srielau commented on code in PR #59873: URL: https://github.com/apache/spark/pull/59873#discussion_r4467787714
########## sql/core/src/test/scala/org/apache/spark/sql/catalyst/expressions/ParseSqlAnalyzeSuite.scala: ########## @@ -0,0 +1,414 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.sql.catalyst.expressions + +import java.util.Collections +import java.util.concurrent.atomic.AtomicInteger + +import org.json4s._ +import org.json4s.jackson.JsonMethods.parse + +import org.apache.spark.sql.{AnalysisException, QueryTest} +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.connector.catalog.{Identifier, InMemoryCatalog} +import org.apache.spark.sql.connector.catalog.procedures.{ + BoundProcedure, + ProcedureParameter, + UnboundProcedure +} +import org.apache.spark.sql.connector.read.Scan +import org.apache.spark.sql.internal.SQLConf +import org.apache.spark.sql.test.SharedSparkSession +import org.apache.spark.sql.types.{DataTypes, StructType} + +class ParseSqlAnalyzeSuite extends QueryTest with SharedSparkSession { + + private def evalJson(sql: String, mode: String = "ANALYZE"): JValue = { + val result = spark.sql(s"SELECT parse_sql($sql, '$mode')").head().getString(0) + parse(result) + } + + private def evalStatement(sql: String, mode: String = "ANALYZE"): JObject = { + evalJson("'" + sql.replace("'", "''") + "'", mode) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + } + + private def jsonType(name: String): JObject = JObject("name" -> JString(name)) + private val jsonInt: JObject = jsonType("int") + private val jsonBigint: JObject = jsonType("bigint") + private val jsonBoolean: JObject = jsonType("boolean") + private val jsonString: JObject = JObject( + "name" -> JString("string"), + "collation" -> JString("UTF8_BINARY")) + + test("ANALYZE types selected columns and PARSE does not") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val analyzed = evalStatement("SELECT 1 AS a", "ANALYZE") + assert(analyzed \ "parse_success" === JBool(true)) + assert(analyzed \ "analyze_success" === JBool(true)) + assert(analyzed \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))), "type" -> jsonInt)))) + + val parsed = evalStatement("SELECT 1 AS a", "PARSE") + assert(parsed \ "analyze_success" === JNothing) + assert(parsed \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))))))) + } + } + + test("ANALYZE fully qualifies tables and builtin routines") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_analyze_t") { + sql("CREATE TABLE parse_sql_analyze_t (a INT, b STRING) USING parquet") + val j = evalStatement( + "SELECT count(a) AS c FROM parse_sql_analyze_t", "analyze") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), JString("default"), JString("parse_sql_analyze_t")))))) + assert(j \ "function_references" === JArray(List(JArray(List( + JString("system"), JString("builtin"), JString("count")))))) + assert(j \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("c"))), "type" -> jsonBigint)))) + } + } + } + + test("ANALYZE qualifies local temp views as system.session") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTempView("parse_sql_analyze_v") { + sql("CREATE OR REPLACE TEMP VIEW parse_sql_analyze_v AS SELECT 1 AS a, 'x' AS b") + val j = evalStatement("SELECT a, b FROM parse_sql_analyze_v") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "source_table_references" === JArray(List(JArray(List( + JString("system"), JString("session"), JString("parse_sql_analyze_v")))))) + assert(j \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))), "type" -> jsonInt), + JObject("name" -> JArray(List(JString("b"))), "type" -> jsonString)))) + } + } + } + + test("ANALYZE keeps parse metadata when analysis fails") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val j = evalStatement("SELECT a FROM missing_parse_sql_analyze_t") + assert(j \ "parse_success" === JBool(true)) + assert(j \ "analyze_success" === JBool(false)) + assert(j \ "statement_identifier" === JString("SELECT")) + assert(j \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), + JString("default"), + JString("missing_parse_sql_analyze_t")))))) + val errorClass = (j \ "error" \ "errorClass").asInstanceOf[JString].s + assert(errorClass == "TABLE_OR_VIEW_NOT_FOUND" || + errorClass.contains("TABLE_OR_VIEW_NOT_FOUND")) + } + } + + test("ANALYZE inherits the invoker schema") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + sql("CREATE DATABASE parse_sql_ctx_a") + sql("CREATE DATABASE parse_sql_ctx_b") + try { + sql("CREATE TABLE parse_sql_ctx_a.t (a INT) USING parquet") + sql("USE parse_sql_ctx_b") + try { + val missing = evalStatement("SELECT a FROM t") + assert(missing \ "analyze_success" === JBool(false)) + assert(missing \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), JString("parse_sql_ctx_b"), JString("t")))))) + } finally { + sql("USE parse_sql_ctx_a") + } + try { + val found = evalStatement("SELECT a FROM t") + assert(found \ "analyze_success" === JBool(true)) + assert(found \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), JString("parse_sql_ctx_a"), JString("t")))))) + assert(spark.sql("SELECT current_schema()").head().getString(0) === + "parse_sql_ctx_a") + } finally { + sql("USE default") + } + } finally { + sql("DROP TABLE IF EXISTS parse_sql_ctx_a.t") + sql("DROP DATABASE IF EXISTS parse_sql_ctx_a CASCADE") + sql("DROP DATABASE IF EXISTS parse_sql_ctx_b CASCADE") + } + } + } + + test("ANALYZE named mode argument") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val json = spark.sql( + """SELECT parse_sql( + | 'SELECT 1 AS a', + | mode => 'ANALYZE')""".stripMargin).head().getString(0) + val analyzed = parse(json) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + assert(analyzed \ "analyze_success" === JBool(true)) + assert(analyzed \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))), "type" -> jsonInt)))) + } + } + + test("ANALYZE without parameter marker types leaves the query unresolved") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val json = spark.sql("SELECT parse_sql('SELECT abs(?) AS x', 'ANALYZE')") + .head().getString(0) + val analyzed = parse(json) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + assert(analyzed \ "parse_success" === JBool(true)) + assert(analyzed \ "analyze_success" === JBool(false)) + } + } + + private def evalBatch(sql: String): List[JObject] = + evalJson("'" + sql.replace("'", "''") + "'") match { + case JArray(values) => values.map(_.asInstanceOf[JObject]) + case other => fail(s"expected a JSON array, got: $other") + } + + private def typedColumns(columns: (String, JObject)*): JArray = JArray(columns.map { + case (name, dataType) => JObject("name" -> JArray(List(JString(name))), "type" -> dataType) + }.toList) + + private def ref(parts: String*): JArray = JArray(parts.map(JString).toList) + + /** Run a full `SELECT parse_sql(...)` and return its statement objects. */ + private def evalCall(query: String): List[JObject] = + parse(spark.sql(query).head().getString(0)) match { + case JArray(values) => values.map(_.asInstanceOf[JObject]) + case other => fail(s"expected a JSON array, got: $other") + } + + private def errorClass(statement: JObject): JValue = statement \ "error" \ "errorClass" + + test("ANALYZE does not run CALL, directly or through EXECUTE IMMEDIATE") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + spark.conf.set("spark.sql.catalog.parse_sql_cat", classOf[InMemoryCatalog].getName) + try { + val catalog = spark.sessionState.catalogManager.catalog("parse_sql_cat") + .asInstanceOf[InMemoryCatalog] + catalog.createProcedure( + Identifier.of(Array("ns"), "record"), ParseSqlAnalyzeSuite.UnboundRecord) + ParseSqlAnalyzeSuite.calls.set(0) + val results = evalBatch( + """CALL parse_sql_cat.ns.record(1); + |EXECUTE IMMEDIATE 'CALL parse_sql_cat.ns.record(2)'""".stripMargin) + assert(results.map(_ \ "parse_success") === List(JBool(true), JBool(true))) + assert(results.map(_ \ "analyze_success") === List(JBool(true), JBool(true))) + assert(ParseSqlAnalyzeSuite.calls.get === 0) + sql("CALL parse_sql_cat.ns.record(3)").collect() + assert(ParseSqlAnalyzeSuite.calls.get === 1) + } finally { + spark.sessionState.catalogManager.reset() + spark.sessionState.conf.unsetConf("spark.sql.catalog.parse_sql_cat") + } + } + } + + test("ANALYZE inside a SQL script does not see the script's locals") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val json = spark.sql( + """BEGIN + | DECLARE parse_sql_outer INT DEFAULT 5; + | SELECT parse_sql('SELECT parse_sql_outer AS x', 'ANALYZE'); + |END""".stripMargin).head().getString(0) + val statement = parse(json) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + assert(statement \ "analyze_success" === JBool(false)) + val cls = (statement \ "error" \ "errorClass").asInstanceOf[JString].s + assert(cls === "UNRESOLVED_COLUMN.WITHOUT_SUGGESTION" || + cls === "UNRESOLVED_COLUMN.WITH_SUGGESTION" || + cls.startsWith("UNRESOLVED_COLUMN")) + } + } + + test("ANALYZE requires constant arguments; PARSE does not") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val query = "SELECT parse_sql(c, '%s') FROM VALUES ('SELECT 1') AS t(c)" + val e = intercept[AnalysisException](spark.sql(query.format("ANALYZE")).collect()) + assert(e.getCondition === "DATATYPE_MISMATCH.NON_FOLDABLE_INPUT") + assert(spark.sql(query.format("PARSE")).collect().length === 1) + val modeCol = intercept[AnalysisException]( + spark.sql("SELECT parse_sql(c, m) FROM VALUES ('SELECT 1', 'PARSE') AS t(c, m)") + .collect()) + assert(modeCol.getCondition === "DATATYPE_MISMATCH.NON_FOLDABLE_INPUT") + } + } + + test("ANALYZE does not execute SQL scripts") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_script_t", "parse_sql_script_u") { + sql("CREATE TABLE parse_sql_script_t (a INT) USING parquet") + sql("CREATE TABLE parse_sql_script_u (a INT) USING parquet") + val schema = spark.sql("SELECT current_schema()").head().getString(0) + val j = evalStatement( + """BEGIN + | INSERT INTO parse_sql_script_t VALUES (1); + | DROP TABLE parse_sql_script_u; + |END""".stripMargin) + assert(j \ "parse_success" === JBool(true)) + assert(j \ "analyze_success" === JBool(false)) + assert(errorClass(j) === JString("PARSE_SQL_ANALYZE_UNSUPPORTED_SCRIPT")) + assert(spark.table("parse_sql_script_t").count() === 0) + assert(spark.catalog.tableExists("parse_sql_script_u")) + assert(spark.sql("SELECT current_schema()").head().getString(0) === schema) + } + } + } + + test("ANALYZE reports range as system.builtin and de-duplicates count") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_dedup_t") { + sql("CREATE TABLE parse_sql_dedup_t (a INT, b INT) USING parquet") + val rangeStmt = evalStatement("SELECT * FROM range(3)") + assert(rangeStmt \ "analyze_success" === JBool(true)) + assert(rangeStmt \ "function_references" === JArray(List( + ref("system", "builtin", "range")))) + val counts = evalStatement( + "SELECT count(a), COUNT(b) FROM parse_sql_dedup_t") + assert(counts \ "analyze_success" === JBool(true)) + assert(counts \ "function_references" === JArray(List( + ref("system", "builtin", "count")))) + } + } + } + + test("ANALYZE qualifies DDL targets by statement kind") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val tempView = evalStatement("CREATE TEMPORARY VIEW parse_sql_ddl_v AS SELECT 1 AS a") + assert(tempView \ "analyze_success" === JBool(true)) + assert(tempView \ "target_table_references" === JArray(List( + ref("system", "session", "parse_sql_ddl_v")))) + assert(!spark.catalog.tableExists("parse_sql_ddl_v")) + withTempView("parse_sql_ddl_t") { + sql("CREATE TEMP VIEW parse_sql_ddl_t AS SELECT 1 AS a") + val table = evalStatement("CREATE TABLE parse_sql_ddl_t (a INT) USING parquet") + assert(table \ "analyze_success" === JBool(true)) + assert(table \ "target_table_references" === JArray(List( + ref("spark_catalog", "default", "parse_sql_ddl_t")))) + } + } + } + + test("ANALYZE types session-catalog CTAS query columns") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val j = evalStatement( + "CREATE TABLE parse_sql_ctas_t USING parquet AS SELECT 1 AS a") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "select_list" === typedColumns("a" -> jsonInt)) + assert(!spark.catalog.tableExists("parse_sql_ctas_t")) + } + } + + test("ANALYZE does not execute a mixed DDL/DML batch") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_batch_t") { + sql("CREATE TABLE parse_sql_batch_t (a INT) USING parquet") + val results = evalBatch( + """INSERT INTO parse_sql_batch_t VALUES (1); + |CREATE TABLE parse_sql_batch_u (a INT) USING parquet; + |DROP TABLE parse_sql_batch_t""".stripMargin) + assert(results.map(_ \ "analyze_success") === + List(JBool(true), JBool(true), JBool(true))) + assert(spark.table("parse_sql_batch_t").count() === 0) + assert(!spark.catalog.tableExists("parse_sql_batch_u")) + assert(spark.catalog.tableExists("parse_sql_batch_t")) + } + } + } + + test("ANALYZE qualifies global temp views and DROP of local temp views") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withGlobalTempView("parse_sql_global_v") { + sql("CREATE OR REPLACE GLOBAL TEMP VIEW parse_sql_global_v AS SELECT 1 AS a") + val globalDb = spark.conf.get("spark.sql.globalTempDatabase") + val j = evalStatement(s"SELECT a FROM $globalDb.parse_sql_global_v") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "source_table_references" === JArray(List( + ref("spark_catalog", globalDb, "parse_sql_global_v")))) + } + withTempView("parse_sql_drop_v") { + sql("CREATE TEMP VIEW parse_sql_drop_v AS SELECT 1 AS a") + val drop = evalStatement("DROP VIEW parse_sql_drop_v") + assert(drop \ "analyze_success" === JBool(true)) + assert(drop \ "target_table_references" === JArray(List( + ref("system", "session", "parse_sql_drop_v")))) + assert(spark.catalog.tableExists("parse_sql_drop_v")) + } + } + } + + test("ANALYZE types EXPLAIN and DESCRIBE QUERY inner plans") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val explain = evalStatement("EXPLAIN SELECT 1 AS a") Review Comment: Fixed in e19b9d04cf1. The EXPLAIN test now uses `EXPLAIN SELECT a FROM <table>` (typed column and qualified source) and `EXPLAIN SELECT nope FROM missing_t` (`analyze_success` false, `TABLE_OR_VIEW_NOT_FOUND`). ########## sql/core/src/test/scala/org/apache/spark/sql/catalyst/expressions/ParseSqlAnalyzeSuite.scala: ########## @@ -0,0 +1,414 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.spark.sql.catalyst.expressions + +import java.util.Collections +import java.util.concurrent.atomic.AtomicInteger + +import org.json4s._ +import org.json4s.jackson.JsonMethods.parse + +import org.apache.spark.sql.{AnalysisException, QueryTest} +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.connector.catalog.{Identifier, InMemoryCatalog} +import org.apache.spark.sql.connector.catalog.procedures.{ + BoundProcedure, + ProcedureParameter, + UnboundProcedure +} +import org.apache.spark.sql.connector.read.Scan +import org.apache.spark.sql.internal.SQLConf +import org.apache.spark.sql.test.SharedSparkSession +import org.apache.spark.sql.types.{DataTypes, StructType} + +class ParseSqlAnalyzeSuite extends QueryTest with SharedSparkSession { + + private def evalJson(sql: String, mode: String = "ANALYZE"): JValue = { + val result = spark.sql(s"SELECT parse_sql($sql, '$mode')").head().getString(0) + parse(result) + } + + private def evalStatement(sql: String, mode: String = "ANALYZE"): JObject = { + evalJson("'" + sql.replace("'", "''") + "'", mode) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + } + + private def jsonType(name: String): JObject = JObject("name" -> JString(name)) + private val jsonInt: JObject = jsonType("int") + private val jsonBigint: JObject = jsonType("bigint") + private val jsonBoolean: JObject = jsonType("boolean") + private val jsonString: JObject = JObject( + "name" -> JString("string"), + "collation" -> JString("UTF8_BINARY")) + + test("ANALYZE types selected columns and PARSE does not") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val analyzed = evalStatement("SELECT 1 AS a", "ANALYZE") + assert(analyzed \ "parse_success" === JBool(true)) + assert(analyzed \ "analyze_success" === JBool(true)) + assert(analyzed \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))), "type" -> jsonInt)))) + + val parsed = evalStatement("SELECT 1 AS a", "PARSE") + assert(parsed \ "analyze_success" === JNothing) + assert(parsed \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))))))) + } + } + + test("ANALYZE fully qualifies tables and builtin routines") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_analyze_t") { + sql("CREATE TABLE parse_sql_analyze_t (a INT, b STRING) USING parquet") + val j = evalStatement( + "SELECT count(a) AS c FROM parse_sql_analyze_t", "analyze") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), JString("default"), JString("parse_sql_analyze_t")))))) + assert(j \ "function_references" === JArray(List(JArray(List( + JString("system"), JString("builtin"), JString("count")))))) + assert(j \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("c"))), "type" -> jsonBigint)))) + } + } + } + + test("ANALYZE qualifies local temp views as system.session") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTempView("parse_sql_analyze_v") { + sql("CREATE OR REPLACE TEMP VIEW parse_sql_analyze_v AS SELECT 1 AS a, 'x' AS b") + val j = evalStatement("SELECT a, b FROM parse_sql_analyze_v") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "source_table_references" === JArray(List(JArray(List( + JString("system"), JString("session"), JString("parse_sql_analyze_v")))))) + assert(j \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))), "type" -> jsonInt), + JObject("name" -> JArray(List(JString("b"))), "type" -> jsonString)))) + } + } + } + + test("ANALYZE keeps parse metadata when analysis fails") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val j = evalStatement("SELECT a FROM missing_parse_sql_analyze_t") + assert(j \ "parse_success" === JBool(true)) + assert(j \ "analyze_success" === JBool(false)) + assert(j \ "statement_identifier" === JString("SELECT")) + assert(j \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), + JString("default"), + JString("missing_parse_sql_analyze_t")))))) + val errorClass = (j \ "error" \ "errorClass").asInstanceOf[JString].s + assert(errorClass == "TABLE_OR_VIEW_NOT_FOUND" || + errorClass.contains("TABLE_OR_VIEW_NOT_FOUND")) + } + } + + test("ANALYZE inherits the invoker schema") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + sql("CREATE DATABASE parse_sql_ctx_a") + sql("CREATE DATABASE parse_sql_ctx_b") + try { + sql("CREATE TABLE parse_sql_ctx_a.t (a INT) USING parquet") + sql("USE parse_sql_ctx_b") + try { + val missing = evalStatement("SELECT a FROM t") + assert(missing \ "analyze_success" === JBool(false)) + assert(missing \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), JString("parse_sql_ctx_b"), JString("t")))))) + } finally { + sql("USE parse_sql_ctx_a") + } + try { + val found = evalStatement("SELECT a FROM t") + assert(found \ "analyze_success" === JBool(true)) + assert(found \ "source_table_references" === JArray(List(JArray(List( + JString("spark_catalog"), JString("parse_sql_ctx_a"), JString("t")))))) + assert(spark.sql("SELECT current_schema()").head().getString(0) === + "parse_sql_ctx_a") + } finally { + sql("USE default") + } + } finally { + sql("DROP TABLE IF EXISTS parse_sql_ctx_a.t") + sql("DROP DATABASE IF EXISTS parse_sql_ctx_a CASCADE") + sql("DROP DATABASE IF EXISTS parse_sql_ctx_b CASCADE") + } + } + } + + test("ANALYZE named mode argument") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val json = spark.sql( + """SELECT parse_sql( + | 'SELECT 1 AS a', + | mode => 'ANALYZE')""".stripMargin).head().getString(0) + val analyzed = parse(json) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + assert(analyzed \ "analyze_success" === JBool(true)) + assert(analyzed \ "select_list" === JArray(List( + JObject("name" -> JArray(List(JString("a"))), "type" -> jsonInt)))) + } + } + + test("ANALYZE without parameter marker types leaves the query unresolved") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val json = spark.sql("SELECT parse_sql('SELECT abs(?) AS x', 'ANALYZE')") + .head().getString(0) + val analyzed = parse(json) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + assert(analyzed \ "parse_success" === JBool(true)) + assert(analyzed \ "analyze_success" === JBool(false)) + } + } + + private def evalBatch(sql: String): List[JObject] = + evalJson("'" + sql.replace("'", "''") + "'") match { + case JArray(values) => values.map(_.asInstanceOf[JObject]) + case other => fail(s"expected a JSON array, got: $other") + } + + private def typedColumns(columns: (String, JObject)*): JArray = JArray(columns.map { + case (name, dataType) => JObject("name" -> JArray(List(JString(name))), "type" -> dataType) + }.toList) + + private def ref(parts: String*): JArray = JArray(parts.map(JString).toList) + + /** Run a full `SELECT parse_sql(...)` and return its statement objects. */ + private def evalCall(query: String): List[JObject] = + parse(spark.sql(query).head().getString(0)) match { + case JArray(values) => values.map(_.asInstanceOf[JObject]) + case other => fail(s"expected a JSON array, got: $other") + } + + private def errorClass(statement: JObject): JValue = statement \ "error" \ "errorClass" + + test("ANALYZE does not run CALL, directly or through EXECUTE IMMEDIATE") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + spark.conf.set("spark.sql.catalog.parse_sql_cat", classOf[InMemoryCatalog].getName) + try { + val catalog = spark.sessionState.catalogManager.catalog("parse_sql_cat") + .asInstanceOf[InMemoryCatalog] + catalog.createProcedure( + Identifier.of(Array("ns"), "record"), ParseSqlAnalyzeSuite.UnboundRecord) + ParseSqlAnalyzeSuite.calls.set(0) + val results = evalBatch( + """CALL parse_sql_cat.ns.record(1); + |EXECUTE IMMEDIATE 'CALL parse_sql_cat.ns.record(2)'""".stripMargin) + assert(results.map(_ \ "parse_success") === List(JBool(true), JBool(true))) + assert(results.map(_ \ "analyze_success") === List(JBool(true), JBool(true))) + assert(ParseSqlAnalyzeSuite.calls.get === 0) + sql("CALL parse_sql_cat.ns.record(3)").collect() + assert(ParseSqlAnalyzeSuite.calls.get === 1) + } finally { + spark.sessionState.catalogManager.reset() + spark.sessionState.conf.unsetConf("spark.sql.catalog.parse_sql_cat") + } + } + } + + test("ANALYZE inside a SQL script does not see the script's locals") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val json = spark.sql( + """BEGIN + | DECLARE parse_sql_outer INT DEFAULT 5; + | SELECT parse_sql('SELECT parse_sql_outer AS x', 'ANALYZE'); + |END""".stripMargin).head().getString(0) + val statement = parse(json) match { + case JArray(List(value: JObject)) => value + case other => fail(s"expected one statement object, got: $other") + } + assert(statement \ "analyze_success" === JBool(false)) + val cls = (statement \ "error" \ "errorClass").asInstanceOf[JString].s + assert(cls === "UNRESOLVED_COLUMN.WITHOUT_SUGGESTION" || + cls === "UNRESOLVED_COLUMN.WITH_SUGGESTION" || + cls.startsWith("UNRESOLVED_COLUMN")) + } + } + + test("ANALYZE requires constant arguments; PARSE does not") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val query = "SELECT parse_sql(c, '%s') FROM VALUES ('SELECT 1') AS t(c)" + val e = intercept[AnalysisException](spark.sql(query.format("ANALYZE")).collect()) + assert(e.getCondition === "DATATYPE_MISMATCH.NON_FOLDABLE_INPUT") + assert(spark.sql(query.format("PARSE")).collect().length === 1) + val modeCol = intercept[AnalysisException]( + spark.sql("SELECT parse_sql(c, m) FROM VALUES ('SELECT 1', 'PARSE') AS t(c, m)") + .collect()) + assert(modeCol.getCondition === "DATATYPE_MISMATCH.NON_FOLDABLE_INPUT") + } + } + + test("ANALYZE does not execute SQL scripts") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_script_t", "parse_sql_script_u") { + sql("CREATE TABLE parse_sql_script_t (a INT) USING parquet") + sql("CREATE TABLE parse_sql_script_u (a INT) USING parquet") + val schema = spark.sql("SELECT current_schema()").head().getString(0) + val j = evalStatement( + """BEGIN + | INSERT INTO parse_sql_script_t VALUES (1); + | DROP TABLE parse_sql_script_u; + |END""".stripMargin) + assert(j \ "parse_success" === JBool(true)) + assert(j \ "analyze_success" === JBool(false)) + assert(errorClass(j) === JString("PARSE_SQL_ANALYZE_UNSUPPORTED_SCRIPT")) + assert(spark.table("parse_sql_script_t").count() === 0) + assert(spark.catalog.tableExists("parse_sql_script_u")) + assert(spark.sql("SELECT current_schema()").head().getString(0) === schema) + } + } + } + + test("ANALYZE reports range as system.builtin and de-duplicates count") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_dedup_t") { + sql("CREATE TABLE parse_sql_dedup_t (a INT, b INT) USING parquet") + val rangeStmt = evalStatement("SELECT * FROM range(3)") + assert(rangeStmt \ "analyze_success" === JBool(true)) + assert(rangeStmt \ "function_references" === JArray(List( + ref("system", "builtin", "range")))) + val counts = evalStatement( + "SELECT count(a), COUNT(b) FROM parse_sql_dedup_t") + assert(counts \ "analyze_success" === JBool(true)) + assert(counts \ "function_references" === JArray(List( + ref("system", "builtin", "count")))) + } + } + } + + test("ANALYZE qualifies DDL targets by statement kind") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val tempView = evalStatement("CREATE TEMPORARY VIEW parse_sql_ddl_v AS SELECT 1 AS a") + assert(tempView \ "analyze_success" === JBool(true)) + assert(tempView \ "target_table_references" === JArray(List( + ref("system", "session", "parse_sql_ddl_v")))) + assert(!spark.catalog.tableExists("parse_sql_ddl_v")) + withTempView("parse_sql_ddl_t") { + sql("CREATE TEMP VIEW parse_sql_ddl_t AS SELECT 1 AS a") + val table = evalStatement("CREATE TABLE parse_sql_ddl_t (a INT) USING parquet") + assert(table \ "analyze_success" === JBool(true)) + assert(table \ "target_table_references" === JArray(List( + ref("spark_catalog", "default", "parse_sql_ddl_t")))) + } + } + } + + test("ANALYZE types session-catalog CTAS query columns") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val j = evalStatement( + "CREATE TABLE parse_sql_ctas_t USING parquet AS SELECT 1 AS a") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "select_list" === typedColumns("a" -> jsonInt)) + assert(!spark.catalog.tableExists("parse_sql_ctas_t")) + } + } + + test("ANALYZE does not execute a mixed DDL/DML batch") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTable("parse_sql_batch_t") { + sql("CREATE TABLE parse_sql_batch_t (a INT) USING parquet") + val results = evalBatch( + """INSERT INTO parse_sql_batch_t VALUES (1); + |CREATE TABLE parse_sql_batch_u (a INT) USING parquet; + |DROP TABLE parse_sql_batch_t""".stripMargin) + assert(results.map(_ \ "analyze_success") === + List(JBool(true), JBool(true), JBool(true))) + assert(spark.table("parse_sql_batch_t").count() === 0) + assert(!spark.catalog.tableExists("parse_sql_batch_u")) + assert(spark.catalog.tableExists("parse_sql_batch_t")) + } + } + } + + test("ANALYZE qualifies global temp views and DROP of local temp views") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withGlobalTempView("parse_sql_global_v") { + sql("CREATE OR REPLACE GLOBAL TEMP VIEW parse_sql_global_v AS SELECT 1 AS a") + val globalDb = spark.conf.get("spark.sql.globalTempDatabase") + val j = evalStatement(s"SELECT a FROM $globalDb.parse_sql_global_v") + assert(j \ "analyze_success" === JBool(true)) + assert(j \ "source_table_references" === JArray(List( + ref("spark_catalog", globalDb, "parse_sql_global_v")))) + } + withTempView("parse_sql_drop_v") { + sql("CREATE TEMP VIEW parse_sql_drop_v AS SELECT 1 AS a") + val drop = evalStatement("DROP VIEW parse_sql_drop_v") + assert(drop \ "analyze_success" === JBool(true)) + assert(drop \ "target_table_references" === JArray(List( + ref("system", "session", "parse_sql_drop_v")))) + assert(spark.catalog.tableExists("parse_sql_drop_v")) + } + } + } + + test("ANALYZE types EXPLAIN and DESCRIBE QUERY inner plans") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + val explain = evalStatement("EXPLAIN SELECT 1 AS a") + assert(explain \ "analyze_success" === JBool(true)) + assert(explain \ "select_list" === typedColumns("a" -> jsonInt)) + val describe = evalStatement("DESCRIBE QUERY SELECT 1 AS a") + assert(describe \ "analyze_success" === JBool(true)) + assert(describe \ "select_list" === typedColumns("a" -> jsonInt)) + } + } + + test("ANALYZE reports path-based datasource tables from the analyzed plan") { + withSQLConf(SQLConf.PARSE_SQL_ENABLED.key -> "true") { + withTempDir { dir => + val path = dir.getCanonicalPath + spark.range(1).write.mode("overwrite").parquet(path) + val j = evalStatement(s"SELECT id FROM parquet.`$path`") + assert(j \ "analyze_success" === JBool(true)) + val sources = (j \ "source_table_references").asInstanceOf[JArray].arr + assert(sources.nonEmpty) + val parts = sources.head.asInstanceOf[JArray].arr.map(_.asInstanceOf[JString].s) + assert(parts.exists(_.contains(path)) || parts == Seq("parquet", path)) Review Comment: Fixed in e19b9d04cf1. The path-based test now asserts the full `source_table_references` array is `[format, path]`. ########## sql/core/src/main/scala/org/apache/spark/sql/catalyst/parser/ParseSqlResult.scala: ########## @@ -489,19 +660,183 @@ object ParseSqlResult { named.map(selectListItem) } + /** Select-list items with SQL types from an analyzed query's output. */ + private def collectTypedSelectList( + analyzed: LogicalPlan, + parsed: LogicalPlan): Seq[JObject] = { + val query = primaryQueryPlan(analyzed) + if (query.resolved && query.output.nonEmpty) { + query.output.map { attr => + JObject( + "name" -> partsToJArray(Seq(attr.name)), + "type" -> DescribeRelationJsonCommand.jsonType( + CharVarcharUtils.getRawType(attr.metadata).getOrElse(attr.dataType))) + } + } else { + collectSelectList(parsed) + } + } + + private def uniqueIdentities( + ids: Seq[Seq[String]], + session: ClassicSparkSession): Seq[Seq[String]] = { + val sensitive = session.sessionState.conf.caseSensitiveAnalysis + val seen = mutable.LinkedHashSet.empty[Seq[String]] + ids.filter { parts => + val key = if (sensitive) parts else parts.map(_.toLowerCase(Locale.ROOT)) + seen.add(key) + } + } + + private def sessionIdent(name: String): Seq[String] = + Seq(CatalogManager.SYSTEM_CATALOG_NAME, CatalogManager.SESSION_NAMESPACE, name) + + private def catalogIdentParts(catalog: String, ident: Identifier): Seq[String] = + Seq(catalog) ++ ident.namespace() :+ ident.name() + + private def v2RelationParts(r: DataSourceV2Relation): Option[Seq[String]] = + (r.catalog, r.identifier) match { + case (Some(catalog), Some(ident)) => Some(catalogIdentParts(catalog.name(), ident)) + case _ => None + } + + private def logicalRelationParts(r: LogicalRelation): Option[Seq[String]] = + r.catalogTable.map { t => + val id = t.identifier + if (id.catalog.isEmpty && id.database.isDefined) { + CatalogManager.SESSION_CATALOG_NAME +: tableIdentifierParts(id) + } else { + tableIdentifierParts(id) + } + }.orElse { + r.relation match { + case fs: HadoopFsRelation => + fs.location.rootPaths.headOption.map(p => Seq(p.toString)) Review Comment: Fixed in e19b9d04cf1. Path-based relations keep `[format, path]` as written; the analyzed plan is not used to replace that shape. Documented in the note. ########## sql/core/src/main/scala/org/apache/spark/sql/catalyst/parser/ParseSqlResult.scala: ########## @@ -65,148 +114,215 @@ object ParseSqlResult { ThreadLocal.withInitial(() => new SparkSqlParser()) /** Parse `sql` as a batch and render the JSON result array. */ - def fromSql(sql: String): String = { + def fromSql( + sql: String, + mode: String = ParseSqlMode.Parse): String = { val split = parser.get().splitStatementsWithPositions(sql) val statements = split.completeStatements ++ split.partialStatement - val starts = charStarts(sql, statements) - compact(render(JArray(statements.zip(starts).map { - case (segment, start) => parseStatement(segment, start) - }.toList))) + val spans = statementSpans(sql, statements) + val json = if (mode == ParseSqlMode.Analyze) { + analyzeBatch(statements, spans) + } else { + JArray(statements.zip(spans).map { case (segment, span) => + parsePlanOrError(segment, span) match { + case ParseFailed(errorObj) => errorObj + case Parsed(_, plan) => + successJson(plan, analyzed = None, analyzeError = None, None, span) + } + }.toList) + } + compact(render(json)) + } + + private sealed trait ParsedStatement + private case class Parsed(segment: PositionedSqlStatement, plan: LogicalPlan) + extends ParsedStatement + private case class ParseFailed(errorObj: JObject) extends ParsedStatement + + /** + * Parse every statement, then analyze each against the invoker session. + * Statements are not executed, so one does not see the effect of an earlier + * one. Analysis failures stay JSON (`analyze_success` false) per statement. + */ + private def analyzeBatch( + statements: Seq[PositionedSqlStatement], + spans: Seq[StatementSpan]): JArray = { + val parsed = statements.zip(spans).map { case (segment, span) => + parsePlanOrError(segment, span) + } + val session = ClassicSparkSession.getActiveSession + .getOrElse(throw QueryExecutionErrors.parseSqlAnalyzeRequiresSession()) + // The batch is analyzed on its own: a SQL script that calls parse_sql does + // not lend it its locals, and CALL is not run. + SqlScriptingContextManager.create(null).runWith { + InvokeProcedures.withoutInvocation { + JArray(parsed.zip(spans).map { + case (ParseFailed(errorObj), _) => errorObj + case (Parsed(_, plan), span) => + val (analyzed, error) = analyze(session, plan) + successJson(plan, analyzed, error, Some(session), span) + }.toList) + } + } } - private def parseStatement(segment: PositionedSqlStatement, charStart: Int): JObject = { + /** INTERNAL_ERROR is a Spark bug, not an analysis result: it fails the call. */ + private def isInternalError(e: SparkThrowable): Boolean = + Option(e.getCondition).exists(_.startsWith("INTERNAL_ERROR")) + + /** Character span of a statement object in the original batch. */ + private case class StatementSpan(start: Int, length: Int) + + /** + * Code-point `start` for each statement in one forward pass over `batch`. + * Statements are in source order, so prefix counts are not repeated. + */ + private def statementSpans( + batch: String, + statements: Seq[PositionedSqlStatement]): Seq[StatementSpan] = { + var lastUtf16 = 0 + var lastCp = 0 + statements.map { segment => + val utf16 = math.max(0, math.min(segment.start, batch.length)) + if (utf16 >= lastUtf16) { + lastCp += batch.codePointCount(lastUtf16, utf16) + lastUtf16 = utf16 + } else { + lastCp = batch.codePointCount(0, utf16) + lastUtf16 = utf16 + } + StatementSpan(lastCp + 1, charLength(segment)) + } + } + + /** Parsed plan, or parse-failure JSON for this statement. */ + private def parsePlanOrError( + segment: PositionedSqlStatement, + span: StatementSpan): ParsedStatement = { val sql = segment.statement - val length = sql.codePointCount(0, sql.length) try { // Do not inherit the outer query's origin from the parse_sql expression. - // Errors and parsed nodes refer to the individual statement. Parser - // origins are code-point offsets, so this one is too. + // Errors and parsed nodes refer to the individual statement. val origin = if (sql.nonEmpty) { Origin( startIndex = Some(0), - stopIndex = Some(length - 1), + stopIndex = Some(sql.codePointCount(0, sql.length) - 1), sqlText = Some(sql)) } else { Origin(sqlText = Some(sql)) } CurrentOrigin.withOrigin(origin) { - val plan = parser.get().parsePlan(sql) - successJson(plan, charStart, length) + Parsed(segment, parser.get().parsePlan(sql)) } } catch { - // User-facing parse / scripting failures become JSON; everything else fails. case e: ParseException => - errorJson(e, segment, charStart, length) + ParseFailed(errorJson(e, span)) case e: SqlScriptingException => - errorJson(e, segment, charStart, length) + ParseFailed(errorJson(e, span)) + } + } + + /** Analyze `plan` without executing it. Scripts never reach QueryExecution. */ + private def analyze( + session: ClassicSparkSession, + plan: LogicalPlan): (Option[LogicalPlan], Option[SparkThrowable with Throwable]) = { + plan match { + case _: CompoundBody => + (None, Some(QueryExecutionErrors.parseSqlAnalyzeUnsupportedScript())) + case _ => + try { + val toAnalyze = plan match { + case s: SupervisingCommand => Review Comment: Fixed in e19b9d04cf1. Supervising commands are transformed recursively, so `EXPLAIN EXPLAIN SELECT nope FROM missing_t` now fails analysis. Covered by the EXPLAIN test. ########## sql/core/src/main/scala/org/apache/spark/sql/catalyst/parser/ParseSqlResult.scala: ########## @@ -452,11 +582,52 @@ object ParseSqlResult { case u: UnresolvedView => add(u.multipartIdentifier) case u: UnresolvedTableOrView => add(u.multipartIdentifier) case u: UnresolvedIdentifier if role == TableRefRole.Target => add(u.nameParts) - case c: CreateViewCommand => addTarget(tableIdentifierParts(c.name)) - case c: CreateTempViewUsing => addTarget(tableIdentifierParts(c.tableIdent)) + case c: CreateViewCommand => + addTarget(c.viewType match { + case LocalTempView => sessionIdent(c.name.table) + case GlobalTempView => + Seq( + CatalogManager.SESSION_CATALOG_NAME, + SQLConf.get.getConf(StaticSQLConf.GLOBAL_TEMP_DATABASE), + c.name.table) + case _ => tableIdentifierParts(c.name) + }) + case c: CreateTempViewUsing => addTarget(sessionIdent(c.tableIdent.table)) + case c: CreateDataSourceTableCommand => + addTarget(persistedTableParts(c.table.identifier)) + case c: CreateTableCommand => + addTarget(persistedTableParts(c.table.identifier)) + case c: DropTableCommand => + addTarget( + if (c.isView && c.tableName.database.isEmpty) { + sessionIdent(c.tableName.table) + } else { + persistedTableParts(c.tableName) + }) + case c: DropTempViewCommand => + addTarget(tempViewParts(c.ident)) case c: CacheTable if c.multipartIdentifier.nonEmpty => addTarget(c.multipartIdentifier) case u: UnresolvedTableValuedFunction => addFunction(u.name) + case _: CTERelationRef => + case r: DataSourceV2Relation => + v2RelationParts(r).foreach(add) + case r: LogicalRelation => + logicalRelationParts(r).foreach(add) + case h: HiveTableRelation => + add(tableIdentifierParts(h.tableMeta.identifier)) + case v: View => + add(viewParts(v)) + case r: ResolvedTable => + add(catalogIdentParts(r.catalog.name(), r.identifier)) + case r: ResolvedPersistentView => + add(catalogIdentParts(r.catalog.name(), r.identifier)) + case r: ResolvedTempView => + add(tempViewParts(r.identifier)) + case r: ResolvedIdentifier if role == TableRefRole.Target => + add(catalogIdentParts(r.catalog.name(), r.identifier)) + case _: Range => Review Comment: Cleaned up in e19b9d04cf1: dropped the dead `Range` / `CTERelationRef` arms, unused `ParseSql.parameterNames`, the duplicate CTAS/INSERT `primaryQueryPlan` arms, and set `nullIntolerant` to true. `eval` still validates `mode` because `EvaluateParseSql` goes through it. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
