使用.NET SDK创建Azure AI Search索引时，如何将float转为double？

问题背景 #

使用.NET SDK创建Azure AI Search Service索引时，模型中的float?字段（示例如下）会触发不支持Edm.Single类型的错误：

[SimpleField(IsFilterable = true, IsSortable = true, IsFacetable = true)]
public float? MonitoringFee { get; set; }

错误信息：

{"error":{"code":"InvalidRequestParameter","message":"The request is invalid. Details: definition : The field 'MonitoringFee' cannot specify a field type 'Edm.Single' because it is not supported.","details":[{"code":"InvalidField","message":"The field 'MonitoringFee' cannot specify a field type 'Edm.Single' because it is not supported. Parameters: definition"}]}}

已尝试通过自定义JsonSerializer和[JsonConverter(typeof(FloatToDoubleConverter))]注解解决，但未生效，且无法修改模型的数据类型。

解决方案 #

要解决这个问题，需要分两步处理：确保索引字段类型为Edm.Double，以及确保数据导入时float值被序列化为double。

1. 修改索引字段类型为Edm.Double

默认FieldBuilder会将float/float?映射为Edm.Single，需要手动替换为Edm.Double，有两种方式：

方式一：自定义FieldBuilder

继承FieldBuilder并重写字段类型映射逻辑：

public class CustomFloatToDoubleFieldBuilder : FieldBuilder
{
    protected override Field CreateField(MemberInfo memberInfo, IEnumerable<Attribute> attributes)
    {
        var field = base.CreateField(memberInfo, attributes);
        // 将Edm.Single替换为Edm.Double
        if (field.Type == EdmType.Single)
        {
            field.Type = EdmType.Double;
        }
        return field;
    }
}

创建索引时使用自定义FieldBuilder：

var fieldBuilder = new CustomFloatToDoubleFieldBuilder();
var indexFields = fieldBuilder.Build(typeof(YourModel));
var searchIndex = new SearchIndex("your-index-name")
{
    Fields = indexFields
};

await searchIndexClient.CreateOrUpdateIndexAsync(searchIndex);

方式二：手动修改生成的字段列表

如果不想自定义FieldBuilder，可在生成字段后遍历修改类型：

var fieldBuilder = new FieldBuilder();
var indexFields = fieldBuilder.Build(typeof(YourModel)).ToList();

foreach (var field in indexFields)
{
    if (field.Type == EdmType.Single)
    {
        field.Type = EdmType.Double;
    }
}

var searchIndex = new SearchIndex("your-index-name") { Fields = indexFields };
await searchIndexClient.CreateOrUpdateIndexAsync(searchIndex);

2. 确保数据导入时float值序列化为double

索引字段类型改为Edm.Double后，需要确保导入的float值被正确转换为double类型，可通过以下方式实现：

方式一：正确配置JsonConverter

重新实现针对float?的转换器（处理可空类型场景）：

public class NullableFloatToDoubleConverter : JsonConverter<float?>
{
    public override float? Read(ref Utf8JsonReader reader, Type typeToConvert, JsonSerializerOptions options)
    {
        return reader.TokenType == JsonTokenType.Null ? null : (float?)reader.GetDouble();
    }

    public override void Write(Utf8JsonWriter writer, float? value, JsonSerializerOptions options)
    {
        if (value.HasValue)
        {
            writer.WriteNumberValue((double)value.Value);
        }
        else
        {
            writer.WriteNullValue();
        }
    }
}

将转换器配置到SearchClientOptions中：

var jsonOptions = new JsonSerializerOptions();
jsonOptions.Converters.Add(new NullableFloatToDoubleConverter());

var clientOptions = new SearchClientOptions
{
    Serializer = new JsonObjectSerializer(jsonOptions)
};

var searchIndexClient = new SearchIndexClient(
    new Uri("your-search-service-endpoint"),
    new AzureKeyCredential("your-api-key"),
    clientOptions);

方式二：导入前转换对象

如果序列化配置仍有问题，可在导入前将模型转换为匿名类型，手动将float字段转为double：

var documents = yourModelList.Select(model => new
{
    Id = model.Id,
    MonitoringFee = (double?)model.MonitoringFee,
    // 复制其他字段...
});

await searchIndexClient.IndexDocumentsAsync(IndexDocumentsBatch.Upload(documents));

内容的提问来源于stack exchange，提问作者Mujahid Ali