Best Practices: Structs vs Inline Components

edit Introduced: 5.0

Best Practices: Structs vs Inline Components

Performance guide for choosing between structs and components. Structs are ~16x faster but components offer type safety and validation.

Note: These performance characteristics apply to all component types (CFCs, sub-components, inline components).

Basic Examples

Struct Approach

// Simple data container
personData = {
    personId: 123,
    firstName: "Susi",
    lastName: "Sorglos",
    email: "susi.sorglos@lucee.org"
};

// Access data directly
fullName = personData.firstName & " " & personData.lastName;

Inline Component Approach

// Type-safe data container
personBean = new component accessors=true {
    property name="personId" type="numeric";
    property name="firstName" type="string";
    property name="lastName" type="string";
    property name="email" type="string";

    function getFullName() {
        return getFirstName() & " " & getLastName();
    }
};

personBean.setPersonId(123);
personBean.setFirstName("Susi");
personBean.setLastName("Sorglos");

Bean Factory Approach

// Custom bean structure with getters/setters
function createBean(boolean readOnly=false) {
    var bean = [:];
    var args = arguments;

    // Remove readOnly from arguments
    var readOnly = arguments.readOnly;
    structDelete(arguments, "readOnly");

    loop struct=arguments index="local.k" item="local.v" {
        // Create "private" value
        bean["_" & k] = v;

        // Getter - uses function name to determine property
        bean["get" & ucFirst(k)] = function() {
            var key = mid(getPageContext().getActiveUDFCalledName(), 4);
            return bean["_" & key];
        };

        // Setter (if not read-only) - uses function name to determine property
        if (!readOnly) {
            bean["set" & ucFirst(k)] = function(val) {
                var key = mid(getPageContext().getActiveUDFCalledName(), 4);
                bean["_" & key] = val;
            };
        }
    }
    return bean;
}

// Usage
personBean = createBean(
    readOnly: false,
    personId: 123,
    firstName: "Susi",
    lastName: "Sorglos",
    email: "susi.sorglos@lucee.org"
);

dump(personBean.getFirstName());
dump(personBean.getLastName());

Common Use Cases

High-Frequency Object Creation

// Recommended: Structs for query loops and bulk processing
loop query=qryPersons {
    personRecord = {
        id: qryPersons.id,
        fullName: qryPersons.firstName & " " & qryPersons.lastName,
        email: qryPersons.email,
        status: "active"
    };
    processPerson(personRecord);
}

// Bean factory approach: Better performance than components, no metaspace overhead
loop query=qryPersons {
    personBean = createBean(
        readOnly: true,
        id: qryPersons.id,
        firstName: qryPersons.firstName,
        lastName: qryPersons.lastName,
        email: qryPersons.email
    );
    processPerson(personBean);
}

// Component approach: Slower and uses more memory
loop query=qryPersons {
    personRecord = new component accessors=true {
        property name="id" type="numeric";
        property name="firstName" type="string";
        property name="lastName" type="string";
        property name="email" type="string";
    };
    // Set values and process
}

// Avoid: Components in high-frequency scenarios
loop query=qryPersons {
    personRecord = new component accessors=true {
        property name="id" type="numeric";
        property name="fullName" type="string";
        property name="email" type="string";
        property name="status" type="string";
    };
}

Configuration Objects

// Appropriate: Components for configuration with validation
config = new component accessors=true {
    property name="host" type="string" required=true;
    property name="port" type="numeric" default=3306;
    property name="timeout" type="numeric" default=30000;

    function validate() {
        if (!len(trim(getHost()))) {
            throw("Host is required");
        }
        return true;
    }
};

Simple Data Transfer

// Recommended: Structs for API responses and data transfer
function getPersonData(personId) {
    return {
        personId: personId,
        firstName: getFirstName(personId),
        lastName: getLastName(personId),
        email: getEmail(personId),
        lastLogin: getLastLogin(personId)
    };
}

Performance Testing

Basic Benchmark

function performanceTest() {
    iterations = 100000;

    // Query creation
    var qryPersons = queryNew("id,firstName,lastName,email");
    loop from=1 to=iterations index="local.idx" {
        var row = queryAddRow(qryPersons);
        querySetCell(qryPersons, "id", idx, row);
        querySetCell(qryPersons, "firstName", "Susi" & idx, row);
        querySetCell(qryPersons, "lastName", "Sorglos" & idx, row);
        querySetCell(qryPersons, "email", "Susi" & idx & "@lucee.org", row);
    }

    // Bean factory function
    function createBean(boolean readOnly=false) {
        var bean = [:];
        var args = arguments;
        var readOnly = arguments.readOnly;
        structDelete(arguments, "readOnly");

        loop struct=arguments index="local.k" item="local.v" {
            bean["_" & k] = v;
            bean["get" & ucFirst(k)] = function() {
                var key = mid(getPageContext().getActiveUDFCalledName(), 4);
                return bean["_" & key];
            };
            if (!readOnly) {
                bean["set" & ucFirst(k)] = function(val) {
                    var key = mid(getPageContext().getActiveUDFCalledName(), 4);
                    bean["_" & key] = val;
                };
            }
        }
        return bean;
    }

    // Test struct creation
    start = getTickCount();
    loop query=qryPersons {
        x = {
            id: qryPersons.id,
            firstName: qryPersons.firstName,
            lastName: qryPersons.lastName,
            email: qryPersons.email
        };
    }
    structTime = getTickCount() - start;

    // Test bean factory creation
    start = getTickCount();
    loop query=qryPersons {
        x = createBean(
            readOnly: false,
            id: qryPersons.id,
            firstName: qryPersons.firstName,
            lastName: qryPersons.lastName,
            email: qryPersons.email
        );
    }
    beanTime = getTickCount() - start;

    // Test component creation
    start = getTickCount();
    loop query=qryPersons {
        x = new component accessors=true {
            property name="id" type="numeric";
            property name="firstName" type="string";
            property name="lastName" type="string";
            property name="email" type="string";
        };
    }
    componentTime = getTickCount() - start;

    echo("Struct time: " & structTime & "ms<br>");
    echo("Bean factory time: " & beanTime & "ms<br>");
    echo("Component time: " & componentTime & "ms<br>");
    echo("Bean vs Struct ratio: " & (beanTime/structTime) & "x<br>");
    echo("Component vs Struct ratio: " & (componentTime/structTime) & "x<br>");
}

Performance Analysis Results

Based on testing with 100,000 iterations (multiple runs):

Execution Time Comparison

Struct creation: ~20ms (19-23ms range)
Bean factory creation: ~320ms (315-347ms range)
Inline component creation: ~335ms (328-336ms range)
Performance impact: Bean factories are ~16x slower than structs, components are ~17x slower than structs

Memory Usage Patterns

Struct Approach:

Minimal memory overhead
No class loading
Lightweight objects
Fast garbage collection

Bean Factory Approach:

Moderate memory overhead (function closures only)
No class loading required
No metaspace allocation
Slightly faster than components on average

Inline Component Approach:

Creates physical class definitions
Metaspace allocation for class metadata
Component lifecycle overhead
Highest memory consumption due to class creation

Key Findings

Performance: Structs are dramatically faster (16-17x) than both bean factories and components
Bean vs Component: Bean factories are slightly faster than components and have less memory overhead
Memory Efficiency: Bean factories avoid class loading and metaspace allocation
Closure vs Class Cost: Function closures are less expensive than class creation overhead
Universal Impact: These performance characteristics apply to all component types in Lucee

Recommendations

Use Structs When:

High-frequency object creation
Performance is critical
Simple data containers
Temporary objects
API responses
Memory optimization needed

Use Inline Components When:

Type safety is important
Complex validation required
Long-lived objects
Configuration management
Code clarity is priority

Note: The same performance considerations apply when choosing between structs and regular components (CFCs) or sub-components.

Use components of any type when the benefits of type safety, validation, and encapsulation outweigh the performance costs.

Testing Guidelines

For accurate performance testing:

Use fresh JVM instances when possible
Run multiple iterations
Consider garbage collection impact
Test with realistic data sizes
Monitor both time and memory usage
Run at least 30k rounds to warmup, before benchmarkings, to allow the JVM optimizations to be applied

Summary

Choose structs for performance-critical scenarios and simple data containers.

Use inline components when type safety and validation are more important than raw performance.

Consider hybrid approaches that combine the benefits of both.