feat(cubesql): support GREATEST/LEAST SQL functions (#8325)

* feat(cubesql): allow float scalars in LEAST/GREATEST functions * feat(cubesql): add support for the greatest function with 2 parameters * feat(cubesql): support GREATEST sql function with arbitrary number of arguments * feat(cubesql): support LEAST sql function with arbitrary number of arguments * feat(cubesql): propogate GREATEST/LEAST support to Postgres, MS SQL, DuckDB and Databricks drivers
cube-js · Jun 13, 2024 · c13a28e · c13a28e
1 parent a4ca555
commit c13a28e
Show file tree

Hide file tree

Showing 9 changed files with 415 additions and 85 deletions.
diff --git a/packages/cubejs-databricks-jdbc-driver/src/DatabricksQuery.ts b/packages/cubejs-databricks-jdbc-driver/src/DatabricksQuery.ts
@@ -124,6 +124,8 @@ export class DatabricksQuery extends BaseQuery {
     templates.functions.LTRIM = 'LTRIM({{ args|reverse|join(", ") }})';
     templates.functions.RTRIM = 'RTRIM({{ args|reverse|join(", ") }})';
     templates.functions.DATEDIFF = 'DATEDIFF({{ date_part }}, DATE_TRUNC(\'{{ date_part }}\', {{ args[1] }}), DATE_TRUNC(\'{{ date_part }}\', {{ args[2] }}))';
+    templates.functions.LEAST = 'LEAST({{ args_concat }})';
+    templates.functions.GREATEST = 'GREATEST({{ args_concat }})';
     templates.expressions.timestamp_literal = 'from_utc_timestamp(\'{{ value }}\', \'UTC\')';
     templates.quotes.identifiers = '`';
     templates.quotes.escape = '``';

diff --git a/packages/cubejs-duckdb-driver/src/DuckDBQuery.ts b/packages/cubejs-duckdb-driver/src/DuckDBQuery.ts
@@ -34,6 +34,8 @@ export class DuckDBQuery extends BaseQuery {
   public sqlTemplates() {
     const templates = super.sqlTemplates();
     templates.functions.DATETRUNC = 'DATE_TRUNC({{ args_concat }})';
+    templates.functions.LEAST = 'LEAST({{ args_concat }})';
+    templates.functions.GREATEST = 'GREATEST({{ args_concat }})';
     return templates;
   }
 }
diff --git a/packages/cubejs-schema-compiler/src/adapter/BaseQuery.js b/packages/cubejs-schema-compiler/src/adapter/BaseQuery.js
@@ -2954,7 +2954,16 @@ export class BaseQuery {
         FLOOR: 'FLOOR({{ args_concat }})',
         CEIL: 'CEIL({{ args_concat }})',
         TRUNC: 'TRUNC({{ args_concat }})',
-        LEAST: 'LEAST({{ args_concat }})',
+
+        // There is a difference in behaviour of these function processing in different DBs and DWHs.
+        // The SQL standard requires greatest and least to return null in case one argument is null.
+        // However, many DBMS ignore NULL values (mostly because greatest and least were often supported
+        // decades before they were added to the SQL standard in 2023).
+        // Cube follows the Postgres implementation (as we mimic the Postgres protocol) and ignores NULL values.
+        // So these functions are enabled on a driver-specific basis for databases that ignores NULLs.
+        // LEAST: 'LEAST({{ args_concat }})',
+        // GREATEST: 'GREATEST({{ args_concat }})',
+
         LOWER: 'LOWER({{ args_concat }})',
         UPPER: 'UPPER({{ args_concat }})',
         LEFT: 'LEFT({{ args_concat }})',
@@ -2976,7 +2985,6 @@ export class BaseQuery {
         REPEAT: 'REPEAT({{ args_concat }})',
         NULLIF: 'NULLIF({{ args_concat }})',
         ROUND: 'ROUND({{ args_concat }})',
-        GREATEST: 'GREATEST({{ args_concat }})',
 
         STDDEV: 'STDDEV_SAMP({{ args_concat }})',
         SUBSTR: 'SUBSTRING({{ args_concat }})',

diff --git a/packages/cubejs-schema-compiler/src/adapter/MssqlQuery.ts b/packages/cubejs-schema-compiler/src/adapter/MssqlQuery.ts
@@ -128,15 +128,15 @@ export class MssqlQuery extends BaseQuery {
     const timeDimensionsColumns = this.timeDimensions.map(
       (t) => `${t.dateSeriesAliasName()}.${this.escapeColumnName('date_from')}`
     );
-  
+
     // Group by regular dimensions
     const dimensionColumns = R.flatten(
       this.dimensions.map(s => s.selectColumns() && s.dimensionSql() && s.aliasName())
     ).filter(s => !!s);
-  
+
     // Combine time dimensions and regular dimensions for GROUP BY clause
     const allGroupByColumns = timeDimensionsColumns.concat(dimensionColumns);
-  
+
     const forSelect = this.overTimeSeriesForSelect(cumulativeMeasures);
     return (
       `SELECT ${forSelect} FROM ${dateSeriesSql}` +
@@ -180,4 +180,11 @@ export class MssqlQuery extends BaseQuery {
     const amountInterval = interval.split(' ', 2);
     return `DATEADD(${amountInterval[1]}, ${amountInterval[0]}, ${date})`;
   }
+
+  public sqlTemplates() {
+    const templates = super.sqlTemplates();
+    templates.functions.LEAST = 'LEAST({{ args_concat }})';
+    templates.functions.GREATEST = 'GREATEST({{ args_concat }})';
+    return templates;
+  }
 }
diff --git a/packages/cubejs-schema-compiler/src/adapter/PostgresQuery.ts b/packages/cubejs-schema-compiler/src/adapter/PostgresQuery.ts
@@ -51,6 +51,8 @@ export class PostgresQuery extends BaseQuery {
     templates.functions.CONCAT = 'CONCAT({% for arg in args %}CAST({{arg}} AS TEXT){% if not loop.last %},{% endif %}{% endfor %})';
     templates.functions.DATEPART = 'DATE_PART({{ args_concat }})';
     templates.functions.CURRENTDATE = 'CURRENT_DATE';
+    templates.functions.LEAST = 'LEAST({{ args_concat }})';
+    templates.functions.GREATEST = 'GREATEST({{ args_concat }})';
     templates.functions.NOW = 'NOW({{ args_concat }})';
     // DATEADD is being rewritten to DATE_ADD
     // templates.functions.DATEADD = '({{ args[2] }} + \'{{ interval }} {{ date_part }}\'::interval)';

diff --git a/rust/cubesql/cubesql/src/compile/engine/df/coerce.rs b/rust/cubesql/cubesql/src/compile/engine/df/coerce.rs
@@ -37,6 +37,9 @@ pub fn numerical_coercion(lhs_type: &DataType, rhs_type: &DataType) -> Option<Da
         (_, DataType::Null) => Some(lhs_type.clone()),
         (DataType::Null, _) => Some(rhs_type.clone()),
         //
+        (_, DataType::Float64) => Some(DataType::Float64),
+        (DataType::Float64, _) => Some(DataType::Float64),
+        //
         (_, DataType::UInt64) => Some(DataType::UInt64),
         (DataType::UInt64, _) => Some(DataType::UInt64),
         //