Github user DaveBirdsall commented on a diff in the pull request: https://github.com/apache/trafodion/pull/1399#discussion_r162194005 --- Diff: docs/sql_reference/src/asciidoc/_chapters/sql_functions_and_expressions.adoc --- @@ -6337,6 +6337,300 @@ UPDATE persnl.job SET jobdesc = RIGHT (jobdesc, 12); ``` +<<< +[[rollup_function]] +== ROLLUP Function + +The ROLLUP function calculates multiple levels of subtotals aggregating from right to left through the comma-separated list of columns, and provides a grand total. It is a an extension to the `GROUP BY` clause and can be used with `ORDER BY` to sort the results. + +``` +SELECTâ€¦GROUP BY ROLLUP (column 1, [column 2,]â€¦[column n]) +``` + +ROLLUP generates n+1 levels of subtotals and grand total, where n is the number of the selected column(s). + +For example, a query that contains three rollup columns returns the following rows: + +* First-level: stand aggregate values calculated by GROUP BY clause without using ROLLUP. +* Second-level: subtotals aggregating across column 3 for each combination of column 1 and column 2. +* Third-level: subtotals aggregating across column 2 and column 3 for each column 1. +* Fourth-level: the grand total row. + +NOTE: Trafodion does not support CUBE function which works slightly differently from ROLLUP. + +[[considerations_for_rollup]] +=== Considerations for ROLLUP + +[[null_in_result_sets]] +==== NULL in Result Sets + +* The NULLs in each super-aggregate row represent subtotals and grand total. +* The NULLs in selected columns are considered equal and sorted into one NULL group in result sets. + +[[using_rollup_with_the_column_order_reversed]] +==== Using ROLLUP with the Column Order Reversed + +ROLLUP removes the right-most column at each step, therefore the result sets vary with the column order specified in the comma-separated list. + +[cols="50%,50%"] +|=== +| If the column order is _country_, _state_, _city_ and _name_, ROLLUP returns following groupings. +| If the column order is _name_, _city_, _state_ and _country_, ROLLUP returns following groupings. +| _country_, _state_, _city_ and _name_ | _name_, _city_, _state_ and _country_ +| _country_, _state_ and _city_ | _name_, _city_ and _state_ +| _country_ and _state_ | _name_ and _city_ +| _country_ | _name_ +| grand total | grand total +|=== + +[[examples_of_rollup]] +=== Examples of ROLLUP + +[[examples_of_grouping_by_one_or_multiple_rollup_columns]] +==== Examples of Grouping By One or Multiple Rollup Columns + +Suppose that we have a _sales1_ table like this: + +``` +SELECT * FROM sales1; + +DELIVERY_YEAR REGION PRODUCT REVENUE +------------- ------ -------------------------------- ----------- + 2016 A Dress 100 + 2016 A Dress 200 + 2016 A Pullover 300 + 2016 B Dress 400 + 2017 A Pullover 500 + 2017 B Dress 600 + 2017 B Pullover 700 + 2017 B Pullover 800 + +--- 8 row(s) selected. +``` + +* This is an example of grouping by one rollup column. ++ +``` +SELECT delivery_year, SUM (revenue) AS total_revenue +FROM sales1 +GROUP BY ROLLUP (delivery_year); +``` + ++ +``` +DELIVERY_YEAR TOTAL_REVENUE +------------- -------------------- + 2016 1000 + 2017 2600 + NULL 3600 + +--- 3 row(s) selected. +``` + +* This is an example of grouping by two rollup columns. ++ +ROLLUP firstly aggregates at the lowest level (_region_) and then rollup those aggregations to the next +level (_delivery_year_), finally it produces a grand total across these two levels. + ++ +``` +SELECT delivery_year, region, SUM (revenue) AS total_revenue +FROM sales1 +GROUP BY ROLLUP (delivery_year, region); +``` + ++ +``` +DELIVERY_YEAR REGION TOTAL_REVENUE +------------- ------ -------------------- + 2016 A 600 + 2016 B 400 + 2016 NULL 1000 + 2017 A 500 + 2017 B 2100 + 2017 NULL 2600 + NULL NULL 3600 + +--- 7 row(s) selected. +``` ++ + +* This is an example of grouping by three rollup columns. ++ +``` +SELECT delivery_year, region, product, SUM (revenue) AS total_revenue +FROM sales1 +GROUP BY ROLLUP (delivery_year, region, product); +``` + ++ +.Grouping By Three Rollup Columns +image::grouping-by-three-rollup-columns.jpg[700,700] + ++ +** First-level: the rows marked in *blue* are the total revenue for each year (_2016_ and _2017_), each region (_A_ and _B_) and each product (_Dress_ and _Pullover_), they are caculated by GROUP BY instead of ROLLUP. + ++ +** Second-level: the rows marked in *red* provide the total revenue for the given _delivery_year_ and _region_ by _product_. ++ +These rows have the _product_ columns set to NULL. + ++ +** Third-level: the rows marked in *yellow* show the total revenue in each year (_2016_ and _2017_). ++ +These rows have the _region_ and _product_ columns set to NULL. + ++ +** Fourth-level: the row marked in *purple* aggregates over all rows in the _delivery_year_, _region_ and _product_ columns. ++ +This row has the _delivery_year_, _region_ and _product_ columns set to NULL. + +[[examples_of_null]] +=== Examples of NULL + +The example below demonstrates how ROLLUP treats NULLs in the selected columns and generates NULLs for super-aggregate rows. + +Suppose that we have a _sales2_ table like this: + +``` +SELECT * FROM sales2; + +DELIVERY_YEAR REGION PRODUCT REVENUE +------------- ------ -------------------------------- ----------- + NULL A Dress 100 + NULL A Dress 200 + 2016 A Pullover 300 + 2016 B Dress 400 + 2017 A Pullover 500 + 2017 B Dress 600 + NULL B Pullover 700 + NULL B Pullover 800 + +--- 8 row(s) selected. +``` + +``` +SELECT delivery_year, region, product, SUM (revenue) AS total_revenue +FROM sales2 +GROUP BY ROLLUP (delivery_year, region, product); +``` + +``` +DELIVERY_YEAR REGION PRODUCT TOTAL_REVENUE +------------- ------ -------------------------------- -------------------- + 2016 A Pullover 300 + 2016 A NULL 300 + 2016 B Dress 400 + 2016 B NULL 400 + 2016 NULL NULL 700 + 2017 A Pullover 500 + 2017 A NULL 500 + 2017 B Dress 600 + 2017 B NULL 600 + 2017 NULL NULL 1100 + NULL A Dress 300 + NULL A NULL 300 + NULL B Pullover 1500 + NULL B NULL 1500 + NULL NULL NULL 1800 + NULL NULL NULL 3600 + +--- 16 row(s) selected. +``` + +[[examples_of_using_rollup_with_the_column_order_reversed]] +==== Examples of Using ROLLUP with the Column Order Reversed + +Suppose that we have the same _sale1_ table as shown in the <>. --- End diff -- "_sale1_" should be "_sales1_" ---