akashrn5 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document URL: https://github.com/apache/carbondata/pull/3689#discussion_r401417882
########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,155 @@ +<!-- + Licensed to the Apache Software Foundation (ASF) under one or more + contributor license agreements. See the NOTICE file distributed with + this work for additional information regarding copyright ownership. + The ASF licenses this file to you under the Apache License, Version 2.0 + (the "License"); you may not use this file except in compliance with + the License. You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. +--> + +# CarbonData Secondary Index + +* [Quick Example](#quick-example) +* [Secondary Index Table](#Secondary-Index-Introduction) +* [Loading Data](#loading-data) +* [Querying Data](#querying-data) +* [Compaction](#compacting-SI-table) +* [Data Management](#data-management-with-SI-tables) + +## Quick example + +Start spark-sql in terminal and run the following queries, +``` +CREATE TABLE maintable(a int, b string, c string) stored as carbondata; +insert into maintable select 1, 'ab', 'cd'; +CREATE index inex1 on table maintable(c) AS 'carbondata'; +SELECT a from maintable where c = 'cd'; +// NOTE: run explain query and check if query hits the SI table from the plan +EXPLAIN SELECT a from maintable where c = 'cd'; +``` + +## Secondary Index Introduction + Sencondary index tables are created as a child table and manager as tables internally by Carbondata. + User can create the SI tables based on the column position in table(Recommended for right columns) + and the queries should have filter on that column to improve the filter query performance. + + SI tables will always be loaded non-lazy way. Once SI table is created, Carbondata's + CarbonOptimizer with the help of CarbonSITransformationRule, transforms the query plan to hit the + SI table based on the filter condition or set of filter conditions present in the query. + So first level of pruning will be done on SI table as it stores blocklets and main table/parent + table pruning will be based on the SI output, which helps in giving the faster query results with + better pruning. + + For instance, main table called **sales** which is defined as + + ``` + CREATE TABLE sales ( + order_time timestamp, + user_id string, + sex string, + country string, + quantity int, + price bigint) + STORED AS carbondata + ``` + + User can create SI table using the Create Index DDL + + ``` + CREATE INDEX index_sales + ON TABLE sales(user_id) + AS + 'carbondata' + TBLPROPERTIES('table_blocksize'='1') + ``` +**NOTE**: + * Secondary Index tables are registered with hive and stored n HiveSERDEPROPERTIES s json formatted + value. Maximum characters supported for SERDEPROPERTIES by hive is 4000 characters which cannot be + changed. + + +#### How SI tables are selected + +When a user filter query is submitted, during query planning phase, CarbonData with help of +CarbonSITransformationRule, checkes if there are any index tables present on the filter column of +query. If there are any, then filter query plan will be transformed such a way that, execution will +first hit the corresponding SI table and give input to main table for further pruning. + + +For the main table **sales** and SI table **index_sales** created above, following queries +``` +SELECT country, sex from sales where user_id = 'xxx' + +SELECT country, sex from sales where user_id = 'xxx' and country = 'INDIA' +``` + +will be transformed by CarbonData's CarbonSITransformationRule to query against SI table +**index_sales** first which will be input to the main table **sales** + + +## Loading data + +### Loading data to Secondary Index table(s). + +*case1:* When SI table is created and the main table does not have any data. In this case every +consecutive load will load to SI table once main table data load is finished. + +*case2:* When SI table is created and main table already contains some data, then SI creation will +also load to SI table with same number of segments as main table. There after, consecutive load to +main table will load to SI table also. + + **NOTE**: + * In case of any failure to SI table, then we make the SI table disable by setting a hive serde + property and subsequent load to main table will also load to SI with failed load data also and + make it enable and available for query. + +## Querying data +Direct query can be made on SI tables to see the data present in position reference columns. +When a filter query is fired, if the filter column is a secondary index column, then plan is +transformed accordingly to hit SI table first to make better pruning with main table and in turn +helps for faster query results. + +User can verify whether a query can leverage SI table or not by executing `EXPLAIN` +command, which will show the transformed logical plan, and thus user can check whether SI table +table is selected. Review comment: i think, we have many things like that like, filter combinations of or , and so, better to avoid these kind in user documents , it will be better to mention in developer document ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services