Add a generic, extensible rest-endpoint provider SPI

common/src/main/java/org/opensearch/sql/common/setting/Settings.java

@@ -36,6 +36,7 @@ public enum Key {
     PPL_SYNTAX_LEGACY_PREFERRED("plugins.ppl.syntax.legacy.preferred"),
     PPL_SUBSEARCH_MAXOUT("plugins.ppl.subsearch.maxout"),
     PPL_JOIN_SUBSEARCH_MAXOUT("plugins.ppl.join.subsearch_maxout"),
+    PPL_REST_ALLOWED_ENDPOINTS("plugins.ppl.rest.allowed_endpoints"),
 
     /** Enable Calcite as execution engine */
     CALCITE_ENGINE_ENABLED("plugins.calcite.enabled"),

core/src/main/java/org/opensearch/sql/ast/tree/RestRelation.java

@@ -0,0 +1,25 @@
+/*
+ * Copyright OpenSearch Contributors
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+package org.opensearch.sql.ast.tree;
+
+import java.util.Collections;
+import lombok.EqualsAndHashCode;
+import lombok.ToString;
+import org.opensearch.sql.ast.expression.UnresolvedExpression;
+
+/**
+ * Extend Relation to mark a {@code rest} leading command. The single table name is a reserved,
+ * encoded token (produced by {@link org.opensearch.sql.utils.SystemIndexUtils#restTable}) that
+ * carries the validated REST endpoint spec; it resolves through the storage engine to a REST source
+ * table on the Calcite path, exactly as {@link DescribeRelation} resolves to a system index.
+ */
+@ToString
+@EqualsAndHashCode(callSuper = false)
+public class RestRelation extends Relation {
+  public RestRelation(UnresolvedExpression tableName) {
+    super(Collections.singletonList(tableName));
+  }
+}

core/src/main/java/org/opensearch/sql/utils/SystemIndexUtils.java

@@ -5,6 +5,10 @@
 
 package org.opensearch.sql.utils;
 
+import java.nio.charset.StandardCharsets;
+import java.util.HexFormat;
+import java.util.LinkedHashMap;
+import java.util.Map;
 import lombok.Getter;
 import lombok.RequiredArgsConstructor;
 import lombok.experimental.UtilityClass;
@@ -34,6 +38,110 @@ public static Boolean isSystemIndex(String indexName) {
     return indexName.endsWith(SYS_TABLES_SUFFIX);
   }
 
+  /**
+   * Reserved suffix marking a {@code rest} source table. Distinct from {@link #SYS_TABLES_SUFFIX}
+   * so {@link #isSystemIndex} and {@link #isRestSource} never overlap. The whole reserved name is a
+   * single Calcite identifier (REST + lowercase hex + this suffix), so it survives name resolution
+   * the same way the uppercase system-mapping names do.
+   */
+  private static final String REST_SOURCE_SUFFIX = "__REST_SOURCE";
+
+  private static final String REST_SOURCE_PREFIX = "REST";
+
+  /** True if the resolved table name is a {@code rest} source token. */
+  public static boolean isRestSource(String indexName) {
+    return indexName.endsWith(REST_SOURCE_SUFFIX);
+  }
+
+  /**
+   * Encode a validated {@link RestSpec} into a single reserved table name. Mirrors {@link
+   * #mappingTable}: structured metadata travels inside a reserved name rather than a side channel.
+   * The endpoint/args have already been allow-list-validated before this is called.
+   */
+  public static String restTable(RestSpec spec) {
+    StringBuilder sb = new StringBuilder();
+    sb.append("endpoint=").append(spec.getEndpoint());
+    if (spec.getCount() != null) {
+      sb.append('\n').append("count=").append(spec.getCount());
+    }
+    if (spec.getTimeout() != null) {
+      sb.append('\n').append("timeout=").append(spec.getTimeout());
+    }
+    if (spec.getArgs() != null) {
+      for (Map.Entry<String, String> e : spec.getArgs().entrySet()) {
+        sb.append('\n').append("arg.").append(e.getKey()).append('=').append(e.getValue());
+      }
+    }
+    return REST_SOURCE_PREFIX + toHex(sb.toString()) + REST_SOURCE_SUFFIX;
+  }
+
+  /** Decode a reserved {@code rest} table name back into its {@link RestSpec}. */
+  public static RestSpec decodeRestSpec(String indexName) {
+    // Validate the token shape before slicing it: callers gate on isRestSource today, but a
+    // public decoder must not assume its precondition, otherwise a malformed token would throw an
+    // opaque StringIndexOutOfBoundsException from substring rather than a clear input error.
+    if (!isRestSource(indexName)) {
+      throw new IllegalArgumentException("not a valid rest source token: " + indexName);
+    }
+    String body =
+        indexName.substring(
+            REST_SOURCE_PREFIX.length(), indexName.length() - REST_SOURCE_SUFFIX.length());
+    String decoded = fromHex(body);
+    String endpoint = null;
+    Integer count = null;
+    String timeout = null;
+    LinkedHashMap<String, String> args = new LinkedHashMap<>();
+    for (String line : decoded.split("\n")) {
+      if (line.isEmpty()) {
+        continue;
+      }
+      int eq = line.indexOf('=');
+      if (eq < 0) {
+        continue;
+      }
+      String k = line.substring(0, eq);
+      String v = line.substring(eq + 1);
+      if (k.equals("endpoint")) {
+        endpoint = v;
+      } else if (k.equals("count")) {
+        count = Integer.parseInt(v);
+      } else if (k.equals("timeout")) {
+        timeout = v;
+      } else if (k.startsWith("arg.")) {
+        args.put(k.substring("arg.".length()), v);
+      }
+    }
+    if (endpoint == null) {
+      throw new IllegalArgumentException("rest source token is missing the endpoint: " + indexName);
+    }
+    return new RestSpec(endpoint, args, count, timeout);
+  }
+
+  private static String toHex(String s) {
+    return HexFormat.of().formatHex(s.getBytes(StandardCharsets.UTF_8));
+  }
+
+  private static String fromHex(String h) {
+    if (h.length() % 2 != 0) {
+      throw new IllegalArgumentException("not a valid rest source token: odd-length hex body");
+    }
+    return new String(HexFormat.of().parseHex(h), StandardCharsets.UTF_8);
+  }
+
+  /**
+   * The validated spec for a {@code rest} command: an allow-listed read-only endpoint plus optional
+   * count/timeout/query args. Lives in core so the parser (encode) and the storage engine (decode)
+   * share it without a cross-module dependency.
+   */
+  @Getter
+  @RequiredArgsConstructor
+  public static class RestSpec {
+    private final String endpoint;
+    private final Map<String, String> args;
+    private final Integer count;
+    private final String timeout;
+  }
+
   /**
    * Compose system mapping table.
    *

docs/category.json

@@ -35,6 +35,7 @@
     "user/ppl/cmd/rename.md",
     "user/ppl/cmd/multisearch.md",
     "user/ppl/cmd/replace.md",
+    "user/ppl/cmd/rest.md",
     "user/ppl/cmd/rex.md",
     "user/ppl/cmd/search.md",
     "user/ppl/cmd/showdatasources.md",

docs/user/ppl/cmd/rest.md

@@ -0,0 +1,63 @@
+# rest
+
+The `rest` command is a leading command that reads an allow-listed, read-only in-cluster management endpoint and emits the response as PPL rows. Its rows come from the endpoint dispatch, not from an index, so `rest` appears at the start of a query.
+
+> **Note**: The `rest` command is supported only on the Calcite query engine (`plugins.calcite.enabled=true`). Each endpoint has a fixed output schema, and the dispatch runs under the caller's security context, so a user who cannot call an endpoint directly cannot call it through `rest`. The command is read-only; mutating and non-allow-listed endpoints are rejected. Each endpoint requires the same cluster-monitor privilege as calling it natively, so `rest` grants no extra access.
+
+The `rest` command is a generic, extensible framework: a plugin contributes additional read-only endpoints through the `RestEndpointProvider` extension point without changing the grammar. This first version ships a single built-in endpoint, `/_cluster/health`. Additional endpoints (for example `/_cat/nodes`, `/_cat/shards`, `/_cluster/state`, `/_cluster/settings`) can be added in follow-ups, with any response redaction handled inside the provider's own handler.
+
+## Enabling the command
+
+`/_cluster/health` is **enabled by default**: `plugins.ppl.rest.allowed_endpoints` defaults to `["/_cluster/health"]`. Any other endpoint is rejected until a deployment adds it to the allow-list (a node-level setting, applied at node startup and not changeable at runtime):
+
+```yaml
+plugins.ppl.rest.allowed_endpoints: ["/_cluster/health"]
+```
+
+Every endpoint must be listed explicitly by name; there is no wildcard, so a newly installed or upgraded provider is never enabled without an explicit allow-list change. Set an empty list to disable the command entirely.
+
+## Syntax
+
+```syntax
+rest <endpoint-path> [count=<int>] [<get-arg>=<value> ...]
+```
+
+## Parameters
+
+| Parameter | Required/Optional | Description |
+| --- | --- | --- |
+| `<endpoint-path>` | Required | An allow-listed, read-only endpoint path (see the allow-list below), for example `/_cluster/health`. |
+| `count=<int>` | Optional | Caps the number of emitted rows. |
+| `<get-arg>=<value>` | Optional | Endpoint query arguments, validated per endpoint by both key and value (for example `local=true` for `/_cluster/health`). |
+
+## Allow-list
+
+`rest` resolves only an explicit, curated set of read-only endpoints. Anything outside the list, including any mutating endpoint, is rejected with a clear error.
+
+| Endpoint | Output columns | Accepted args |
+| --- | --- | --- |
+| `/_cluster/health` | `response` (string): the full cluster-health response as JSON. Extract fields with `json_extract` or the `spath` command (see the example below). | `local` |
+
+## Example: Reading fields from the response
+
+`/_cluster/health` returns the full health response in a single `response` column as JSON. Extract the fields you need with `json_extract` (or the `spath` command):
+
+```ppl ignore
+| rest '/_cluster/health'
+| eval status = json_extract(response, 'status'),
+       number_of_nodes = json_extract(response, 'number_of_nodes')
+| fields status, number_of_nodes
+```
+
+The query returns the following results:
+
+```text
+fetched rows / total rows = 1/1
++--------+-----------------+
+| status | number_of_nodes |
+|--------+-----------------|
+| green  | 1               |
++--------+-----------------+
+```
+
+Because the whole response is available, a query can read any field it exposes (for example `active_shards`, `active_primary_shards`, `unassigned_shards`) without the endpoint pre-declaring a column for it. The extracted columns then compose with downstream `where`, `sort`, `stats`, and `fields` exactly like an index scan, for example `| rest '/_cluster/health' | spath input=response path=status output=status | where status = 'green'`.

docs/user/ppl/index.md

@@ -75,6 +75,7 @@ source=accounts
 | [lookup command](cmd/lookup.md) | 3.0 | experimental (since 3.0) | Add or replace data from a lookup index. |
 | [multisearch command](cmd/multisearch.md) | 3.4 | experimental (since 3.4) | Execute multiple search queries and combine their results. |
 | [union command](cmd/union.md) | 3.7 | experimental (since 3.7) | Combine results from multiple datasets using UNION ALL semantics. |
+| [rest command](cmd/rest.md) | 3.9 | experimental (since 3.9) | Read an allow-listed, read-only in-cluster management endpoint (cluster/cat/nodes) as rows. Calcite engine only. |
 | [ml command](cmd/ml.md) | 2.5 | stable (since 2.5) | Apply machine learning algorithms to analyze data. |
 | [kmeans command](cmd/kmeans.md) | 1.3 | stable (since 1.3) | Apply the kmeans algorithm on the search result returned by a PPL command. |
 | [ad command](cmd/ad.md) | 1.3 | deprecated (since 2.5) | Apply Random Cut Forest algorithm on the search result returned by a PPL command. |

doctest/build.gradle

@@ -205,6 +205,8 @@ testClusters {
         plugin(getJobSchedulerPlugin())
         plugin ':opensearch-sql-plugin'
         testDistribution = 'archive'
+        // Only /_cluster/health is registered; pin the allow-list to it so the rest.md examples run.
+        setting 'plugins.ppl.rest.allowed_endpoints', '/_cluster/health'
     }
 }
 tasks.register("runRestTestCluster", RunTask) {

integ-test/build.gradle

@@ -387,6 +387,8 @@ testClusters {
         plugin(getGeoSpatialPlugin())
         plugin ":opensearch-sql-plugin"
         setting "plugins.query.datasources.encryption.masterkey", "1234567812345678"
+        // Only /_cluster/health is registered; pin the allow-list to it for the rest ITs.
+        setting 'plugins.ppl.rest.allowed_endpoints', '/_cluster/health'
     }
     yamlRestTest {
         testDistribution = 'archive'
@@ -405,6 +407,8 @@ testClusters {
         testDistribution = 'archive'
         plugin(getJobSchedulerPlugin())
         plugin ":opensearch-sql-plugin"
+        // Only /_cluster/health is registered; pin the allow-list to it for RestCommandSecurityIT.
+        setting 'plugins.ppl.rest.allowed_endpoints', '/_cluster/health'
     }
     remoteIntegTestWithSecurity {
         testDistribution = 'archive'
@@ -419,6 +423,8 @@ testClusters {
         plugin(getArrowFlightRpcPlugin())
         plugin(getAnalyticsEnginePlugin())
         plugin ":opensearch-sql-plugin"
+        // Composite-default cluster: PPL queries route to the analytics engine unless excluded.
+        setting 'cluster.pluggable.dataformat', 'composite'
     }
 }
 

integ-test/src/test/java/org/opensearch/sql/calcite/CalciteNoPushdownIT.java

@@ -59,6 +59,7 @@
   CalciteObjectFieldOperateIT.class,
   CalciteOperatorIT.class,
   CalciteParseCommandIT.class,
+  CalcitePPLRestIT.class,
   CalcitePPLAggregationIT.class,
   CalcitePPLAppendcolIT.class,
   CalcitePPLAppendCommandIT.class,

integ-test/src/test/java/org/opensearch/sql/calcite/remote/CalciteExplainIT.java

@@ -66,6 +66,15 @@ public void init() throws Exception {
     loadIndex(Index.GRAPH_EMPLOYEES);
   }
 
+  // Only for Calcite: the rest row source explains as a CalciteScannableCatalogScan.
+  @Test
+  public void explainRestCommand() throws IOException {
+    String result = explainQueryToString("| rest '/_cluster/health' | fields response");
+    Assert.assertTrue(
+        "Expected a rest scan node in the explain output, got: " + result,
+        result.contains("CatalogScan"));
+  }
+
   @Override
   @Ignore("test only in v2")
   public void testExplainModeUnsupportedInV2() throws IOException {}