Elastic.Mapping 0.4.0

Elastic.Mapping

Compile-time Elasticsearch mappings for .NET. Native AOT ready. Define your index mappings, analysis chains, and field metadata with C# attributes and get reflection-free generated code that works with System.Text.Json source generation out of the box.

Why?

Elasticsearch field names are strings. Typos are silent. Refactors break queries. Manual JSON mappings drift from your code.

Elastic.Mapping fixes this with a source generator that turns your POCOs into type-safe, pre-computed mapping infrastructure at build time -- zero reflection, zero runtime overhead, fully AOT compatible.

Quick Start

1. Define your domain types (clean POCOs)

public class Product
{
    [Keyword]
    public string Id { get; set; }

    [Text(Analyzer = "standard")]
    public string Name { get; set; }

    public double Price { get; set; }

    public bool InStock { get; set; }

    [Nested]
    public List<Category> Categories { get; set; }
}

2. Register them in a mapping context

[ElasticsearchMappingContext]
[Index<Product>(Name = "products", SearchPattern = "products*")]
[DataStream<ApplicationLog>(Type = "logs", Dataset = "myapp", Namespace = "production")]
public static partial class MyContext;

3. Use generated field constants and metadata

// Type-safe field names -- rename the C# property, these update automatically
MyContext.Product.Fields.Name      // "name"
MyContext.Product.Fields.Price     // "price"
MyContext.Product.Fields.InStock   // "inStock"

// Index targets
MyContext.Product.IndexStrategy.WriteTarget   // "products"
MyContext.Product.SearchStrategy.Pattern      // "products*"

// Data stream naming follows Elastic conventions
MyContext.ApplicationLog.IndexStrategy.DataStreamName  // "logs-myapp-production"

// Pre-built JSON for index creation
var json = MyContext.Product.Context.GetIndexJson();

// Change detection -- only update when mappings actually change
if (clusterHash != MyContext.Product.Hash) UpdateMappings();

System.Text.Json Integration

Elastic.Mapping is built around System.Text.Json. Link your STJ source-generated JsonSerializerContext and the mapping generator inherits your serialization configuration automatically -- one source of truth for both JSON serialization and Elasticsearch field names.

// Your STJ source-generated context
[JsonSourceGenerationOptions(
    PropertyNamingPolicy = JsonKnownNamingPolicy.CamelCase,
    UseStringEnumConverter = true,
    DefaultIgnoreCondition = JsonIgnoreCondition.WhenWritingNull)]
[JsonSerializable(typeof(Product))]
[JsonSerializable(typeof(Order))]
public partial class MyJsonContext : JsonSerializerContext;

// Link it to the mapping context
[ElasticsearchMappingContext(JsonContext = typeof(MyJsonContext))]
[Index<Product>(Name = "products")]
[Index<Order>(Name = "orders")]
public static partial class MyContext;

The generator reads [JsonSourceGenerationOptions] at compile time and applies:

Per-property [JsonPropertyName("custom_name")] and [JsonIgnore] attributes are always respected, with or without a linked context.

This means your Elasticsearch field names, your JSON wire format, and your C# properties all stay in sync -- at compile time, with no reflection.

Native AOT

Every feature in Elastic.Mapping is AOT compatible:

• No reflection at runtime -- all field names, mappings JSON, and type metadata are generated as constants at compile time
• No dynamic code generation -- source generators run during build, not at runtime
• Pre-computed JSON -- settings and mappings are embedded as string literals, ready to send to Elasticsearch
• STJ source generation -- link your JsonSerializerContext for a fully AOT serialization pipeline

Publish with dotnet publish -p:PublishAot=true and everything works.

Field Type Attributes

Control how properties map to Elasticsearch field types:

Properties without attributes are inferred from their CLR type (string → keyword, int → integer, DateTime → date, etc.).

Analysis Configuration

Build custom analyzers, tokenizers, and filters with a fluent API:

public static class ProductConfig
{
    public static AnalysisBuilder ConfigureAnalysis(AnalysisBuilder analysis) => analysis
        .Analyzer("product_search", a => a.Custom()
            .Tokenizer(BuiltIn.Tokenizers.Standard)
            .Filter(BuiltIn.TokenFilters.Lowercase, "english_stemmer", "edge_ngram_3_8"))
        .TokenFilter("english_stemmer", f => f.Stemmer()
            .Language(BuiltIn.StemmerLanguages.English))
        .TokenFilter("edge_ngram_3_8", f => f.EdgeNGram()
            .MinGram(3).MaxGram(8));

    public static ProductMappingsBuilder ConfigureMappings(ProductMappingsBuilder m) => m
        .Name(f => f.Analyzer("product_search")
            .MultiField("keyword", mf => mf.Keyword().IgnoreAbove(256)));
}

// Reference it from the context attribute:
[Index<Product>(Name = "products", Configuration = typeof(ProductConfig))]

The generated ProductAnalysis class gives you type-safe constants for your custom components:

MyContext.Product.Analysis.Analyzers.ProductSearch     // "product_search"
MyContext.Product.Analysis.TokenFilters.EnglishStemmer  // "english_stemmer"

Index Strategies

Traditional Index

[Index<Product>(
    Name = "products",
    WriteAlias = "products-write",
    ReadAlias = "products-read",
    SearchPattern = "products*",
    Shards = 3,
    RefreshInterval = "5s"
)]

Rolling Date Index

[Index<Order>(Name = "orders", DatePattern = "yyyy.MM")]
// Write target: orders-2025.02
// Search pattern: orders-*

Data Stream (logs, metrics, traces)

[DataStream<ApplicationLog>(Type = "logs", Dataset = "ecommerce.app", Namespace = "production")]
// Data stream: logs-ecommerce.app-production
// Search pattern: logs-ecommerce.app-*

Mappings Builder

The source generator creates a typed builder for each registered type. Customize mappings at the property level with full IntelliSense:

public static ProductMappingsBuilder ConfigureMappings(ProductMappingsBuilder m) => m
    .Name(f => f.Analyzer("product_search"))
    .Price(f => f.DocValues(true))
    .AddRuntimeField("discount_pct", r => r.Double()
        .Script("emit((doc['price'].value - doc['sale_price'].value) / doc['price'].value * 100)"))
    .AddDynamicTemplate("labels_as_keyword", dt => dt
        .PathMatch("labels.*")
        .Mapping(m => m.Keyword()));

What Gets Generated

For each registered type, the source generator produces:

• Field constants -- MyContext.Product.Fields.Name (compile-time safe field names)
• Bidirectional field mapping -- PropertyToField / FieldToProperty dictionaries
• Index/search strategy -- write targets, search patterns, data stream names
• Settings + mappings JSON -- pre-computed, ready for index creation
• Content hashes -- detect when mappings change
• Analysis accessors -- type-safe constants for custom analyzers/filters
• Mappings builder -- per-property fluent API for customization

All generated at compile time. Zero reflection at runtime. AOT compatible. Aligned with your System.Text.Json configuration.