Skillet — “Lightning-fast formulas, Rust-powered.”

Skillet is a high-performance, embeddable expression engine written in Rust, inspired by Excel formulas with Ruby-style method chaining. It parses expressions into an AST and evaluates them with an optimized runtime.

✨ New Features:

• Ruby-style Type Conversion Methods: null.to_s(), "123".to_i(), [1,2,3].to_bool() - available on all types
• Safe Navigation Operator: obj&.property&.method() - prevents null reference errors
• String Helpers: SUBSTITUTE(text, substr, replacement), SUBSTITUTEM(text, substr, replacement) and Excel-style REPLACE(old_text, start_num, num_chars, new_text)
• JSON Digging: DIG(json, ['path','to','key'], default) and method form json.dig([...], default)
• Enhanced Null Safety: Conversion methods provide safe defaults for null values
• Performance Optimized: ~3ms evaluation time (100x+ improvement from original 300ms)

Core Features

• 🚀 Lightning Fast: Optimized parser with string interning and memory pooling
• 🛡️ Null Safe: Safe navigation (&.) and conversion methods handle null gracefully
• 🔧 Extensible: JavaScript plugins for runtime extensibility without recompilation
• 📊 Excel-like: Familiar syntax with advanced features like array operations
• 🦀 Rust-powered: Memory safe with zero-cost abstractions
• 🎯 Type Smart: Ruby-style conversions with automatic type coercion

Supported Types: Numbers, strings, booleans, nulls, arrays, JSON objects, dates, currency Operations: Arithmetic, logical, comparisons, method chaining, array operations, lambdas Extensions: JavaScript plugins, Rust custom functions, HTTP/TCP server modes

📚 Full Documentation | 📖 API Reference

Build

• Requirements: Rust stable (2021 edition)
• Build and test:

cargo build
cargo test

Quick Examples

Traditional Excel-style formulas:

cargo run --bin sk -- "SUM(1, 2, 3, 4, 5)"                    # 15
cargo run --bin sk -- "IF(10 > 5, \"Yes\", \"No\")"           # "Yes"
cargo run --bin sk -- "AVERAGE([85, 92, 78, 90])"             # 86.25

✨ New: Null-safe operations with conversion methods:

cargo run --bin sk -- "null.to_s().length()"                  # 0 (no error!)
cargo run --bin sk -- "\"123\".to_i() + 10"                   # 133
cargo run --bin sk -- "[null, \"hello\"].map(:x.to_s())"      # ["", "hello"]

✨ New: Safe navigation operator:

cargo run --bin sk -- ":data := {\"name\": null}; :data&.name&.length()"  # null (no error!)

✨ New: String helpers and JSON dig:

# SUBSTITUTE replaces all occurrences of a substring
cargo run --bin sk -- "SUBSTITUTE('foo bar foo', 'foo', 'baz')"      # "baz bar baz"

# SUBSTITUTEM: same as SUBSTITUTE (replace multiple occurrences)
cargo run --bin sk -- "SUBSTITUTEM('a-a-a', '-', '_')"               # "a_a_a"

# REPLACE: positional, 1-based start
cargo run --bin sk -- "REPLACE('abcdef', 3, 2, 'XY')"                 # "abXYef"

# DIG: navigate JSON safely with a path (arrays supported)
cargo run --bin sk -- ":obj := {\"user\": {\"posts\": [{\"title\": \"First\"}]}}; DIG(:obj, ['user','posts',0,'title'])"  # "First"

# Method form
cargo run --bin sk -- ":obj := {\"user\": {\"name\": \"Jane\"}}; :obj.dig(['user','name'])"   # "Jane"

Advanced array operations:

cargo run --bin sk -- "[30,60,80,100].filter(:x > 50).map(:x * 0.9).sum()"  # 216

Notes:

• Wrap expressions in quotes in your shell
• A leading = is optional (supported for Excel-style familiarity)

Library Usage

Add to your Cargo project (from crates.io):

[dependencies]
skillet = "0.5.2"

Or with cargo-edit:

cargo add skillet@0.4.1

Server Modes

Skillet includes production-ready HTTP and TCP servers for high-performance expression evaluation.

🌐 HTTP Server (sk_http_server)

Run the HTTP server for REST API access:

# Basic usage
cargo run --bin sk_http_server 5074

# Production deployment
cargo run --bin sk_http_server 5074 --host 0.0.0.0 --token your_secret_token

# Background daemon
cargo run --bin sk_http_server 5074 -d --host 0.0.0.0 --token secret123 --admin-token admin456

Parameters:

• <port> - Port to bind (required)
• -H, --host <addr> - Bind address (default: 127.0.0.1)
• -d, --daemon - Run as background daemon
• --token <value> - Require token for eval requests
• --admin-token <value> - Require admin token for JS function management
• --pid-file <file> - PID file for daemon mode
• --log-file <file> - Log file for daemon mode

HTTP Endpoints:

• GET /health - Health check
• GET / - API documentation
• POST /eval - Evaluate expressions (JSON body)
• GET /eval?expr=... - Evaluate expressions (query params)
• POST /js/functions - Upload JavaScript functions (admin)
• GET /js/functions - List JavaScript functions
• DELETE /js/functions/{name} - Delete JavaScript function (admin)

Example API calls:

# Basic evaluation
curl -X POST http://localhost:5074/eval \
  -H "Content-Type: application/json" \
  -d '{"expression": "2 + 3 * 4"}'

# With variables and null safety
curl -X POST http://localhost:5074/eval \
  -H "Content-Type: application/json" \
  -d '{
    "expression": ":data.filter(:x.value.to_s().length() > 0)",
    "arguments": {
      "data": [{"value": null}, {"value": "hello"}, {"value": ""}]
    },
    "include_variables": true
  }'

# GET request with query params
curl "http://localhost:5074/eval?expr=SUM(1,2,3,4,5)"

# With authentication
curl -X POST http://localhost:5074/eval \
  -H "Authorization: Bearer your_secret_token" \
  -H "Content-Type: application/json" \
  -d '{"expression": "null.to_s().length()"}'

⚡ TCP Server (sk_server)

High-performance TCP server for custom protocol access:

# Basic usage
cargo run --bin sk_server 8080

# With worker threads
cargo run --bin sk_server 8080 16

# Production daemon
cargo run --bin sk_server 8080 8 -d --host 0.0.0.0 --token secret123

Parameters:

• <port> - Port to bind (required)
• [num_threads] - Worker threads (optional)
• -H, --host <addr> - Bind address (default: 127.0.0.1)
• -d, --daemon - Run as background daemon
• --token <value> - Require authentication token
• --pid-file <file> - PID file for daemon mode
• --log-file <file> - Log file for daemon mode

Protocol: JSON-based TCP protocol for maximum performance

Library Usage

Evaluate expressions:

use skillet::{evaluate, evaluate_with, Value};
use std::collections::HashMap;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Numeric
    let v = evaluate("= 2 + 3 * 4")?; // -> Value::Number(14.0)

    // Variables
    let mut vars = HashMap::new();
    vars.insert("sales".to_string(), Value::Number(5000.0));
    let v = evaluate_with("=SUM(:sales, 1000)", &vars)?; // -> 6000.0

    // Strings and chaining
    let v = evaluate("= '  john  '.trim().upper()")?; // -> "JOHN"

    // Arrays + F/M/R
    let v = evaluate("= [30,60,80,100].filter(:x > 50).map(:x * 0.9).sum()")?; // -> 216.0

    // Named lambda parameters
    let v = evaluate("= FILTER([1,2,3,4], :n % 2 == 0, 'n')")?; // -> [2,4]

    // Type casting
    let v = evaluate("= '42'::Integer")?; // -> 42
    Ok(())
}

Language Features (MVP)

• Numbers, booleans (TRUE/FALSE), strings ('...' or "..."), NULL
• Operators: + - * / % ^, > < >= <= == !=, AND/OR/NOT (also &&/||/!), ternary ? :
• Variables: :name (provided via evaluate_with map)
• Functions (subset):
  • Math: SUM, AVG/AVERAGE, MIN, MAX, ROUND, CEIL, FLOOR, ABS, SQRT, POW
  • Arrays: ARRAY, FIRST, LAST, CONTAINS, UNIQUE, SORT, REVERSE, JOIN, FLATTEN
  • Strings: CONCAT, UPPER, LOWER, TRIM, LENGTH, SPLIT, SUBSTITUTE, REPLACE
  • JSON: DIG(json, path_array, [default])
  • Logic: ISBLANK
  • Functional: FILTER(array, expr, [param]), MAP(array, expr, [param]), REDUCE(array, expr, initial, [valParam], [accParam])
  • Conditional aggregations: SUMIF(array, expr), AVGIF(array, expr), COUNTIF(array, expr)
• Methods (subset): chaining with . and predicates ?
  • Numbers: .abs() .round(n) .floor() .ceil(); predicates .positive? .negative? .zero? .even? .odd? .numeric?
  • Arrays: .length() .size() .first() .last() .sum() .avg() .min() .max() .sort() .unique() .reverse() .compact() .flatten()
  • Strings: .upper() .lower() .trim() .reverse()
• Arrays: literals [1, 2, 3]; indexing arr[i] (negatives allowed); slicing arr[a:b]
• Spread: ...expr inside arg lists
• Casting: expr::Integer|Float|String|Boolean|Array|Currency|DateTime|Json

Examples

• Arithmetic precedence: = 2 + 3 * 4 → 14
• Ternary: = :score >= 90 ? 'A' : 'B'
• Named lambda param: = [1,2,3,4].map(:v * 10, 'v') → [10,20,30,40]
• Reduce with named params: = [1,2,3].reduce(:a + :v, 0, 'v', 'a') → 6
• PRODUCT: = PRODUCT(2, 3, 4) → 24
• SUMIF: = SUMIF([1,-2,3,-4], :x > 0) → 4 (lambda-style)
• SUMIF Excel-style: = SUMIF([10,20,30,40], ">25") → 70
• FLATTEN: = FLATTEN([1,[2,[3]],4]) → [1,2,3,4]

Notes

• This is an MVP; error messages and type coverage are intentionally minimal.
• For variables beyond numbers/strings/arrays (e.g., dates, currency), see Value in src/types.rs.

Install Binaries

If you want the binaries such as sk, sk_server and sk_client installed system-wide:

cargo install skillet

Server Mode

Skillet includes a high-performance evaluation server that keeps the interpreter warm and eliminates per-process overhead.

• Start the server: sk_server 8080 (binds to 127.0.0.1:8080)
• Daemonize (Unix): sk_server 8080 -d (writes PID to skillet-server.pid in CWD)
• Stop daemon: kill $(cat skillet-server.pid)
• Bind host/IP: sk_server 8080 --host 0.0.0.0 (listen on all interfaces)
• Optional token auth: sk_server 8080 --host 0.0.0.0 --token <secret> (or set SKILLET_AUTH_TOKEN)

Client and benchmarks:

• One-off eval: sk_client localhost:8080 '=2+3*4'
• With variables: sk_client localhost:8080 '=SUM(:a,:b)' a=10 b=5
• JSON vars: sk_client localhost:8080 '=:user.name' --json '{"user":{"name":"Alice"}}'
• Benchmark: sk_client localhost:8080 --benchmark '=2+3*4' 10000
• With token: sk_client localhost:8080 '=2+3*4' --token <secret> (or set SKILLET_SERVER_TOKEN)

Scripts:

• Build + run multi-test benchmark: bash scripts/benchmark_server.sh [port] [iterations] [threads]

take a look at the Server Usage Guide for more details about how to use it and consume in different langauages

Built-in Functions

• Arithmetic: SUM, PRODUCT/MULTIPLY, AVG/AVERAGE, MIN, MAX, ROUND, CEIL, CEILING, FLOOR, ABS, SQRT, POW/POWER, MOD, INT
• Logical: AND, OR, NOT, XOR, IF, IFS
• String: LENGTH, CONCAT, UPPER, LOWER, TRIM, SUBSTRING, SPLIT, REPLACE, REVERSE, ISBLANK, ISNUMBER, ISTEXT
• Array: ARRAY, FLATTEN, FIRST, LAST, CONTAINS, IN, COUNT, UNIQUE, SORT, REVERSE, JOIN
• Date/Time: NOW, DATE, TIME, YEAR, MONTH, DAY, DATEADD, DATEDIFF
• Financial: PMT, DB, FV, IPMT
• Statistical: MEDIAN, MODE.SNGL (MODESNGL, MODE_SNGL), STDEV.P (STDEVP, STDEV_P), VAR.P (VARP, VAR_P), PERCENTILE.INC (PERCENTILEINC, PERCENTILE_INC), QUARTILE.INC (QUARTILEINC, QUARTILE_INC)
• Functional: FILTER(array, expr, [param]), MAP(array, expr, [param]), REDUCE(array, expr, initial, [valParam], [accParam]), SUMIF(array, expr_or_criteria [, sum_array]), AVGIF(array, expr), COUNTIF(array, expr)

API Surface (Rust)

• parse(input: &str) -> Result<Expr, Error>: parse into AST
• evaluate(input: &str) -> Result<Value, Error>: evaluate without variables
• evaluate_with(input: &str, vars: &HashMap<String, Value>) -> Result<Value, Error>
• evaluate_with_json(input: &str, json_vars: &str) -> Result<Value, Error>
• evaluate_with_custom(input: &str, vars: &HashMap<String, Value>) -> Result<Value, Error>
• evaluate_with_json_custom(input: &str, json_vars: &str) -> Result<Value, Error>
• Custom functions:
  • register_function(Box<dyn CustomFunction>) -> Result<(), Error>
  • unregister_function(name: &str) -> bool
  • list_custom_functions() -> Vec<String>
• Types:

  • Value enum: Number(f64) | Array(Vec<Value>) | Boolean(bool) | String(String) | Null | Currency(f64) | DateTime(i64) | Json(String)

  • Error with message and optional position

Tests

Run the test suite:

cargo test

Author

@zenbakiak

Github repo

License

MIT OR Apache-2.0