#75 - Polishing.

Clarified code a little.

Added a warning to `AnonymousBindMarkers`:

 Anonymous bind markers are problematic because the have to appear in generated SQL in the same order they get generated.
 This might cause challenges in the future with complex generate statements.
 For example those containing subselects which limit the freedom of arranging bind markers.

Original pull request: #84.

src/test/java/org/springframework/data/r2dbc/function/AbstractTransactionalDatabaseClientIntegrationTests.java

@@ -209,13 +209,19 @@ public abstract class AbstractTransactionalDatabaseClientIntegrationTests extend
 
 		Flux<Object> transactionIds = databaseClient.inTransaction(db -> {
 
+			// We have to execute a sql statement first.
+			// Otherwise some databases (MySql) don't have a transaction id.
 			Mono<Integer> insert = db.execute().sql(getInsertIntoLegosetStatement()) //
 					.bind(0, 42055) //
 					.bind(1, "SCHAUFELRADBAGGER") //
 					.bindNull(2, Integer.class) //
 					.fetch().rowsUpdated();
 
-			Flux<Object> txId = db.execute().sql(getCurrentTransactionIdStatement()).map((r, md) -> r.get(0)).all();
+			Flux<Object> txId = db.execute() //
+					.sql(getCurrentTransactionIdStatement()) //
+					.map((row, md) -> row.get(0)) //
+					.all();
+
 			return insert.thenMany(txId.concatWith(txId));
 		});
 
@@ -223,7 +229,7 @@ public abstract class AbstractTransactionalDatabaseClientIntegrationTests extend
 				.consumeNextWith(actual -> {
 
 					assertThat(actual).hasSize(2);
-					assertThat(actual).containsExactly(actual.get(1), actual.get(0));
+					assertThat(actual.get(0)).isEqualTo(actual.get(1));
 				}) //
 				.verifyComplete();
 	}