Github user paul-rogers commented on a diff in the pull request:
https://github.com/apache/drill/pull/666#discussion_r91013251
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/rpc/user/UserSession.java ---
@@ -207,25 +239,165 @@ public SchemaPlus getDefaultSchema(SchemaPlus
rootSchema) {
return null;
}
- final SchemaPlus defaultSchema = SchemaUtilites.findSchema(rootSchema,
defaultSchemaPath);
-
- if (defaultSchema == null) {
- // If the current schema resolves to null, return root schema as the
current default schema.
- return defaultSchema;
- }
-
- return defaultSchema;
+ return SchemaUtilites.findSchema(rootSchema, defaultSchemaPath);
}
public boolean setSessionOption(String name, String value) {
return true;
}
+ /**
+ * @return unique session identifier
+ */
+ public String getUuid() { return uuid; }
+
+ /**
+ * Adds temporary table to temporary tables cache.
+ *
+ * @param schema table schema
+ * @param tableName original table name
+ * @return generated temporary table name
+ */
+ public String registerTemporaryTable(AbstractSchema schema, String
tableName) {
+ return temporaryTablesCache.add(schema, tableName);
+ }
+
+ /**
+ * Looks for temporary table in temporary tables cache by its name in
specified schema.
+ *
+ * @param fullSchemaName table full schema name (example, dfs.tmp)
+ * @param tableName original table name
+ * @return temporary table name if found, null otherwise
+ */
+ public String findTemporaryTable(String fullSchemaName, String
tableName) {
+ return temporaryTablesCache.find(fullSchemaName, tableName);
+ }
+
+ /**
+ * Before removing temporary table from temporary tables cache,
+ * checks if table exists physically on disk, if yes, removes it.
+ *
+ * @param fullSchemaName full table schema name (example, dfs.tmp)
+ * @param tableName original table name
+ * @return true if table was physically removed, false otherwise
+ */
+ public boolean removeTemporaryTable(String fullSchemaName, String
tableName) {
+ final AtomicBoolean result = new AtomicBoolean();
+ temporaryTablesCache.remove(fullSchemaName, tableName, new
BiConsumer<AbstractSchema, String>() {
+ @Override
+ public void accept(AbstractSchema schema, String temporaryTableName)
{
+ if (schema.getTable(temporaryTableName) != null) {
+ schema.dropTable(temporaryTableName);
+ result.set(true);
+ }
+ }
+ });
+ return result.get();
+ }
+
private String getProp(String key) {
return properties.get(key) != null ? properties.get(key) : "";
}
private void setProp(String key, String value) {
properties.put(key, value);
}
+
+ /**
+ * Temporary tables cache stores data by full schema name (schema and
workspace separated by dot
+ * (example: dfs.tmp)) as key, and map of generated temporary tables
names
+ * and its schemas represented by {@link AbstractSchema} as values.
+ * Schemas represented by {@link AbstractSchema} are used to drop
temporary tables.
+ * Generated temporary tables consists of original table name and unique
session id.
+ * Cache is represented by {@link ConcurrentMap} so if is thread-safe
and can be used
+ * in multi-threaded environment.
+ *
+ * Temporary tables cache is used to find temporary table by its name
and schema,
+ * to drop all existing temporary tables on session close
+ * or remove temporary table from cache on user demand.
+ */
+ public static class TemporaryTablesCache {
+
+ private final String uuid;
+ private final ConcurrentMap<String, ConcurrentMap<String,
AbstractSchema>> temporaryTables;
--- End diff --
Please explain the keys here. The first key is... schema name? The second
key the internal name of that table? So, this is a map from temp table name to
a map of (on disk name->schema)? (I may still not get this right, have revised
this note several time as I try to infer the key meanings...)
If this is a map of schema names to (internal name, schema) pairs, doesn't
this mean we duplicate the schema for ever internal name? No harm, but seems a
bit awkward...
Also, does this handle the case where the same temp table is created twice
in the same schema (seems to.) How about this: a nefarious (or clueless) user
has schema A and B both refer to the same disk directory. How will we tell that
tabname_uuid in B is already taken by a temp table of the same name in A?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
