jmespath_extensions

Extended functions for JMESPath queries in Rust. 320+ functions for strings, arrays, dates, hashing, encoding, and more.

Built on jmespath.rs

This crate extends the jmespath crate by @mtdowling, which provides the complete Rust implementation of the JMESPath specification. All spec-compliant parsing, evaluation, and the 26 built-in functions come from that foundational library—we simply add extra functions on top.

If you only need standard JMESPath functionality, use jmespath directly.

Want to try it out? Install the jpx CLI tool:

# Homebrew (macOS/Linux)
brew install joshrotenberg/brew/jpx

# Or cargo
cargo install jpx

# Then use it!
echo '{"name": "world"}' | jpx 'upper(name)'
# "WORLD"

Non-Standard Extensions - Not Portable

This crate provides custom extension functions that are NOT part of the JMESPath specification. Queries using these functions will NOT work in other JMESPath implementations (Python, JavaScript, Go, AWS CLI, Ansible, etc.).

JMESPath Spec vs This Library

What This Means

1. Not portable: Queries using upper(), map_expr(), haversine(), etc. won't work in AWS CLI's --query, Ansible filters, or any other JMESPath implementation.

2. No spec backing: Function names, signatures, and behaviors are our decisions. While we align with JEPs where possible (items, find_first), many functions are novel.

3. Expression functions are unique: map_expr, filter_expr, sort_by_expr etc. leverage Rust runtime access—these don't exist in any JMESPath spec or implementation.

Use Cases

This library is ideal for:

• Backend data transformation: Reshape API responses, filter datasets, compute derived fields
• Configuration processing: Query and transform JSON/YAML configs with complex logic
• ETL pipelines: Extract, transform, and validate data with expressive queries
• Log/event processing: Filter and aggregate structured log data
• CLI tools: Build jq-like tools with domain-specific functions
• Embedded queries: Let users write safe, sandboxed data queries in your application

For Portable Queries

Use only the 26 standard JMESPath built-in functions: abs, avg, ceil, contains, ends_with, floor, join, keys, length, map, max, max_by, merge, min, min_by, not_null, reverse, sort, sort_by, starts_with, sum, to_array, to_number, to_string, type, values

Overview

This crate provides 320+ additional functions beyond the standard JMESPath built-ins, organized into feature-gated categories.

Full API Documentation →

Quick Start

use jmespath::Runtime;
use jmespath_extensions::register_all;

let mut runtime = Runtime::new();
runtime.register_builtin_functions();
register_all(&mut runtime);

// Now you can use extended functions in queries
let expr = runtime.compile("items[*].name | upper(@)").unwrap();

Runtime Function Registry

For applications that need runtime control over function availability (ACLs, config-based gating, introspection):

use jmespath::Runtime;
use jmespath_extensions::registry::{FunctionRegistry, Category};

let mut registry = FunctionRegistry::new();

// Register specific categories
registry.register_category(Category::String);
registry.register_category(Category::Math);

// Or register all available functions
// registry.register_all();

// Disable specific functions (e.g., for security policies)
registry.disable_function("md5");
registry.disable_function("sha256");

// Apply to runtime
let mut runtime = Runtime::new();
runtime.register_builtin_functions();
registry.apply(&mut runtime);

// Introspection - list available functions
for func in registry.functions() {
    let type_label = if func.is_standard { "standard" } else { "extension" };
    println!("[{}] {}: {}", type_label, func.name, func.description);
}

This enables:

• Runtime gating: Enable/disable functions via config instead of compile-time features
• ACL support: Disable specific functions for security policies
• Introspection: Query available functions with signatures, descriptions, examples, and whether they are standard JMESPath or extensions

jpx CLI

See jpx/README.md for full CLI documentation, or use jpx --help.

# Expression functions (the novel stuff!)
echo '{"users": [{"name": "alice", "age": 30}, {"name": "bob", "age": 25}]}' \
  | jpx 'filter_expr(users, &age > `26`) | [].name'
# ["alice"]

# Strict mode - only standard JMESPath functions
echo '[1, 2, 3]' | jpx --strict 'length(@)'
# 3

# Function discovery
jpx --list-functions           # List all 320+ functions
jpx --list-category expression # List expression functions
jpx --describe map_expr        # Detailed function info

Features

All features are opt-in. Use default-features = false to select only what you need.

Minimal Dependencies

[dependencies]
jmespath_extensions = { version = "0.2", default-features = false, features = ["core"] }

Specific Features

[dependencies]
jmespath_extensions = { version = "0.2", default-features = false, features = ["string", "array", "datetime"] }

Examples

String Manipulation

upper('hello')                    → "HELLO"
split('a,b,c', ',')               → ["a", "b", "c"]
camel_case('hello_world')         → "helloWorld"

Array Operations

first([1, 2, 3])                  → 1
unique([1, 2, 1, 3])              → [1, 2, 3]
chunk([1, 2, 3, 4], `2`)          → [[1, 2], [3, 4]]

Expression Functions

map_expr('name', users)           → ["alice", "bob"]
filter_expr('age >= `18`', users) → [{...}, {...}]
sort_by_expr('score', items)      → [{score: 1}, {score: 2}, ...]

Date/Time

now()                             → 1699900000
format_date(`0`, '%Y-%m-%d')      → "1970-01-01"
date_add(`0`, `1`, 'days')        → 86400

Fuzzy Matching

levenshtein('kitten', 'sitting')  → 3
jaro_winkler('hello', 'hallo')    → 0.88
sounds_like('Robert', 'Rupert')   → true

Geospatial

geo_distance_km(`40.7128`, `-74.0060`, `51.5074`, `-0.1278`) → 5570.2
geo_bearing(`40.7128`, `-74.0060`, `51.5074`, `-0.1278`)     → 51.2

Network

cidr_contains('192.168.0.0/16', '192.168.1.1') → true
is_private_ip('10.0.0.1')                      → true

JSON Patch (RFC 6902/7386)

json_patch({a: 1}, [{op: 'add', path: '/b', value: 2}])  → {a: 1, b: 2}
json_merge_patch({a: 1}, {b: 2})                        → {a: 1, b: 2}
json_diff({a: 1}, {a: 2})                               → [{op: 'replace', path: '/a', value: 2}]

Multi-Pattern Matching (Aho-Corasick)

match_any('hello world', ['world', 'foo'])              → true
match_all('hello world', ['hello', 'world'])            → true
match_which('hello world', ['hello', 'foo', 'world'])   → ["hello", "world"]
match_count('abcabc', ['a', 'b'])                       → 4
replace_many('hello world', {hello: 'hi', world: 'earth'}) → "hi earth"

See the API documentation for complete function reference with examples.

JMESPath Community JEP Alignment

This crate aligns with several JMESPath Enhancement Proposals:

• JEP-014 (String Functions): lower, upper, trim, trim_left, trim_right, pad_left, pad_right, replace, split, find_first, find_last
• JEP-013 (Object Functions): items, from_items, zip

Functions that align with JEPs have jep: Some("JEP-XXX") in their FunctionInfo metadata, accessible via the registry.

Additional functions extend well beyond these proposals. Some JEPs (like arithmetic operators) require parser changes and cannot be implemented as extension functions.

Development

Adding a New Function

1. Add metadata to jmespath_extensions/functions.toml:

   [[functions]]
   name = "my_function"
   category = "string"  # Must match an existing category
   description = "Brief description of what it does"
   signature = "string, number -> string"
   example = "my_function('hello', `3`) -> 'hellohellohello'"
   features = ["string"]  # Feature flags that enable this function
   # Optional fields:
   # jep = "JEP-014"      # If aligned with a JMESPath Enhancement Proposal
   # aliases = ["my_func"] # Alternative names
   

2. Implement the function in the appropriate module (e.g., src/string.rs):

   define_function! {
       name = "my_function";
       doc = "Brief description of what it does.";
       args = [value: Value, count: usize];
       fn run(value, count) {
           // Implementation
           let s = value.as_string().ok_or_else(|| /* error */)?;
           Ok(Rc::new(Variable::String(s.repeat(count))))
       }
   }
   

3. Register the function in the module's register_* function:

   pub fn register_string(runtime: &mut Runtime) {
       // ... existing registrations
       runtime.register_function(Box::new(MyFunction));
   }
   

4. Run the build to regenerate documentation:

   cargo build  # build.rs generates docs from functions.toml
   cargo test   # Verify everything works
   cargo clippy --all-features
   

Adding a New Feature/Category

1. Add the feature to Cargo.toml:

   [features]
   my_category = ["dep:some-crate"]  # If it needs dependencies
   full = ["my_category", ...]       # Add to full feature
   

2. Create the module src/my_category.rs with the standard structure

3. Add to src/lib.rs:

   #[cfg(feature = "my_category")]
   pub mod my_category;
   

4. Add the category variant to Category enum in src/registry.rs

5. Update build.rs to handle the new category in category_variant()

Contribution Guidelines

When proposing new functions, please ensure they meet these criteria:

1. Generally useful: Functions should solve common problems that many users encounter. Avoid highly specialized or niche use cases.

2. Minimal dependencies: Prefer zero dependencies. If a dependency is needed, limit to one well-maintained crate. Dependencies should be feature-gated.

3. Thoughtful naming: Function names should be:

   • Clear and descriptive
   • General enough to not imply overly specific behavior
   • Consistent with existing naming conventions in the category

4. Fit existing categories: New functions should naturally belong to an existing feature/category. Proposing a new category requires strong justification and multiple related functions.

5. No duplicate functionality: Check that the function doesn't duplicate existing functionality or can be trivially composed from existing functions.

6. Include tests and examples: All new functions must include tests and a working example in functions.toml.

Benchmarks

Run benchmarks with:

cargo bench --all-features

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.