteration index
node()	Current node value	The whole current object
literal(value)	Constant	literal(42)
concat(*parts)	Join strings	concat(get("first"), literal(" "), get("last"))
coalesce(*transforms)	First non-None	coalesce(get("nickname"), get("name"))
format_id(*parts, sep="_")	Join with separator	format_id(get("type"), get("id"))
len_of(inner)	Length of list/dict/string	len_of(get("tags"))

Common Mistakes

• Empty results?
  • Check your goto() path matches the JSON structure exactly
  • Make sure you called .each() to iterate
• Missing parent data?
  • Check the depth parameter in get_from_parent()
  • Ensure the parent context exists in your navigation chain
• Duplicate or missing rows?
  • Verify TempField values are unique for each row
  • Check that join keys don’t contain None values

Next Steps

• Navigation - Deep dive into goto(), each(), and goto_root()
• Transforms - All built-in transforms and how to use them
• Mapping Tables - Field, TempField, merge policies
• Relationships - Link tables with link_to()
• Database Loading - Persist with load().run()

Glossary

• Pipeline: The chain of operations from etl() to run()
• Navigation: Methods like goto() and each() that position you in the JSON
• Transform: A function that extracts values from the current context
• Field: An output column in your table
• TempField: A field used for joins/linking but not in final output
• Join Key: Values that uniquely identify a row (derived from TempFields)

License

MIT

Need help? Open an issue on GitHub!