Github user karanmehta93 commented on a diff in the pull request:
https://github.com/apache/phoenix/pull/309#discussion_r219310231
--- Diff: phoenix-core/src/test/java/org/apache/phoenix/query/BaseTest.java
---
@@ -1894,10 +1877,36 @@ protected static void
splitSystemCatalog(Map<String, List<String>> tenantToTable
}
}
}
+
+ }
+
+ /**
+ * Splits SYSTEM.CATALOG into multiple regions based on the table or
view names passed in.
+ * Metadata for each table or view is moved to a separate region,
+ * @param tenantToTableAndViewMap map from tenant to tables and views
owned by the tenant
+ */
+ protected static void splitSystemCatalog(Map<String, List<String>>
tenantToTableAndViewMap) throws Exception {
+ List<byte[]> splitPoints = Lists.newArrayListWithExpectedSize(5);
+ // add the rows keys of the table or view metadata rows
+ Set<String> schemaNameSet=Sets.newHashSetWithExpectedSize(15);
+ for (Entry<String, List<String>> entrySet :
tenantToTableAndViewMap.entrySet()) {
+ String tenantId = entrySet.getKey();
+ for (String fullName : entrySet.getValue()) {
+ String schemaName =
SchemaUtil.getSchemaNameFromFullName(fullName);
+ // we don't allow SYSTEM.CATALOG to split within a schema,
so to ensure each table
+ // or view is on a separate region they need to have a
unique tenant and schema name
+ assertTrue("Schema names of tables/view must be unique ",
schemaNameSet.add(tenantId+"."+schemaName));
+ String tableName =
SchemaUtil.getTableNameFromFullName(fullName);
+ splitPoints.add(
+ SchemaUtil.getTableKey(tenantId, "".equals(schemaName)
? null : schemaName, tableName));
+ }
+ }
+ Collections.sort(splitPoints, Bytes.BYTES_COMPARATOR);
+
splitTable(PhoenixDatabaseMetaData.SYSTEM_CATALOG_HBASE_TABLE_NAME,
splitPoints);
--- End diff --
Does calling `split()` method explicitly check whether the table is allowed
to split on that key or not by consulting the `MetaDataSplitPolicy `
---
