feat(tesseract): Athena support (#9707)

Author: waralexrom

.github/workflows/drivers-tests.yml

@@ -271,6 +271,8 @@ jobs:
             use_tesseract_sql_planner: true
           - database: bigquery-export-bucket-gcs
             use_tesseract_sql_planner: true
+          - database: athena-export-bucket-s3
+            use_tesseract_sql_planner: true
       fail-fast: false
 
     steps:

.github/workflows/rust-cubesql.yml

@@ -325,7 +325,7 @@ jobs:
       matrix:
         # We do not need to test under all versions, we do it under linux
         node-version: [22.x]
-        os-version: [windows-2019]
+        os-version: [windows-2022]
         python-version: ["fallback"]
       fail-fast: false
 

packages/cubejs-schema-compiler/src/adapter/BaseQuery.js

@@ -1078,7 +1078,8 @@ export class BaseQuery {
    * @returns {string}
    */
   subtractInterval(date, interval) {
-    return `${date} - interval '${interval}'`;
+    const intervalStr = this.intervalString(interval);
+    return `${date} - interval ${intervalStr}`;
   }
 
   /**
@@ -1087,7 +1088,16 @@ export class BaseQuery {
    * @returns {string}
    */
   addInterval(date, interval) {
-    return `${date} + interval '${interval}'`;
+    const intervalStr = this.intervalString(interval);
+    return `${date} + interval ${intervalStr}`;
+  }
+
+  /**
+   * @param {string} interval
+   * @returns {string}
+   */
+  intervalString(interval) {
+    return `'${interval}'`;
   }
 
   /**
@@ -4111,6 +4121,7 @@ export class BaseQuery {
       },
       tesseract: {
         ilike: '{{ expr }} {% if negated %}NOT {% endif %}ILIKE {{ pattern }}', // May require different overloads in Tesseract than the ilike from expressions used in SQLAPI.
+        series_bounds_cast: '{{ expr }}'
       },
       filters: {
         equals: '{{ column }} = {{ value }}{{ is_null_check }}',

packages/cubejs-schema-compiler/src/adapter/BigqueryQuery.ts

@@ -203,6 +203,10 @@ export class BigqueryQuery extends BaseQuery {
     return this.subtractInterval(timestamp, interval);
   }
 
+  public intervalString(interval: string): string {
+    return `${interval}`;
+  }
+
   public addTimestampInterval(timestamp, interval) {
     return this.addInterval(timestamp, interval);
   }
@@ -353,6 +357,7 @@ export class BigqueryQuery extends BaseQuery {
     delete templates.expressions.like_escape;
     templates.filters.like_pattern = 'CONCAT({% if start_wild %}\'%\'{% else %}\'\'{% endif %}, LOWER({{ value }}), {% if end_wild %}\'%\'{% else %}\'\'{% endif %})';
     templates.tesseract.ilike = 'LOWER({{ expr }}) {% if negated %}NOT {% endif %} LIKE {{ pattern }}';
+    templates.tesseract.series_bounds_cast = 'TIMESTAMP({{ expr }})';
     templates.types.boolean = 'BOOL';
     templates.types.float = 'FLOAT64';
     templates.types.double = 'FLOAT64';

packages/cubejs-schema-compiler/src/adapter/PostgresQuery.ts

@@ -83,11 +83,11 @@ export class PostgresQuery extends BaseQuery {
     templates.types.binary = 'BYTEA';
     templates.operators.is_not_distinct_from = 'IS NOT DISTINCT FROM';
     templates.statements.generated_time_series_select = 'SELECT d AS "date_from",\n' +
-      'd + interval \'{{ granularity }}\' - interval \'1 millisecond\' AS "date_to" \n' +
-      'FROM generate_series({{ start }}::timestamp, {{ end }}:: timestamp, \'{{ granularity }}\'::interval) d ';
+      'd + interval {{ granularity }} - interval \'1 millisecond\' AS "date_to" \n' +
+      'FROM generate_series({{ start }}::timestamp, {{ end }}:: timestamp, {{ granularity }}::interval) d ';
     templates.statements.generated_time_series_with_cte_range_source = 'SELECT d AS "date_from",\n' +
-      'd + interval \'{{ granularity }}\' - interval \'1 millisecond\' AS "date_to" \n' +
-      'FROM {{ range_source }}, LATERAL generate_series({{ range_source }}.{{ min_name }}, {{ range_source }}.{{ max_name }}, \'{{ granularity }}\'::interval) d ';
+      'd + interval {{ granularity }} - interval \'1 millisecond\' AS "date_to" \n' +
+      'FROM {{ range_source }}, LATERAL generate_series({{ range_source }}.{{ min_name }}, {{ range_source }}.{{ max_name }}, {{ granularity }}::interval) d ';
     return templates;
   }
 

packages/cubejs-schema-compiler/src/adapter/PrestodbQuery.ts

@@ -84,14 +84,9 @@ export class PrestodbQuery extends BaseQuery {
     return `date_trunc('${GRANULARITY_TO_INTERVAL[granularity]}', ${dimension})`;
   }
 
-  public subtractInterval(date, interval) {
+  public intervalString(interval: string): string {
     const [intervalValue, intervalUnit] = interval.split(' ');
-    return `${date} - interval '${intervalValue}' ${intervalUnit}`;
-  }
-
-  public addInterval(date, interval) {
-    const [intervalValue, intervalUnit] = interval.split(' ');
-    return `${date} + interval '${intervalValue}' ${intervalUnit}`;
+    return `'${intervalValue}' ${intervalUnit}`;
   }
 
   public seriesSql(timeDimension) {
@@ -138,9 +133,16 @@ export class PrestodbQuery extends BaseQuery {
     templates.functions.DATETRUNC = 'DATE_TRUNC({{ args_concat }})';
     templates.functions.DATEPART = 'DATE_PART({{ args_concat }})';
     delete templates.functions.PERCENTILECONT;
-    templates.statements.select = 'SELECT {{ select_concat | map(attribute=\'aliased\') | join(\', \') }} \n' +
-      'FROM (\n  {{ from }}\n) AS {{ from_alias }} \n' +
+    templates.statements.select = '{% if ctes %} WITH \n' +
+          '{{ ctes | join(\',\n\') }}\n' +
+          '{% endif %}' +
+      'SELECT {{ select_concat | map(attribute=\'aliased\') | join(\', \') }}  {% if from %}\n' +
+      'FROM (\n  {{ from }}\n) AS {{ from_alias }} {% elif from_prepared %}\n' +
+      'FROM {{ from_prepared }}' +
+      '{% endif %}' +
+      '{% if filter %}\nWHERE {{ filter }}{% endif %}' +
       '{% if group_by %} GROUP BY {{ group_by }}{% endif %}' +
+      '{% if having %}\nHAVING {{ having }}{% endif %}' +
       '{% if order_by %} ORDER BY {{ order_by | map(attribute=\'expr\') | join(\', \') }}{% endif %}' +
       '{% if offset is not none %}\nOFFSET {{ offset }}{% endif %}' +
       '{% if limit is not none %}\nLIMIT {{ limit }}{% endif %}';
@@ -153,6 +155,25 @@ export class PrestodbQuery extends BaseQuery {
     // Presto intervals have a YearMonth or DayTime type variants, but no universal type
     delete templates.types.interval;
     templates.types.binary = 'VARBINARY';
+    templates.tesseract.ilike = 'LOWER({{ expr }}) {% if negated %}NOT {% endif %} LIKE {{ pattern }}';
+    templates.filters.like_pattern = 'CONCAT({% if start_wild %}\'%\'{% else %}\'\'{% endif %}, LOWER({{ value }}), {% if end_wild %}\'%\'{% else %}\'\'{% endif %}) ESCAPE \'\\\'';
+    templates.statements.time_series_select = 'SELECT from_iso8601_timestamp(dates.f) date_from, from_iso8601_timestamp(dates.t) date_to \n' +
+    'FROM (\n' +
+    '{% for time_item in seria  %}' +
+    '    select \'{{ time_item[0] }}\' f, \'{{ time_item[1] }}\' t \n' +
+    '{% if not loop.last %} UNION ALL\n{% endif %}' +
+    '{% endfor %}' +
+    ') AS dates';
+    templates.statements.generated_time_series_select = 'SELECT d AS date_from,\n' +
+    'date_add(\'MILLISECOND\', -1, d + interval {{ granularity }}) AS date_to\n' +
+    'FROM UNNEST(\n' +
+    'SEQUENCE(CAST(from_iso8601_timestamp({{ start }}) AS TIMESTAMP), CAST(from_iso8601_timestamp({{ end }}) AS TIMESTAMP), INTERVAL {{ granularity }})\n' +
+    ') AS dates(d)';
+    templates.statements.generated_time_series_with_cte_range_source = 'SELECT d AS date_from,\n' +
+    'date_add(\'MILLISECOND\', -1, d + interval {{ granularity }}) AS date_to\n' +
+    'FROM {{ range_source }} CROSS JOIN UNNEST(\n' +
+    'SEQUENCE(CAST({{ range_source }}.{{ min_name }} AS TIMESTAMP), CAST({{ range_source }}.{{ max_name }} AS TIMESTAMP), INTERVAL {{ granularity }})\n' +
+    ') AS dates(d)';
     return templates;
   }
 

packages/cubejs-testing-drivers/fixtures/athena.json

@@ -174,5 +174,41 @@
     "SQL API: Nested Rollup with aliases",
     "SQL API: Nested Rollup over asterisk",
     "SQL API: Extended nested Rollup over asterisk"
+  ],
+  "tesseractSkip": [
+    "for the ECommerce.TimeAnalysisExternal",
+    "for the ECommerce.TimeAnalysisInternal",
+
+    "querying Products: dimensions -- doesn't work wo ordering",
+    "querying ECommerce: total quantity, avg discount, total sales, total profit by product + order + total -- rounding in athena",
+    "querying ECommerce: total sales, total profit by month + order (date) + total -- doesn't work with the BigQuery",
+    "querying ECommerce: total quantity, avg discount, total sales, total profit by product + order + total -- noisy test",
+    "querying BigECommerce: partitioned pre-agg",
+    "querying BigECommerce: null sum",
+    "querying BigECommerce: null boolean",
+    "--------------------",
+
+
+    "querying BigECommerce: rolling window by 2 week",
+    "querying custom granularities ECommerce: count by three_months_by_march + no dimension",
+    "querying custom granularities ECommerce: count by three_months_by_march + dimension",
+    "querying custom granularities (with preaggregation) ECommerce: totalQuantity by half_year + no dimension",
+    "querying custom granularities (with preaggregation) ECommerce: totalQuantity by half_year + dimension",
+    "querying custom granularities ECommerce: count by two_mo_by_feb + no dimension + rollingCountByUnbounded",
+    "querying custom granularities ECommerce: count by two_mo_by_feb + no dimension + rollingCountByTrailing",
+    "querying custom granularities ECommerce: count by two_mo_by_feb + no dimension + rollingCountByLeading",
+    "pre-aggregations Customers: running total without time dimension",
+    "querying BigECommerce: totalProfitYearAgo",
+    "SQL API: post-aggregate percentage of total",
+    "SQL API: Simple Rollup",
+    "SQL API: Complex Rollup",
+    "SQL API: Nested Rollup",
+    "SQL API: Rollup with aliases",
+    "SQL API: Rollup over exprs",
+    "SQL API: Nested Rollup with aliases",
+    "SQL API: Nested Rollup over asterisk",
+    "SQL API: Extended nested Rollup over asterisk",
+    "SQL API: Timeshift measure from cube",
+    "SQL API: SQL push down push to cube quoted alias"
   ]
 }