Skip to main content

json permutation library

Project description

J-Perm

A composable JSON transformation DSL with a powerful, extensible architecture.

J-Perm lets you describe data transformations as executable specifications — a list of steps that can be applied to input documents. It supports JSON Pointer addressing, template interpolation with ${...} syntax, special constructs ($ref, $eval, $and, $or, $not), and a rich set of built-in operations.


Quick Example

from j_perm import build_default_engine

engine = build_default_engine()

# Source data
source = {
    "users": [
        {"name": "Alice", "age": "17"},
        {"name": "Bob", "age": "22"}
    ]
}

# Transformation spec (using shorthands)
spec = {
    "op": "foreach",
    "in": "/users",
    "do": {
        "/adults[]": {
            "$eval": [
                {"/name": "/item/name", "/age": "${int:/item/age}"},
                {"op": "if", "path": "/age", "cond": "${?dest.age >= `18`}", "else": {"~delete": "/age"}}
            ]
        }
    }
}

result = engine.apply(spec, source=source, dest={})
# → {"adults": [{"name": "Bob", "age": 22}]}

Installation

pip install j-perm

(or copy the package into your project)


Architecture Overview

J-Perm is built on a pipeline architecture with two main levels:

┌─────────────────────────────────────────────────────────┐
│  spec (user input)                                      │
│    │                                                    │
│    ▼                                                    │
│  ┌──────────────────────────────────────────────────┐   │
│  │ STAGES (batch preprocessing, priority order)     │   │
│  │  • ShorthandExpansion → expand ~delete, etc      │   │
│  │  • YourCustomStage                               │   │
│  └──────────────────────────────────────────────────┘   │
│    │                                                    │
│    ▼                                                    │
│  List[step]                                             │
│    │                                                    │
│    ▼  for each step:                                    │
│  ┌──────────────────────────────────────────────────┐   │
│  │ MIDDLEWARES (per-step, priority order)           │   │
│  │  • Validation, logging, etc.                     │   │
│  └──────────────────────────────────────────────────┘   │
│    │                                                    │
│    ▼                                                    │
│  ┌──────────────────────────────────────────────────┐   │
│  │ REGISTRY (hierarchical dispatch tree)            │   │
│  │  • SetHandler, CopyHandler, ForeachHandler, ...  │   │
│  └──────────────────────────────────────────────────┘   │
│    │                                                    │
│    │  handlers call ctx.engine.process_value(...)       │
│    └─────────────────────────────────────┐              │
│                                          ▼              │
│  ┌──────────────────────────────────────────────────┐   │
│  │ VALUE PIPELINE (stabilization loop)              │   │
│  │  • SpecialResolveHandler ($ref, $eval)           │   │
│  │  • TemplSubstHandler (${...})                    │   │
│  │  • RecursiveDescentHandler (containers)          │   │
│  │  • IdentityHandler (scalars)                     │   │
│  └──────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────┘

Core Components

Component Purpose
Engine Orchestrates pipelines, manages context, runs stabilization loop
Pipeline Runs stages → middlewares → registry dispatch for each step
StageRegistry Tree of batch preprocessors (run-all, priority order)
ActionTypeRegistry Tree of action handlers (first-match or run-all)
ValueResolver Abstraction for addressing (JSON Pointer implementation)

Core API

Building an Engine

from j_perm import build_default_engine

# Default engine with all built-ins
engine = build_default_engine()

# Custom specials (None = use defaults: $ref, $eval, $and, $or, $not)
engine = build_default_engine(
    specials={"$ref": my_ref_handler, "$custom": my_handler},
    casters={"int": lambda x: int(x), "json": lambda x: json.loads(x)},
    jmes_options=jmespath.Options(custom_functions=CustomFunctions())
)

Applying Transformations

result = engine.apply(
    spec,  # DSL script (dict or list)
    source=source,  # Source context (for pointers, templates)
    dest=dest,  # Initial destination (default: {})
)

Returns: Deep copy of the final dest after all transformations.


Features

1. JSON Pointer Addressing

J-Perm uses RFC 6901 JSON Pointer with extensions:

from j_perm import PointerResolver

resolver = PointerResolver()

# Basic pointers
resolver.get("/users/0/name", data)  # → "Alice"

# Root references (work on scalars too!)
resolver.get(".", 42)  # → 42
resolver.get("/", "text")  # → "text"

# Parent navigation
resolver.get("/a/b/../c", data)  # → data["a"]["c"]

# List slices
resolver.get("/items[1:3]", data)  # → [item1, item2]

# Append notation
resolver.set("/items/-", data, "new")  # Append to list

Key feature: Unlike standard JSON Pointer, PointerResolver works on any type (scalars, lists, dicts) for root references.

Data Source Prefixes

J-Perm supports prefixes to specify which context to read from:

Prefix Source Description
/path source Default - read from source context
@:/path dest Read from destination being built
_:/path metadata Read from execution metadata

Example: Accessing destination in templates

# Build incrementally, referencing previous values
spec = [
    {"/name": "Alice"},
    {"/greeting": "Hello, ${@:/name}!"}  # Reference dest value
]

result = engine.apply(spec, source={}, dest={})
# → {"name": "Alice", "greeting": "Hello, Alice!"}

Example: Comparing source vs dest

spec = [
    {"/dest_value": "from_dest"},
    {"/comparison": "Source: ${/source_value}, Dest: ${@:/dest_value}"}
]

result = engine.apply(
    spec,
    source={"source_value": "from_source"},
    dest={}
)
# → {"dest_value": "from_dest", "comparison": "Source: from_source, Dest: from_dest"}

Example: Using in conditionals

# Check destination state in conditions
spec = [
    {"/status": "ready"},
    {
        "op": "if",
        "path": "@:/status",
        "equals": "ready",
        "then": [{"/message": "System is ready"}]
    }
]

2. Template Interpolation (${...})

Templates are resolved by TemplSubstHandler in the value pipeline.

JSON Pointer lookup

"${/user/name}"  # → Resolve pointer from source

Type casters (built-in)

"${int:/age}"  # → int(value)
"${float:/price}"
"${bool:/flag}"  # → bool(int(value)) if int/str, else bool(value)
"${str:/id}"

JMESPath queries

"${?source.items[?price > `10`].name}"  # → Query source with JMESPath
"${?dest.total}"                         # → Query destination
"${?add(dest.x, source.y)}"              # → Mix source and dest

Built-in JMESPath functions: add(a, b), subtract(a, b)

Data structure: JMESPath expressions use explicit namespaces:

  • source.* – access source document
  • dest.* – access destination document

Nested templates

"${${/path_to_field}}"  # → Resolve inner template first

Escaping

$${ → ${ (literal)
$$  → $  (literal)

3. Special Constructs

Special values are resolved by SpecialResolveHandler.

$ref — Reference resolution

{
    "$ref": "/path/to/value",
    "$default": "fallback"
}
  • Resolves pointer from source context
  • Returns deep copy (no aliasing)
  • Supports $default fallback

$eval — Nested evaluation

{
    "$eval": [
        {
            "op": "set",
            "path": "/x",
            "value": 1
        }
    ],
    "$select": "/x"
}
  • Executes nested DSL with dest={}
  • Optionally selects sub-path from result

$and — Logical AND with short-circuit

{
    "$and": [
        [{"op": "assert", "path": "/x", "return": true}],
        [{"op": "copy", "from": "/y", "path": "/"}]
    ]
}
  • Executes actions in order with empty dest for each
  • Returns last result if all are truthy
  • Short-circuits and returns first falsy result

$or — Logical OR with short-circuit

{
    "$or": [
        [{"op": "assert", "path": "/x", "return": true}],
        [{"op": "copy", "from": "/y", "path": "/"}]
    ]
}
  • Executes actions in order with empty dest for each
  • Returns first truthy result
  • Returns last result if all are falsy

$not — Logical negation

{
    "$not": [{"op": "assert", "path": "/missing", "return": true}]
}
  • Executes action with empty dest
  • Returns logical negation of the result

4. Functions and Error Handling

J-Perm supports defining reusable functions and controlled error handling.

$def — Define a function

{
    "$def": "myFunction",
    "params": ["arg1", "arg2"],
    "body": [
        {"/result": "${arg1}"},
        {"/total": "${int:${arg2}}"}
    ],
    "return": "/total",
    "on_failure": [
        {"/error": "Function failed"}
    ]
}
  • params — list of parameter names (optional, default: [])
  • body — actions to execute when function is called
  • return — path in local context to return (optional, default: entire dest)
  • on_failure — error handler actions (optional)

Special variable _: Inside function body, the original source is accessible via /_/path:

{
    "$def": "getFromSource",
    "body": [{"/value": {"$ref": "/_/data"}}]
}

When called, functions can access:

  • Parameters via /${param_name}
  • Original source via /_/path (source from engine.apply)
  • Current dest via @:/path

Example:

spec = [
    {
        "$def": "combineData",
        "params": ["userInput"],
        "body": [
            {"/input": "${userInput}"},
            {"/global": {"$ref": "/_/config"}},  # From original source
            {"/result": "Input: ${@:/input}, Config: ${@:/global/setting}"}
        ],
        "return": "/result"
    },
    {"/output": {"$func": "combineData", "args": ["test"]}}
]

result = engine.apply(
    spec,
    source={"config": {"setting": "production"}},
    dest={}
)
# → {"output": "Input: test, Config: production"}

$func — Call a function

{
    "$func": "myFunction",
    "args": [10, 20]
}
  • args — list of arguments to pass (optional, default: [])

Functions are stored in the execution context metadata and can be called multiple times within the same transformation.

$raise — Raise an error

{
    "$raise": "Invalid data: ${/error_details}"
}

Raises a JPermError with the specified message. The error can be:

  • Caught by on_failure handlers in function definitions
  • Used for validation and control flow
  • Combined with templates for dynamic error messages

Example with error handling:

[
    {
        "$def": "validateAge",
        "params": ["age"],
        "body": [
            {
                "op": "if",
                "cond": {"$not": [{"op": "assert", "value": "${int:${age}}", "return": true}]},
                "then": [{"$raise": "Age must be a number"}]
            },
            {
                "op": "if",
                "cond": "${?source.age < `0`}",
                "then": [{"$raise": "Age cannot be negative"}]
            }
        ],
        "on_failure": [
            {"/validation_error": "@:/"}
        ]
    },
    {"/result": {"$func": "validateAge", "args": [25]}}
]

5. Shorthand Syntax

Shorthands are expanded by priority-ordered StageProcessors before execution.

~assert

{
    "~assert": {
        "/x": 10,
        "/y": 20
    }
}

Expands to:

[
    {
        "op": "assert",
        "path": "/x",
        "equals": 10
    },
    {
        "op": "assert",
        "path": "/y",
        "equals": 20
    }
]

~delete

{
    "~delete": [
        "/tmp",
        "/cache"
    ]
}

Expands to:

[
    {
        "op": "delete",
        "path": "/tmp"
    },
    {
        "op": "delete",
        "path": "/cache"
    }
]

Append notation (field[])

{
    "/items[]": 123
}

Expands to:

{
    "op": "set",
    "path": "/items/-",
    "value": 123
}

Pointer assignment

{
    "/name": "/user/fullName"
}

Expands to:

{
    "op": "copy",
    "from": "/user/fullName",
    "path": "/name",
    "ignore_missing": true
}

Literal assignment

{
    "/status": "active"
}

Expands to:

{
    "op": "set",
    "path": "/status",
    "value": "active"
}

Priority order: ~assert (100) → ~delete (50) → pointer/literal assignment (0)


Built-in Operations

All operations are registered as ActionHandler instances in the main registry.

set

Write value to destination path.

{
    "op": "set",
    "path": "/target",
    "value": "...",
    "create": true,
    // Auto-create parents (default: true)
    "extend": true
    // Extend lists on append (default: true)
}

Special: path ending with /- appends to list.


copy

Copy value from source to destination.

{
    "op": "copy",
    "from": "/source/path",
    "path": "/dest/path",
    "ignore_missing": false,
    // Skip if missing (default: false)
    "default": "..."
    // Fallback value
}

delete

Remove value at path.

{
    "op": "delete",
    "path": "/remove",
    "ignore_missing": true
    // Don't error if missing (default: true)
}

foreach

Iterate over array/mapping.

{
    "op": "foreach",
    "in": "/items",
    "as": "item",
    // Variable name (default: "item")
    "do": [
        ...
    ],
    // Nested actions
    "skip_empty": true,
    // Skip if empty (default: true)
    "default": []
    // Fallback if missing
}

Note: If source is a dict, iterates over (key, value) tuples.

Special variable _: Inside foreach body, you can access:

  • Current item via /${var} (e.g., /item)
  • Original source via /_/path (source from engine.apply)
  • Current dest via @:/path

Example:

# Enrich items with global config from source
spec = {
    "op": "foreach",
    "in": "/products",
    "as": "product",
    "do": {
        "/results[]": {
            "$eval": [
                {"/name": "${product/name}"},
                {"/price": "${product/price}"},
                {"/tax_rate": {"$ref": "/_/config/tax"}},  # From original source
                {"/total": "${?source.price * source.tax_rate}"}
            ]
        }
    }
}

result = engine.apply(
    spec,
    source={
        "products": [{"name": "A", "price": 100}, {"name": "B", "price": 200}],
        "config": {"tax": 1.2}
    },
    dest={}
)
# → {"results": [{"name": "A", "price": 100, "tax_rate": 1.2, "total": 120}, ...]}

while

Loop while condition holds.

Path mode:

{
    "op": "while",
    "path": "/counter",
    "equals": 0,
    // Or "exists": true
    "do": [
        ...
    ],
    "do_while": false
    // Execute at least once (default: false)
}

Expression mode:

{
    "op": "while",
    "cond": "${?dest.counter < `10`}",
    "do": [
        ...
    ]
}

Note: Condition is checked against destination state. Use do_while: true to execute body at least once before checking condition.


if

Conditional execution.

Path mode:

{
    "op": "if",
    "path": "/check",
    "equals": "value",
    // Optional
    "exists": true,
    // Optional
    "then": [
        ...
    ],
    // Success branch
    "else": [
        ...
    ]
    // Failure branch
}

Expression mode:

{
    "op": "if",
    "cond": "${?source.age >= `18`}",
    "then": [
        ...
    ]
}

exec

Execute nested script.

From source:

{
    "op": "exec",
    "from": "/script",
    "merge": false
    // Replace dest (default) or merge into it
}

Inline:

{
    "op": "exec",
    "actions": [
        ...
    ]
}

update

Merge mapping into target.

{
    "op": "update",
    "path": "/obj",
    "value": {
        "b": 2
    },
    // Or "from": "/source/obj"
    "deep": false
    // Recursive merge (default: false)
}

distinct

Remove duplicates from list.

{
    "op": "distinct",
    "path": "/items",
    "key": "/id"
    // Optional: compare by nested field
}

assert

Assert value existence/equality.

Basic usage:

{
    "op": "assert",
    // Check source
    "path": "/required",
    "equals": "value"
    // Optional
}

With direct value:

{
    "op": "assert",
    "value": "${?source.computed}",
    // Check computed value instead of path
    "equals": "expected"
}

With return mode:

{
    "op": "assert",
    "path": "/optional",
    "return": true,
    // Return value instead of raising error
    "to_path": "/result"
    // Optional: write result to destination
}
  • return: true — returns value on success, false on failure (instead of raising error)
  • to_path — destination path for return value
  • value — alternative to path, checks direct value

Extending J-Perm

Custom Operations

Create a new ActionHandler and register it:

from j_perm import ActionHandler, ActionNode, OpMatcher, ExecutionContext


class MyOpHandler(ActionHandler):
    def execute(self, step, ctx: ExecutionContext):
        # Your logic here
        return ctx.dest


# Register in main registry (in build_default_engine or custom factory)
registry.register(ActionNode(
    name="my_op",
    priority=10,
    matcher=OpMatcher("my_op"),
    handler=MyOpHandler(),
))

Custom Special Constructs

Add a SpecialFn to the specials dict:

def my_special(node, ctx):
    value = ctx.engine.process_value(node["$mySpecial"], ctx)
    return value.upper()


engine = build_default_engine(specials={
    "$ref": ref_handler,
    "$eval": eval_handler,
    "$mySpecial": my_special,
})

Custom Stages

Create a StageProcessor for batch preprocessing:

from j_perm import StageProcessor, StageNode, StageRegistry


class ValidateStage(StageProcessor):
    def apply(self, steps, ctx):
        # Validate/transform steps
        return steps


# Register in main pipeline stages
stages = build_default_shorthand_stages()
stages.register(StageNode(
    name="validate",
    priority=200,  # Higher = runs earlier
    processor=ValidateStage(),
))

# Use in custom engine
main_pipeline = Pipeline(stages=stages, registry=main_registry)

Custom Casters

Provide custom casters to TemplSubstHandler:

from j_perm import TemplSubstHandler

custom_casters = {
    "int": lambda x: int(x),
    "json": lambda x: json.loads(x),
}

handler = TemplSubstHandler(casters=custom_casters)

Or use the default (built-in int, float, bool, str).


Custom Matchers

Implement ActionMatcher or StageMatcher:

from j_perm import ActionMatcher


class PrefixMatcher(ActionMatcher):
    def __init__(self, prefix):
        self.prefix = prefix

    def matches(self, step):
        return isinstance(step, dict) and
            step.get("op", "").startswith(self.prefix)

Advanced Topics

Value Stabilization Loop

When handlers call ctx.engine.process_value(value, ctx), the value pipeline runs repeatedly until:

  1. Output equals input (stable)
  2. value_max_depth iterations reached (default: 50)

This resolves nested templates and special constructs:

# Input: {"$ref": "/path_to_template"}
# Pass 1: {"$ref": ...} → "${/nested}"
# Pass 2: "${/nested}" → "final"
# Pass 3: "final" → "final" (stable ✓)

Hierarchical Registries

Both StageRegistry and ActionTypeRegistry support tree structures:

# Group related operations
math_registry = ActionTypeRegistry()
math_registry.register(ActionNode("add", 10, AddMatcher(), AddHandler()))
math_registry.register(ActionNode("sub", 10, SubMatcher(), SubHandler()))

# Mount as sub-tree
main_registry.register_group(
    "math",
    math_registry,
    matcher=OpMatcher("math"),
    priority=50,
)

Priority and Execution Order

Stages: All matching stages run in priority order (high → low).

Actions: First matching handler executes (unless exclusive=False).

Shorthands:

  1. AssertShorthandProcessor (100) — extracts ~assert
  2. DeleteShorthandProcessor (50) — extracts ~delete
  3. AssignShorthandProcessor (0) — fallback for all remaining keys

Unescape Rules

After value stabilization, registered UnescapeRule callables strip escape sequences:

from j_perm import UnescapeRule

# Built-in: template_unescape (strips $${ → ${, $$ → $)
# Registered at priority 0

# Add custom unescape
engine.unescape_rules.append(
    UnescapeRule(name="custom", priority=10, unescape=my_unescape_fn)
)

API Reference

Core Classes

from j_perm import (
    # Core infrastructure
    ExecutionContext,
    ValueResolver,
    Engine,
    Pipeline,

    # Stage system
    StageProcessor,
    StageMatcher,
    StageNode,
    StageRegistry,

    # Action system
    ActionHandler,
    ActionMatcher,
    ActionNode,
    ActionTypeRegistry,

    # Middleware
    Middleware,

    # Unescape
    UnescapeRule,
)

Handlers

from j_perm import (
    # Value handlers
    TemplMatcher,
    TemplSubstHandler,
    SpecialMatcher,
    SpecialResolveHandler,
    ContainerMatcher,
    RecursiveDescentHandler,
    IdentityHandler,

    # Special construct functions
    ref_handler,
    eval_handler,
    and_handler,
    or_handler,
    not_handler,

    # Function handlers
    DefMatcher,
    CallMatcher,
    DefHandler,
    CallHandler,
    RaiseMatcher,
    RaiseHandler,
    JPermError,

    # Operation handlers
    SetHandler,
    CopyHandler,
    DeleteHandler,
    ForeachHandler,
    WhileHandler,
    IfHandler,
    ExecHandler,
    UpdateHandler,
    DistinctHandler,
    AssertHandler,
)

Utilities

from j_perm import (
    # Matchers
    OpMatcher,
    AlwaysMatcher,

    # Resolver
    PointerResolver,

    # Shorthand stages
    AssertShorthandProcessor,
    DeleteShorthandProcessor,
    AssignShorthandProcessor,

    # Factory
    build_default_engine,
    build_default_shorthand_stages,
)

Examples

Example 1: Data Filtering

spec = {
    "op": "foreach",
    "in": "/products",
    "do": {
        "op": "if",
        "cond": "${?source.item.price < `100`}",
        "then": {"/affordable[]": "/item"}
    }
}

Example 2: Conditional Copy with Default

spec = {
    "/result": {
        "$ref": "/maybe_missing",
        "$default": "not found"
    }
}

Example 3: Nested Evaluation

spec = {
    "/computed": {
        "$eval": [
            {"op": "set", "path": "/x", "value": "${int:/a}"},
            {"op": "set", "path": "/y", "value": "${int:/b}"}
        ],
        "$select": "${?add(dest.x, dest.y)}"
    }
}

Example 4: Mixed Shorthands

spec = [
    {"~assert": {"/user/id": 123}},
    {"~delete": "/temp"},
    {"/output": "/user/name"}
]

Example 5: Functions with Error Handling

spec = [
    # Define a validation function with error recovery
    {
        "$def": "validateAge",
        "params": ["age"],
        "body": [
            # Validate age is a number
            {
                "op": "if",
                "cond": {"$not": [{"op": "assert", "value": "${int:${age}}", "return": True}]},
                "then": [{"$raise": "Age must be a number"}]
            },
            # Validate age is positive
            {
                "op": "if",
                "cond": "${?source.age < `0`}",
                "then": [{"$raise": "Age cannot be negative: ${age}"}]
            },
            {"/valid": True}
        ],
        "return": "/valid",
        "on_failure": [
            # Capture the error in dest
            {"/validation_failed": True},
            {"/last_error": "Validation error occurred"}
        ]
    },
    # Use function to validate user input
    {"/user_age_valid": {"$func": "validateAge", "args": [25]}},
    # Access dest values with @: prefix
    {
        "op": "if",
        "path": "@:/validation_failed",
        "equals": True,
        "then": [{"/message": "Please check your input: ${@:/last_error}"}]
    }
]

License

MIT (or adapt to your project as needed)


Contributing

Issues and pull requests welcome!

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

j_perm-1.2.0.tar.gz (43.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

j_perm-1.2.0-py3-none-any.whl (42.4 kB view details)

Uploaded Python 3

File details

Details for the file j_perm-1.2.0.tar.gz.

File metadata

  • Download URL: j_perm-1.2.0.tar.gz
  • Upload date:
  • Size: 43.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.0

File hashes

Hashes for j_perm-1.2.0.tar.gz
Algorithm Hash digest
SHA256 4f5dffa5799b09577740848614eec46c982381735c9f4ca2fd94a1d59e7ecf2e
MD5 9235e78ade60a1b522aab072acbbf3d3
BLAKE2b-256 9a693bf72c58cdcbac82839c649f8924dc7b7f36a191d981246116f227f5f213

See more details on using hashes here.

File details

Details for the file j_perm-1.2.0-py3-none-any.whl.

File metadata

  • Download URL: j_perm-1.2.0-py3-none-any.whl
  • Upload date:
  • Size: 42.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.11.0

File hashes

Hashes for j_perm-1.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 273ab26e3c6ff8a7d4cce276a2141190a03db6ae1391b37dc1e1442674447243
MD5 db793779a6eae50034839673973143eb
BLAKE2b-256 4335e4e735ab5ad80f98c8c548095fdfac4ef9de7c352af0775f204bd1cb68c8

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page