feat: add BigQuery dialect

packages/mosaic/sql/src/visit/codegen/bigquery.ts

@@ -0,0 +1,106 @@
+import {
+  AggregateNode,
+  CastNode,
+  FunctionNode,
+  JoinNode,
+  SampleClauseNode,
+  UnnestNode,
+} from "../../index.js";
+import { DuckDBCodeGenerator } from "./duckdb.js";
+
+/**
+ * BigQuery SQL dialect visitor for converting AST nodes to BigQuery-compatible SQL.
+ */
+export class BigQueryCodeGenerator extends DuckDBCodeGenerator {
+  visitAggregate(node: AggregateNode): string {
+    const { filter } = node;
+    // TODO: it might be possible to emulate filters using an IF statement
+    if (filter) {
+      throw new Error(`BigQuery does not support aggregate filters. Node: ${JSON.stringify(node)}`);
+    }
+    return super.visitAggregate(node);
+  }
+
+  visitCast(node: CastNode): string {
+    const { expr, cast } = node;
+    return `CAST(${this.toString(expr)} AS ${cast})`;
+  }
+
+  visitFunction(node: FunctionNode): string {
+    const { name, args } = node;
+    switch (name) {
+      case "epoch_ms":
+        return `UNIX_MILLIS(${this.mapToString(args).join(', ')})`;
+
+      case "list_contains":
+        return `${this.toString(args[1])} IN UNNEST(${this.toString(args[0])})`;
+
+      case "list_has_all":
+        return `(SELECT COUNT(l2) = ARRAY_LENGTH(${this.toString(args[1])}) FROM UNNEST(${this.toString(args[0])}) AS l1 JOIN UNNEST(${this.toString(args[1])}) AS l2 ON l1 = l2)"`;
+
+      case "list_has_any":
+        return `EXISTS(SELECT * FROM UNNEST(${this.toString(args[0])}) AS l1 WHERE l1 IN UNNEST(${this.toString(args[1])}))`;
+
+      case "make_date":
+        return `$DATE(${this.mapToString(args).join(', ')})`;
+
+      case "time_bucket":
+        return `TIMESTAMP_BUCKET(${this.toString(args[0])}, ${this.toString(args[1])})`;
+
+      default:
+        return super.visitFunction(node);
+    }
+  }
+
+  visitJoinClause(node: JoinNode): string {
+    const { joinVariant, joinType, sample } = node;
+
+    if (joinVariant !== "REGULAR") {
+      throw new Error(`BigQuery does not support (${joinVariant}) joins (variant). Node: ${JSON.stringify(node)}`);
+    }
+
+    if (joinType === "SEMI" || joinType === "ANTI") {
+      throw new Error(`BigQuery does not support (${joinType}) joins (type). Node: ${JSON.stringify(node)}`);
+    }
+
+    if (sample) {
+      throw new Error(`BigQuery does not support join sampling (${sample}). Node: ${JSON.stringify(node)}`);
+    }
+
+    return super.visitJoinClause(node);
+  }
+
+  visitSampleClause(node: SampleClauseNode): string {
+    const { size, perc, method, seed } = node;
+
+    if (!perc) {
+      throw new Error(`BigQuery does not support row based sampling. Node: ${JSON.stringify(node)}`);
+    }
+
+    if (method !== "system") {
+      throw new Error(`BigQuery does not support (${method}) sampling. Node: ${JSON.stringify(node)}`);
+    }
+
+    if (seed !== null) {
+      throw new Error(`BigQuery does not support setting a seed (${seed}) for sampling. Node: ${JSON.stringify(node)}`);
+    }
+
+    return `TABLESAMPLE SYSTEM (${size} PERCENT)`;
+  }
+
+  visitUnnest(node: UnnestNode): string {
+    const { expr, recursive, maxDepth } = node;
+
+    if (recursive) {
+      throw new Error(`BigQuery does not support recursive unnesting. Node: ${JSON.stringify(node)}`);
+    }
+
+    if (maxDepth != null && maxDepth > 0) {
+      throw new Error(`BigQuery does not support max depth when unnesting. Node: ${JSON.stringify(node)}`);
+    }
+
+    const args = [this.toString(expr)];
+
+    return `UNNEST(${args.join(", ")})`;
+  }
+}