let mnemonic = "dimension_metadata";
  let parameters = (ins
      TFL_DimensionTypeAttr:$format,
      "int32_t":$dense_size,
      I32ArrayParameter:$segments,
      I32ArrayParameter:$indices
  );
  let summary = "Dimension metadata.";
  let assemblyFormat = "`<` struct(params) `>`";
}

def SparsityParameterAttr : AttrDef<TFL_Dialect, "SparsityParameter"> {
  let mnemonic = "sparsity_parameter";

        }

        val substitutions = LinkedHashMap<TypeConstructor, TypeProjection>(parameters.size)
        for (index in parameters.indices) {
            val parameterTypeConstructor = parameters[index].typeConstructor
            substitutions[parameterTypeConstructor] = TypeProjectionImpl(arguments[index] as KotlinType)
        }

    public static void afterClass() throws Exception {
        runner.close();
        runner.clean();
    }

    @Before
    public void before() throws Exception {
        try {
            runner.admin().indices().prepareDelete("_all").execute().actionGet();
        } catch (IndexNotFoundException ignore) {

        }
        runner.refresh();
        settings = Suggester.builder().build(runner.client(), id).settings();
    }

    public static void afterClass() throws Exception {
        runner.close();
        runner.clean();
    }

    @Before
    public void before() throws Exception {
        try {
            runner.admin().indices().prepareDelete("_all").execute().actionGet();
        } catch (IndexNotFoundException ignore) {

        }
        runner.refresh();
        settings = Suggester.builder().build(runner.client(), id).settings();
    }

  // CHECK-NEXT: [[ready_ch:%.*]] = tfrt.new.chain
  // CHECK-NEXT: [[ch:%.*]], [[result:%.*]] = tfrt_fallback_async.get_resource [[ready_ch]] {device = "/device:CPU:0", indices = [0]} : (!tfrt.chain) -> (!tfrt.chain, !tfrt_fallback.tf_tensor)
  // CHECK-NEXT: [[ch1:%.*]], [[var:%.*]] = tfrt_fallback_async.executeop.seq([[in_chain]]) {{.*}} "tf.ReadVariableOp"([[arg1]])

        // requestedAttributes.
        //
        // If the requested attributes are [ A, B, C ]
        // If the attribute precedence is [ C, A ]
        // The indices are [ A: 0, B: 1, C: 2 ]
        // The sorted indices are [ 2, 0 ]
        // The unsorted indices are [ 1 ]
        //
        final AttributeSelectionSchema.PrecedenceResult precedenceResult = schema.orderByPrecedence(requested.keySet());

index3loop:
	VLOADSWAP(R7, R0, V2, V2)  // Load with correct order
	VSLDOI   $1, V2, V3, V4    // string[1:17]
	VSLDOI   $2, V2, V3, V9    // string[2:18]
	VCMPEQUH V1, V2, V5        // compare hw even indices
	VCMPEQUH V1, V4, V6        // compare hw odd indices
	VCMPEQUB V8, V9, V10       // compare 3rd to last byte
	VSEL     V6, V5, V31, V7   // Find 1st matching byte using mask
	VAND     V7, V10, V7       // AND matched bytes with matched 3rd byte

  %indices0 = "tf.Const"() {value = dense<4> : tensor<i32>} : () -> tensor<i32>
  %indices1 = "tf.Const"() {value = dense<[[3, 2], [1, 0]]> : tensor<2x2xi32>} : () -> tensor<2x2xi32>
  %0 = "tf.DynamicStitch"(%indices0, %indices1, %arg0, %arg1) : (tensor<i32>, tensor<2x2xi32>, tensor<2xf32>, tensor<2x2x2xf32>) -> tensor<5x2xf32>

 * implementations, except when the table is sparse.
 *
 * <p>Null row keys or column keys are not permitted.
 *
 * <p>This class provides methods involving the underlying array structure, where the array indices
 * correspond to the position of a row or column in the lists of allowed keys and values. See the
 * {@link #at}, {@link #set}, {@link #toArray}, {@link #rowKeyList}, and {@link #columnKeyList}
 * methods for more details.
 *

import org.opensearch.index.analysis.CharFilterFactory;
import org.opensearch.index.analysis.TokenFilterFactory;
import org.opensearch.index.analysis.TokenizerFactory;
import org.opensearch.indices.analysis.AnalysisModule.AnalysisProvider;
import org.opensearch.plugins.AnalysisPlugin;
import org.opensearch.plugins.Plugin;

public class ExtensionPlugin extends Plugin implements AnalysisPlugin {

    @Override