feat(oscar): enable table structure editing and view source for Shentong

2026-11-15 14:52:36 +00:00 · 2026-08-07 11:58:52 +08:00
1 parent dd5d74fa4f
commit ae13ca9c1a
@@ -2,8 +2,29 @@ package com.dbx.agent.oscar;

 import com.dbx.agent.ConfiguredJdbcAgent;
 import com.dbx.agent.JdbcAgentProfile;
+import com.dbx.agent.JdbcIdentifiers;
 import com.dbx.agent.MultiSessionJsonRpcServer;
+import com.dbx.agent.ObjectSource;
+import java.sql.PreparedStatement;
+import java.sql.ResultSet;
+import java.util.Locale;

+/**
+ * 神通 OSCAR v7 agent。
+ *
+ * <p>神通是 Oracle 兼容的国产数据库，系统视图（ALL_VIEWS/ALL_SOURCE/ALL_TRIGGERS/ALL_SEQUENCES 等）
+ * 与 Oracle 基本一致，但有两处关键差异（实测 v7.0.8，见 issue #5505）：
+ * <ul>
+ *   <li>{@code DBMS_METADATA.GET_DDL} 不可用（"Namespace DBMS_METADATA does not exist"），
+ *       故视图/过程/函数/触发器源码改用 ALL_VIEWS.TEXT、ALL_SOURCE.TEXT、ALL_TRIGGERS.TRIGGER_BODY 读取。</li>
+ *   <li>{@code SET SCHEMA "x"} 与 {@code ALTER SESSION SET CURRENT_SCHEMA} 均不被 parser 接受，
+ *       schema 切换改用 PG 风格的 {@code SET SEARCH_PATH TO "x"}（已验证可用，且不影响 ALL_* 视图的 OWNER 过滤）。</li>
+ * </ul>
+ *
+ * <p>建表/表结构编辑由 DBX 核心的 Oracle 方言 SQL 生成器负责（神通实测支持 ALTER TABLE
+ * ADD/MODIFY/DROP/RENAME COLUMN、DROP/ADD PRIMARY KEY、COMMENT ON、CREATE/DROP INDEX），
+ * 本 agent 仅需提供元数据（继承 ConfiguredJdbcAgent 的标准 JDBC metadata）与对象源码。
+ */
 public final class OscarAgent extends ConfiguredJdbcAgent {
    public static final JdbcAgentProfile OSCAR_PROFILE = new JdbcAgentProfile(
        "com.oscar.Driver",
@@ -18,4 +39,148 @@ public final class OscarAgent extends ConfiguredJdbcAgent {
    public static void main(String[] args) {
        new MultiSessionJsonRpcServer(OscarAgent::new).run();
    }
+
+    /**
+     * 神通用 {@code SET SEARCH_PATH TO "schema"} 切换默认 schema（PG 风格），
+     * 不支持 Oracle 的 {@code ALTER SESSION SET CURRENT_SCHEMA} 或 Dameng 的 {@code SET SCHEMA "x"}。
+     * 切换后 ALL_* 视图仍按 OWNER 正确过滤，不影响 owner 语义。
+     */
+    @Override
+    public String setSchemaSQL(String schema) {
+        if (schema == null || schema.trim().isEmpty()) {
+            return "";
+        }
+        return "SET SEARCH_PATH TO " + JdbcIdentifiers.INSTANCE.doubleQuote(schema);
+    }
+
+    /**
+     * 读取视图/过程/函数/触发器/序列等对象的 DDL 源码。
+     *
+     * <p>神通不支持 DBMS_METADATA.GET_DDL，按对象类型走对应系统视图：
+     * <ul>
+     *   <li>VIEW → ALL_VIEWS.TEXT（返回完整 SELECT 语句）</li>
+     *   <li>PROCEDURE/FUNCTION/PACKAGE/PACKAGE_BODY/TYPE/TYPE_BODY → ALL_SOURCE.TEXT 按 LINE 排序拼接</li>
+     *   <li>TRIGGER → ALL_TRIGGERS.TRIGGER_BODY</li>
+     *   <li>SEQUENCE → 由 ALL_SEQUENCES 元数据重建 CREATE SEQUENCE 语句</li>
+     *   <li>MATERIALIZED_VIEW → 神通无 ALL_MVIEWS，返回空源码且不可编辑</li>
+     * </ul>
+     */
+    @Override
+    public ObjectSource getObjectSource(String schema, String name, String objectType) {
+        String type = objectType == null ? "" : objectType.trim().toUpperCase(Locale.ROOT);
+        return unchecked(() -> {
+            String source = switch (type) {
+                case "VIEW" -> readViewSource(schema, name);
+                case "PROCEDURE", "FUNCTION", "PACKAGE", "PACKAGE_BODY", "TYPE", "TYPE_BODY" -> readSourceText(schema, name, type);
+                case "TRIGGER" -> readTriggerSource(schema, name);
+                case "SEQUENCE" -> readSequenceSource(schema, name);
+                // 神通 v7 无物化视图系统视图（ALL_MVIEWS 不存在）；返回空源码，标记不可编辑，避免 UI 误判为可改。
+                case "MATERIALIZED_VIEW" -> "";
+                default -> throw new IllegalArgumentException("Unsupported object type: " + objectType);
+            };
+            boolean editable = !"MATERIALIZED_VIEW".equals(type);
+            return new ObjectSource(name, objectType, schema, source, editable);
+        });
+    }
+
+    private String readViewSource(String schema, String name) throws Exception {
+        // ALL_VIEWS.TEXT 已包含完整视图定义（神通实测返回带 schema 限定的 SELECT 语句）。
+        String sql = "SELECT TEXT FROM ALL_VIEWS WHERE OWNER = ? AND VIEW_NAME = ?";
+        return scalarText(sql, schema, name);
+    }
+
+    private String readSourceText(String schema, String name, String type) throws Exception {
+        // ALL_SOURCE 按 LINE 存储过程/函数/包/类型的源码行，需按 LINE 排序拼接。
+        // 神通与 Oracle 的关键差异（实测 v7.0.8）：function 也以 TYPE='PROCEDURE' 存储（不区分 FUNCTION），
+        // 包体随包一起存为 TYPE='PACKAGE'，类型体随类型存为 TYPE='TYPE'。故按 objectType 归并到实际 TYPE 值。
+        String sourceType = switch (type) {
+            case "PROCEDURE", "FUNCTION" -> "PROCEDURE";
+            case "PACKAGE", "PACKAGE_BODY" -> "PACKAGE";
+            case "TYPE", "TYPE_BODY" -> "TYPE";
+            default -> type;
+        };
+        String sql = "SELECT TEXT FROM ALL_SOURCE WHERE OWNER = ? AND NAME = ? AND TYPE = ? ORDER BY LINE";
+        StringBuilder sb = new StringBuilder();
+        try (PreparedStatement stmt = requireConnected().prepareStatement(sql)) {
+            stmt.setString(1, schema);
+            stmt.setString(2, name);
+            stmt.setString(3, sourceType);
+            try (ResultSet rs = stmt.executeQuery()) {
+                while (rs.next()) {
+                    String line = rs.getString(1);
+                    if (line != null) {
+                        sb.append(line);
+                    }
+                }
+            }
+        }
+        return sb.toString();
+    }
+
+    private String readTriggerSource(String schema, String name) throws Exception {
+        String sql = "SELECT TRIGGER_BODY FROM ALL_TRIGGERS WHERE OWNER = ? AND TRIGGER_NAME = ?";
+        return scalarText(sql, schema, name);
+    }
+
+    private String readSequenceSource(String schema, String name) throws Exception {
+        // 神通无 DBMS_METADATA，序列源码由 ALL_SEQUENCES 元数据重建为 CREATE SEQUENCE 语句。
+        String sql = "SELECT MIN_VALUE, MAX_VALUE, INCREMENT_BY, CYCLE_FLAG, ORDER_FLAG, CACHE_SIZE "
+            + "FROM ALL_SEQUENCES WHERE SEQUENCE_OWNER = ? AND SEQUENCE_NAME = ?";
+        try (PreparedStatement stmt = requireConnected().prepareStatement(sql)) {
+            stmt.setString(1, schema);
+            stmt.setString(2, name);
+            try (ResultSet rs = stmt.executeQuery()) {
+                if (!rs.next()) {
+                    return "";
+                }
+                String ref = JdbcIdentifiers.INSTANCE.doubleQuote(schema) + "." + JdbcIdentifiers.INSTANCE.doubleQuote(name);
+                StringBuilder sb = new StringBuilder("CREATE SEQUENCE ").append(ref);
+                String minValue = rs.getString("MIN_VALUE");
+                String maxValue = rs.getString("MAX_VALUE");
+                String increment = rs.getString("INCREMENT_BY");
+                if (increment != null && !"1".equals(increment.trim())) {
+                    sb.append(" INCREMENT BY ").append(increment.trim());
+                }
+                if (minValue != null && !"1".equals(minValue.trim())) {
+                    sb.append(" MINVALUE ").append(minValue.trim());
+                }
+                if (maxValue != null) {
+                    sb.append(" MAXVALUE ").append(maxValue.trim());
+                }
+                String cache = rs.getString("CACHE_SIZE");
+                if (cache != null && !"0".equals(cache.trim())) {
+                    sb.append(" CACHE ").append(cache.trim());
+                } else {
+                    sb.append(" NOCACHE");
+                }
+                if ("Y".equalsIgnoreCase(rs.getString("CYCLE_FLAG"))) {
+                    sb.append(" CYCLE");
+                } else {
+                    sb.append(" NOCYCLE");
+                }
+                if ("Y".equalsIgnoreCase(rs.getString("ORDER_FLAG"))) {
+                    sb.append(" ORDER");
+                } else {
+                    sb.append(" NOORDER");
+                }
+                sb.append(";");
+                return sb.toString();
+            }
+        }
+    }
+
+    /** 读取单行单列文本值（视图/触发器源码），空结果返回空串。 */
+    private String scalarText(String sql, String schema, String name) throws Exception {
+        try (PreparedStatement stmt = requireConnected().prepareStatement(sql)) {
+            stmt.setString(1, schema);
+            stmt.setString(2, name);
+            try (ResultSet rs = stmt.executeQuery()) {
+                if (!rs.next()) {
+                    return "";
+                }
+                String value = rs.getString(1);
+                return value == null ? "" : value;
+            }
+        }
+    }
 }