modkit-sdk

Security Context Scoping and Typed OData Query Builder for Clients

Overview

This crate provides two main features:

1. Security Context Scoping: A lightweight, zero-allocation wrapper that binds a SecurityContext to any client type
2. Typed OData Query Builder: A generic, reusable query builder that produces type-safe OData queries with automatic filter hashing

Security Context Scoping

Example

use modkit_sdk::WithSecurityContext;
use modkit_security::SecurityContext;

let client = MyClient::new();
let ctx = SecurityContext::root();

// Bind the security context to the client
let secured = client.security_ctx(&ctx);

// Access the client and context
let client_ref = secured.client();
let ctx_ref = secured.ctx();

Typed OData Query Builder

The typed OData query builder provides a type-safe way to construct OData queries without manually building ODataQuery instances. It ensures compile-time type checking for field references and operations.

Features

• Type-safe field references: Field types are checked at compile time
• Schema trait: Define field enums and their string mappings
• Typed filter constructors: Comparison and string operations with type safety
• Fluent API: Chain methods to build complex queries
• Automatic filter hashing: Stable, deterministic hashing for cursor pagination
• No proc macros: Pure Rust implementation without code generation

Quick Start

1. Define Your Schema

use modkit_sdk::odata::{Schema, FieldRef};

// Define field enum
#[derive(Copy, Clone, Eq, PartialEq)]
enum UserField {
    Id,
    Name,
    Email,
    Age,
}

// Define schema struct
struct UserSchema;

// Implement Schema trait
impl Schema for UserSchema {
    type Field = UserField;

    fn field_name(field: Self::Field) -> &'static str {
        match field {
            UserField::Id => "id",
            UserField::Name => "name",
            UserField::Email => "email",
            UserField::Age => "age",
        }
    }
}

2. Create Typed Field References

// Define typed field constants
const ID: FieldRef<UserSchema, uuid::Uuid> = FieldRef::new(UserField::Id);
const NAME: FieldRef<UserSchema, String> = FieldRef::new(UserField::Name);
const EMAIL: FieldRef<UserSchema, String> = FieldRef::new(UserField::Email);
const AGE: FieldRef<UserSchema, i32> = FieldRef::new(UserField::Age);

3. Build Queries

use modkit_sdk::odata::{QueryBuilder, FilterExpr};
use modkit_odata::SortDir;

// Simple equality filter
let user_id = uuid::Uuid::new_v4();
let query = QueryBuilder::<UserSchema>::new()
    .filter(ID.eq(user_id))
    .build();

// Complex filter with AND/OR
let query = QueryBuilder::<UserSchema>::new()
    .filter(
        AGE.ge(18)
            .and(AGE.le(65))
            .and(NAME.contains("smith"))
    )
    .order_by(NAME, SortDir::Asc)
    .page_size(50)
    .build();

// Full query with all features
let query = QueryBuilder::<UserSchema>::new()
    .filter(ID.eq(user_id).and(AGE.gt(18)))
    .order_by(NAME, SortDir::Asc)
    .order_by(AGE, SortDir::Desc)
    .select([NAME, EMAIL])
    .page_size(25)
    .build();

Supported Operations

Comparison Operators (All Field Types)

• eq(value) - Equality: field eq value
• ne(value) - Not equal: field ne value
• gt(value) - Greater than: field gt value
• ge(value) - Greater or equal: field ge value
• lt(value) - Less than: field lt value
• le(value) - Less or equal: field le value

String Operations (String Fields Only)

• contains(value) - Contains: contains(field, 'value')
• startswith(value) - Starts with: startswith(field, 'value')
• endswith(value) - Ends with: endswith(field, 'value')

Logical Combinators

• and(expr) - Logical AND: expr1 and expr2
• or(expr) - Logical OR: expr1 or expr2
• not() - Logical NOT: not expr

Query Builder Methods

• filter(expr) - Set the filter expression
• order_by(field, dir) - Add an order-by clause (can be called multiple times)
• select(fields) - Set field projection (pass &[&field1, &field2, ...])
• page_size(limit) - Set the page size limit
• build() - Build the final ODataQuery with computed filter hash

Type Safety

The query builder enforces type safety at compile time:

// ✅ Correct: String field with string operations
let query = QueryBuilder::<UserSchema>::new()
    .filter(NAME.contains("john"))
    .build();

// ❌ Compile error: contains() only available for String fields
let query = QueryBuilder::<UserSchema>::new()
    .filter(AGE.contains("test"))  // Won't compile!
    .build();

// ✅ Correct: Comparison operations work on all types
let query = QueryBuilder::<UserSchema>::new()
    .filter(AGE.gt(18))
    .build();

Filter Hash Stability

The query builder automatically computes a stable, deterministic hash for filter expressions using the same algorithm as modkit_odata::pagination::short_filter_hash. This ensures cursor pagination consistency:

let user_id = uuid::Uuid::new_v4();

let query1 = QueryBuilder::<UserSchema>::new()
    .filter(ID.eq(user_id))
    .build();

let query2 = QueryBuilder::<UserSchema>::new()
    .filter(ID.eq(user_id))
    .build();

// Same filter produces same hash
assert_eq!(query1.filter_hash, query2.filter_hash);

Supported Value Types

The following Rust types can be used in filter expressions:

• bool
• uuid::Uuid
• String and &str
• i32, i64, u32, u64

Additional types can be supported by implementing the IntoODataValue trait.

Examples

See examples/typed_odata_query.rs for examples demonstrating:

• Simple equality filters
• String operations (contains, startswith, endswith)
• Complex filters with AND/OR/NOT
• Ordering and field selection
• Page size limits
• Full queries with all features
• Filter hash stability

Run the example:

cargo run --package modkit-sdk --example typed_odata_query

Design Constraints

• No proc macros: Pure Rust implementation without code generation
• Small footprint: Minimal API surface with no large facades
• No DB concepts: Does not expose database or SeaORM concepts
• AST-based: Produces modkit_odata::ast::Expr for maximum flexibility
• Stable hashing: Deterministic filter hashing for cursor pagination

Testing

The query builder includes comprehensive unit tests verifying:

• Field name mapping works correctly
• Building queries sets order/limit/filter properly
• Filter hash is stable for identical filters
• All comparison and string operations work
• Logical combinators (AND/OR/NOT) function correctly
• Field selection handles heterogeneous field types

Run tests:

cargo test --package modkit-sdk --lib odata

License

See workspace license.