What in the Value ?

Rust's primary strength lies in its ability to ensure memory safety and performance guarantees at compile time, which is accomplished through its ownership and borrowing system. However, the capability for developers to manage dynamically defined data types at runtime in Rust applications is not just a pursuit, but a key to enhancing the application's adaptability and versatility, enabling it to address a diverse array of use cases.

Specific interpreted DSL languages devised in Rust utilize a runtime dynamic data type system customized for the specific language. In addition, the reputable crate Serde plays a significant role by incorporating the trait Value, thereby providing a strong sense of security and confidence to Rust developers in the management of dynamic data types at runtime. However, these solutions have limitations, as they cater to specific DSLs or the handling of JSON values.

rust_dynamic, a crate developed for the Rust language, encompasses primitives designed to assist Rust developers in managing dynamically defined data types at runtime. Presently, rust_dynamic supports the following data types:

• Integer, internally represented as i64
• Float, internally represented as f64
• Boolean
• String, internally represented as String
• Lambda functions/Curry functions and pointed on functions
• Pair, as a pair of dynamic values
• List, as a list of dynamic values
• Matrix, as a 2D collection of of vynamic values allocated in a logical grid.
• Binary, as a Vector of u8 values
• Nanosecond grade timestamp
• Any dynamically-typed object wrapped in envelope
• Complex numbers for i64 and f64
• Maps
• Associations
• None
• NODATA
• Error
• Metrics as a Vector of TIMESTAMP->F64 samples
• JSON - dynamic Value object containing JSON
• ValueMap - HashTable where both key and value of type Value
• Class - synonym to the Map
• Object - synonym to the Map

Dynamic values are wrapped and stored inside a Value structure and could be castable back into the original Rust value.

use rust_dynamic::value::Value;

let mut value = Value::from(42).unwrap();
println!("Type of the stored value is {}", &value.type_name());
println!("Dynamic value of the value is {:?}", &value.cast_integer());

What are the properties of dynamically-typed objects

• Value objects can wrap value of any supported type and they are functionally pure. Once you assigned the value, you can not change it, with one exception. Any attempts of change will result in generation of a new Value object. But you can push a new value to the object of List type, without re-creating a new object.
• All dynamically-typed objects are hash-able, so you can use them as the keys in HasMap objects.
• All dynamically-typed objects holds full information about stored data type.
• All dynamically-typed objects stores a unique object identifier.
• All dynamically-typed objects serialize-able to both JSON and a Bincode.
• You can iterate over All dynamically-typed objects.
• You can create attributes attached to wrapped data.
• Basic math operation Add/Sub/Mul/Div supported for FLOAT and INTEGER types.
• String concatenation and multiplication supported.
• LIST-type objects supports car/cdr/at/head and tail functions returning the values or the splices of the list

How to create a dynamically-typed values

rust_dynamic crate supports a number of function-primitives that will take a raw value and return a wrapped Dynamic object.

There are generic function Value::from() that will automatically cast proper data type and ether return object or error message.

How to cast Rust value from dynamically-typed values

rust_dynamic supports a number of casting functions that will try to extract wrapped value from the object that holds dynamically-typed value.

Example:

use rust_dynamic::value::Value;

// First we create a dynamically-typed value from a raw static value
let mut value = Value::from(42).unwrap();

// Then we can cast raw value back from the dynamic object
let raw_value = value.cast_integer().unwrap()

And here is an example of creating and casting JSON data

//
// First, we will create a Value of JSON type
//
let data = Value::json(serde_json::json!(42));
//
// Now, we are converting JSON values to Dynamic types
//
let value = data.cast_json_to_value()

How to set attributes to the Value

Value attributes is a Vector of the values that stored in the Value object. You can assign any number of the Value objects stored as attributes of the Value object. Attributes are serailize-able and wrap-able.

Example:

// Create object
let v = Value::from(42 as i64).unwrap()
                // Set the attributes of the object
                .attr(vec![Value::from(41.0 as f64).unwrap()])
                // And merge some extra attributes. The first attribute now have a value of 42.0
                .attr_merge(vec![Value::from(42.0 as f64).unwrap()]);

How to serialize and deserialize dynamically-typed values

There are two serialization formats that rust_dynamic presently supports: JSON and Bincode. If you are serialize to Biuncode, specail provisioning (wrapping) for objects holding JSON's will be done.

Example:

// This call will create a new dynamic value
let mut data = Value::from(42 as i64).unwrap();
// This call will serialize object to Bincode format
let bin_out = data.to_binary().unwrap();
// This will re-create an object from it's Bincode representation
let mut data2 = Value::from_binary(bin_out).unwrap();

Integration of dynamically-typed values with serde_json crate.

Crate rust_dynamic is capable to export Value object to serde_json::value::Value

Example:

// This call will create a Value object of type PAIR and on-the-fly exports it
// to the serde_json::value::Value object that can be processed as any other Value object created by serde_json crate.
let v = Value::pair(Value::from_int(1), Value::from_int(2)).as_json_value();

Functional operations with dynamically-typed values

While rust_dynamic crate is not strive to provide a full-featured functional interface to the dynamic values, some functionality that specific to a Functional programming has been implemented.

Example of mapping:

// First, we define a function which will cast value of f64 and apply f64.sin() operation
fn comp_sin(value: Value) -> Value {
    match value.data {
        Val::F64(f_val) => {
            return Value::from_float(f64::sin(f_val));
        }
        _ => return value,
    }
}
// Then we create a single value object (map could be epplied to ether list or single value object)
let mut v = Value::from(42.0).unwrap();
// Now v contains result of computation
v = v.map_value(comp_sin);

Example of binding

// First, let's create a "binding function" which will takes two Value objects and return a new Value
// In our example, simple sum() will be performed
fn sum_of_values(v1: Value, v2: Value) -> Value {
    v1 + v2
}
// Then let's create two values
let v1 = Value::from(41.0 as f64).unwrap();
let v2 = Value::from(1.0 as f64).unwrap();
// Value referred as _s_ now contains value of 42.0
let s = Value::bind_values(sum_of_values, v1, v2);

Example of maybe()

// First, we create a function that will check if v is float and is 42
fn if_value_is_42(v: &Value) -> bool {
    if v.cast_float().unwrap() == 42.0 {
        return true;
    }
    false
}
// And because it is, v is object of Value(42.0)
// Otherwise it would be Value::none()
let v = Value::from(42.0 as f64).unwrap()
        .maybe(if_value_is_42);

Example of reducing

// First, we shall create a list of the values
let mut v = Value::from_list(
    vec![Value::from_float(41.0 as f64),
    Value::from_float(1.0 as f64)]);
// Then reduce this list applying function that sums "accumulator" and current value
v = v.freduce(
    |x: Value,y: Value| -> Value { x+y },
    Value::from_float(0.0 as f64));
// This function shall return a single FLOAT value wrapping number 42.0

How to use dynamically-typed objects as HashMap keys

use std::collections::HashMap;

// This call will create a key object. Key object can be of any supported type
let key = Value::from(42.0 as f64).unwrap();

// Then we are creating a HashMap
let mut h: HashMap<Value, String> = HashMap::new();

// and store a key->value association
h.insert(key, "value".to_string());

How to iterate over dynamically-typed objects

let mut c = 0.0;
// Let's create a object of LIST type and push two elements into list
let v = Value::list()
        .push(Value::from(1.0 as f64).unwrap())
        .push(Value::from(41.0 as f64).unwrap());
// We can iterate over dynamically-typed object
for i in v {
    c += i.cast_float().unwrap();
}

How to map over dynamically-typed objects

In this example we are applying f64::sin function to all iterable values of the dynamically-typed object

let mut v = Value::from(42.0).unwrap();
v = v.map_float(f64::sin);

How to use a math operations with dynamically-typed objects

// Let's create x and y objects both of FLOAT type
let mut x = Value::from(1.0 as f64).unwrap();
let y = Value::from(41.0 as f64).unwrap();
// And perform math operation as usual
x = x + y;

How to concatenate string objects

At this moment, only STRING object supports that operation.

// Let's create x and y objects both of STRING type
let mut x = Value::from("Hello").unwrap();
// Then perform operation as usual, x shall be a object of STRING type containing string "Hello world"
let y = Value::from(" world").unwrap();
x = x + y;

How to convert from String to Float

// This call returns the object of type FLOAT
let val = Value::from("42").unwrap().conv(FLOAT).unwrap();

Functional features of rust_dynamic

Create and use of Applicative

You can create an Applicative that is wrapping a function and apply a Value object with wrapped value to Applicative. For example:

// First, as usual, we are defining function, that will be wrapped in Applicative
fn comp_sin(value: Value) -> Value {
    match value.data {
        Val::F64(f_val) => {
            return Value::from_float(f64::sin(f_val));
        }
        _ => return value,
    }
}

// Create applicative and wrap a function
let sin = Applicative::new(comp_sin);
// Then apply a Value to a wrapped function
let res = sin.apply(Value::from(42.0).unwrap());

Neural Network features of rust_dynamic

Conversion Values to Tensors

You can easily convert dynamic value into a Tensor. At this moment, we support this feature for INTEGER, FLOAT and STRING Values

let value = Value::from(42.0);
match value {
  Ok(val) => {
    match val.tensor() {
      Ok(tensor) => {
        // Here will be a tensor
      }
      Err(_) => todo!();
    }
  }
  Err(_) => todo!();
}