[GitHub] [carbondata] akashrn5 opened a new pull request #3689: [WIP]add Secondary Index Document
Locked
1234
1234
 
|
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401411591 ########## 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: with filter on SI and donotPushToSI. For expressions like NOT, ISNOTNULL etc. Can list those expressions ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
akashrn5 commented on issue #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#issuecomment-607091132 @Indhumathi27 handled all comments, please check.Thanks for the review. ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401447490 ########## File path: docs/configuration-parameters.md ########## @@ -147,6 +148,8 @@ This section provides the details of all the configurations required for the Car | carbon.query.stage.input.enable | false | Stage input files are data files written by external applications (such as Flink), but have not been loaded into carbon table. Enabling this configuration makes query to include these files, thus makes query on latest data. However, since these files are not indexed, query maybe slower as full scan is required for these files. | | carbon.driver.pruning.multi.thread.enable.files.count | 100000 | To prune in multi-thread when total number of segment files for a query increases beyond the configured value. | | carbon.load.all.segment.indexes.to.cache | true | Setting this configuration to false, will prune and load only matched segment indexes to cache using segment metadata information such as columnid and it's minmax values, which decreases the usage of driver memory. | +| carbon.secondary.index.creation.threads | 1 | Specifies the number of threads to concurrently process segments during secondary index creation. This property helps fine tuning the system when there are a lot of segments in a table. The value range is 1 to 50. | +| carbon.si.lookup.partialstring | true | When true, it includes starts with, ends with and contains. When false, it includes only starts with secondary indexex. | Review comment: ```suggestion | carbon.si.lookup.partialstring | true | When true, it includes starts with, ends with and contains. When false, it includes only starts with secondary indexes. | ``` ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401449799 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as 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 executes a 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. The subsequent main table load will load the old failed loads along with current load and + makes the SI table 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. + + +## Compacting SI table + +### Compacting SI table table through Main Table compaction +Running Compaction command (`ALTER TABLE COMPACT`)[COMPACTION TYPE-> MINOR/MAJOR] on main table will +automatically delete all the old segments of SI and creates a new segment with same name as main +table compacted segmet and loads data to it. + +### Compacting SI table's individual segment(s) through REBUILD command +Where there are so many small files present in the SI table, then we can use REBUILD command to +compact the files within an SI segment to avoid many small files. + + ``` + REBUILD INDEX sales_index + ``` +This command merges data files in each segment of SI table. + + ``` + REBUILD INDEX sales_index WHERE SEGMENT.ID IN(1) + ``` +This command merges data files within specified segment of SI table. + +## How to skip Secondary Index? +When Secondary indexes are created on a table(s), always data fetching happens from secondary +indexes created on the maintables for better performance. But sometimes, data fetching from the +secondary index might degrade query performance. So to avoid such secondary indexes, we use NI as a +function on flters with in WHERE clause. + + ``` + SELECT country, sex from sales where NI(user_id = 'xxx') + ``` +The above query ignores column user_id from secondary index and fetch data from main table. + +## DDLs on Secondary Index + +### Show index Command +This command is used to get information about all the secondary indexes on a table. + +Syntax + ``` + SHOW INDEXES on sales Review comment: ```suggestion SHOW INDEXES on [db_name.]table_name ``` ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401451711 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as 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 executes a 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. The subsequent main table load will load the old failed loads along with current load and + makes the SI table 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. + + +## Compacting SI table + +### Compacting SI table table through Main Table compaction +Running Compaction command (`ALTER TABLE COMPACT`)[COMPACTION TYPE-> MINOR/MAJOR] on main table will +automatically delete all the old segments of SI and creates a new segment with same name as main +table compacted segmet and loads data to it. + +### Compacting SI table's individual segment(s) through REBUILD command +Where there are so many small files present in the SI table, then we can use REBUILD command to +compact the files within an SI segment to avoid many small files. + + ``` + REBUILD INDEX sales_index + ``` +This command merges data files in each segment of SI table. + + ``` + REBUILD INDEX sales_index WHERE SEGMENT.ID IN(1) + ``` +This command merges data files within specified segment of SI table. + +## How to skip Secondary Index? +When Secondary indexes are created on a table(s), always data fetching happens from secondary +indexes created on the maintables for better performance. But sometimes, data fetching from the +secondary index might degrade query performance. So to avoid such secondary indexes, we use NI as a +function on flters with in WHERE clause. + + ``` + SELECT country, sex from sales where NI(user_id = 'xxx') + ``` +The above query ignores column user_id from secondary index and fetch data from main table. + +## DDLs on Secondary Index + +### Show index Command +This command is used to get information about all the secondary indexes on a table. + +Syntax + ``` + SHOW INDEXES on sales + ``` + +### Drop index Command +This command is used to drop an existing secondary index on a table + +Syntax + ``` + DROP INDEX IF EXISTS index_sales on sales + ``` + +### Register index Command +This command registers the secondary index with the main table in case of compatibility scenarios +where we have old stores. + +Syntax + ``` + REGISTER INDEX TABLE index_sales ON sales Review comment: For syntax: ```suggestion REGISTER INDEX TABLE index_name ON [db_name.]table_name ``` For example: can mention REGISTER INDEX TABLE index_sales ON sales ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401450107 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as 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 executes a 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. The subsequent main table load will load the old failed loads along with current load and + makes the SI table 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. + + +## Compacting SI table + +### Compacting SI table table through Main Table compaction +Running Compaction command (`ALTER TABLE COMPACT`)[COMPACTION TYPE-> MINOR/MAJOR] on main table will +automatically delete all the old segments of SI and creates a new segment with same name as main +table compacted segmet and loads data to it. + +### Compacting SI table's individual segment(s) through REBUILD command +Where there are so many small files present in the SI table, then we can use REBUILD command to +compact the files within an SI segment to avoid many small files. + + ``` + REBUILD INDEX sales_index + ``` +This command merges data files in each segment of SI table. + + ``` + REBUILD INDEX sales_index WHERE SEGMENT.ID IN(1) + ``` +This command merges data files within specified segment of SI table. + +## How to skip Secondary Index? +When Secondary indexes are created on a table(s), always data fetching happens from secondary +indexes created on the maintables for better performance. But sometimes, data fetching from the +secondary index might degrade query performance. So to avoid such secondary indexes, we use NI as a +function on flters with in WHERE clause. + + ``` + SELECT country, sex from sales where NI(user_id = 'xxx') + ``` +The above query ignores column user_id from secondary index and fetch data from main table. + +## DDLs on Secondary Index + +### Show index Command +This command is used to get information about all the secondary indexes on a table. + +Syntax + ``` + SHOW INDEXES on sales + ``` + +### Drop index Command +This command is used to drop an existing secondary index on a table + +Syntax + ``` + DROP INDEX IF EXISTS index_sales on sales Review comment: ```suggestion DROP INDEX IF EXISTS index_name on [db_name.]table_name ``` ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401452823 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 + ``` + Review comment: Can add syntax for create index command and can keep example as below code ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401453741 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as json formatted Review comment: ```suggestion * Secondary Index tables are registered with hive and stored in HiveSERDEPROPERTIES as json formatted ``` ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401458116 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as json formatted Review comment: can add limitations for si in Notes section: 1. not supported for measure column and complex column 2. cannot be created on streaming/partition/non transaction table 3. Index table column indexing order cannot be same as Parent table column start order 4. TableProperties like TABLE_BLOCKSIZE,COLUMN_META_CACHE,CACHE_LEVEL,COMPRESSOR are supported in create index ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401458439 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as 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 executes a filter query is submitted, during query planning phase, CarbonData with help of Review comment: ```suggestion When a user executes a filter query, during query planning phase, CarbonData with help of ``` ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401460400 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) Review comment: Add DDLs on Secondary Index ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
Indhumathi27 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401458549 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as 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 executes a 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 Review comment: ```suggestion CarbonSITransformationRule, checks if there are any index tables present on the filter column of ``` ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
CarbonDataQA1 commented on issue #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#issuecomment-607154069 Build Success with Spark 2.4.5, Please check CI http://121.244.95.60:12545/job/ApacheCarbon_PR_Builder_2.4.5/895/ ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
CarbonDataQA1 commented on issue #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#issuecomment-607154752 Build Success with Spark 2.3.4, Please check CI http://121.244.95.60:12545/job/ApacheCarbonPRBuilder2.3/2604/ ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
akashrn5 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401523505 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as 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 executes a 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. The subsequent main table load will load the old failed loads along with current load and + makes the SI table 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. + + +## Compacting SI table + +### Compacting SI table table through Main Table compaction +Running Compaction command (`ALTER TABLE COMPACT`)[COMPACTION TYPE-> MINOR/MAJOR] on main table will +automatically delete all the old segments of SI and creates a new segment with same name as main +table compacted segmet and loads data to it. + +### Compacting SI table's individual segment(s) through REBUILD command +Where there are so many small files present in the SI table, then we can use REBUILD command to +compact the files within an SI segment to avoid many small files. + + ``` + REBUILD INDEX sales_index + ``` +This command merges data files in each segment of SI table. + + ``` + REBUILD INDEX sales_index WHERE SEGMENT.ID IN(1) + ``` +This command merges data files within specified segment of SI table. + +## How to skip Secondary Index? +When Secondary indexes are created on a table(s), always data fetching happens from secondary +indexes created on the maintables for better performance. But sometimes, data fetching from the +secondary index might degrade query performance. So to avoid such secondary indexes, we use NI as a +function on flters with in WHERE clause. + + ``` + SELECT country, sex from sales where NI(user_id = 'xxx') + ``` +The above query ignores column user_id from secondary index and fetch data from main table. + +## DDLs on Secondary Index + +### Show index Command +This command is used to get information about all the secondary indexes on a table. + +Syntax + ``` + SHOW INDEXES on sales Review comment: since in that doc, i have taken sales tables as example, i gave that, t should be fine right ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
akashrn5 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401523505 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as 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 executes a 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. The subsequent main table load will load the old failed loads along with current load and + makes the SI table 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. + + +## Compacting SI table + +### Compacting SI table table through Main Table compaction +Running Compaction command (`ALTER TABLE COMPACT`)[COMPACTION TYPE-> MINOR/MAJOR] on main table will +automatically delete all the old segments of SI and creates a new segment with same name as main +table compacted segmet and loads data to it. + +### Compacting SI table's individual segment(s) through REBUILD command +Where there are so many small files present in the SI table, then we can use REBUILD command to +compact the files within an SI segment to avoid many small files. + + ``` + REBUILD INDEX sales_index + ``` +This command merges data files in each segment of SI table. + + ``` + REBUILD INDEX sales_index WHERE SEGMENT.ID IN(1) + ``` +This command merges data files within specified segment of SI table. + +## How to skip Secondary Index? +When Secondary indexes are created on a table(s), always data fetching happens from secondary +indexes created on the maintables for better performance. But sometimes, data fetching from the +secondary index might degrade query performance. So to avoid such secondary indexes, we use NI as a +function on flters with in WHERE clause. + + ``` + SELECT country, sex from sales where NI(user_id = 'xxx') + ``` +The above query ignores column user_id from secondary index and fetch data from main table. + +## DDLs on Secondary Index + +### Show index Command +This command is used to get information about all the secondary indexes on a table. + +Syntax + ``` + SHOW INDEXES on sales Review comment: since in that doc, i have taken sales tables as example, i gave that, t should be fine right ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
akashrn5 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401524898 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 + ``` + Review comment: we followed same for all docs, so lets keep same i think ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
akashrn5 commented on a change in pull request #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#discussion_r401526296 ########## File path: docs/index/secondary-index-guide.md ########## @@ -0,0 +1,182 @@ +<!-- + 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) + +## 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 indexes and managed as child tables internally by + Carbondata. Users can create secondary index based on the column position in main 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 as json formatted Review comment: these will be thrown as exceptions right, code validations no need to mention, if there are any thing which not handled in code can be mentioned i think ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
|
|
In reply to this post by GitBox
akashrn5 commented on issue #3689: [CARBONDATA-3680]Add Secondary Index Document
URL: https://github.com/apache/carbondata/pull/3689#issuecomment-607179217 @Indhumathi27 handled all ---------------------------------------------------------------- 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: [hidden email] With regards, Apache Git Services |
«
Return to Apache CarbonData JIRA issues
|
1 view|%1 views
| Free forum by Nabble | Edit this page |