Skip to content

The easy-to-use open source Business Intelligence and Embedded Analytics tool that lets everyone work with data :bar_chart:

BigQuery: reject oversized queries client-side with a humanized error (#79020)

dace9982a7e6ac
diff --git a/modules/drivers/bigquery-cloud-sdk/src/metabase/driver/bigquery_cloud_sdk.clj b/modules/drivers/bigquery-cloud-sdk/src/metabase/driver/bigquery_cloud_sdk.clj
index 8db77324aec..e4c7990ff6e 100644
--- a/modules/drivers/bigquery-cloud-sdk/src/metabase/driver/bigquery_cloud_sdk.clj
+++ b/modules/drivers/bigquery-cloud-sdk/src/metabase/driver/bigquery_cloud_sdk.clj
@@ -723,6 +723,30 @@
page, then adaptive sizing) by default, but override for testing."
nil)
+(def ^:private max-sql-query-length-chars
+ "BigQuery's maximum standard SQL query length, in characters — counting comments and whitespace exactly as
+ BigQuery does. BigQuery rejects jobs above this with a raw 400 `INVALID_ARGUMENT` ('The query is too large ...').
+ See
+ https://cloud.google.com/bigquery/quotas#query_limits ('Maximum query length: 1 MB')."
+ (* 1024 1024))
+
+(defn- validate-query-length!
+ "Throw a localized `invalid-query` error if `sql` exceeds [[max-sql-query-length-chars]], before the request
+ reaches BigQuery. `sql` is the final payload — Metabase's `-- remark` comment, appended in
+ [[driver/execute-reducible-query]], is already included — so it is counted exactly as BigQuery will count it. The
+ raw 400 from BigQuery remains a fallback if this limit ever drifts from BigQuery's own."
+ [^String sql parameters]
+ (let [len (count sql)]
+ (when (> len max-sql-query-length-chars)
+ (throw
+ (ex-info
+ (tru (str "This query is too large for BigQuery ({0} characters; the maximum is {1}). "
+ "Try reducing the number of selected fields, filter values, or rewriting the query to make it shorter.")
+ len max-sql-query-length-chars)
+ {:type driver-api/qp.error-type.invalid-query
+ :sql sql
+ :parameters parameters})))))
+
(defn- throw-invalid-query [e sql parameters]
(throw (ex-info (tru "Error executing query: {0}" (ex-message e))
{:type driver-api/qp.error-type.invalid-query, :sql sql, :parameters parameters}
@@ -879,6 +903,7 @@
(defn- execute-bigquery
[respond database-details ^String sql parameters cancel-chan]
{:pre [(not (str/blank? sql))]}
+ (validate-query-length! sql parameters)
;; Kicking off two async jobs:
;; - Waiting for the cancel-chan to get either a cancel message or to be closed.
;; - Running the BigQuery execution in another thread, since it's blocking.
diff --git a/modules/drivers/bigquery-cloud-sdk/test/metabase/driver/bigquery_cloud_sdk_test.clj b/modules/drivers/bigquery-cloud-sdk/test/metabase/driver/bigquery_cloud_sdk_test.clj
index 1792a7a48af..e8b657eea01 100644
--- a/modules/drivers/bigquery-cloud-sdk/test/metabase/driver/bigquery_cloud_sdk_test.clj
+++ b/modules/drivers/bigquery-cloud-sdk/test/metabase/driver/bigquery_cloud_sdk_test.clj
@@ -894,6 +894,17 @@
(is (re-find #"Too many query parameters" (ex-message ex))
"the underlying BigQuery error message should be preserved"))))))
+(deftest validate-query-length-test
+ (let [limit @#'bigquery/max-sql-query-length-chars]
+ (testing "at the limit, the query is allowed"
+ (is (nil? (#'bigquery/validate-query-length! (str/join (repeat limit "x")) nil))))
+ (testing "over the limit, it throws :invalid-query"
+ (let [ex (try
+ (#'bigquery/validate-query-length! (str/join (repeat (inc limit) "x")) nil)
+ nil
+ (catch Throwable t t))]
+ (is (= :invalid-query (some-> ex ex-data :type)))))))
+
(deftest project-id-override-test
(mt/test-driver :bigquery-cloud-sdk
(testing "Querying a different project-id works"