Pavel Pereslegin created IGNITE-26541:
-----------------------------------------
Summary: Java Thin 3.0. Schema compatibility. Improve unknown
column type handling.
Key: IGNITE-26541
URL: https://issues.apache.org/jira/browse/IGNITE-26541
Project: Ignite
Issue Type: Improvement
Components: thin clients ai3
Reporter: Pavel Pereslegin
If you add a new column type, the thin client cannot work with such a table
through table views.
{code:java}
// Schema has the following columns
// ID INT | VAL UNKNOWN
RecordView<Tuple> view = client.tables()
.table("TEST").recordView();
Tuple key = Tuple.create().set("id", 1);
// Attempt to find a row by key produces exception
view.get(null, key);
{code}
Exception:
{noformat}
[2025-09-28T10:15:41,292][ERROR][main][TcpClientChannel] Failed to deserialize
server response [remoteAddress=127.0.0.1:10800]: Invalid column type id: 14
org.apache.ignite.lang.IgniteException: Invalid column type id: 14
at
org.apache.ignite.internal.client.proto.ColumnTypeConverter.fromIdOrThrow(ColumnTypeConverter.java:39)
at
org.apache.ignite.internal.client.table.ClientTable.readSchema(ClientTable.java:272)
at
org.apache.ignite.internal.client.table.ClientTable.lambda$loadSchema$1(ClientTable.java:249)
at
org.apache.ignite.internal.client.TcpClientChannel.complete(TcpClientChannel.java:474)
at
org.apache.ignite.internal.client.TcpClientChannel.send(TcpClientChannel.java:409)
at
org.apache.ignite.internal.client.TcpClientChannel.serviceAsync(TcpClientChannel.java:335)
at
org.apache.ignite.internal.client.ReliableChannel.serviceAsyncInternal(ReliableChannel.java:360)
at
org.apache.ignite.internal.client.ReliableChannel.lambda$serviceAsync$1(ReliableChannel.java:268)
at
java.base/java.util.concurrent.CompletableFuture.uniComposeStage(CompletableFuture.java:1106)
at
java.base/java.util.concurrent.CompletableFuture.thenCompose(CompletableFuture.java:2237)
at
org.apache.ignite.internal.client.ReliableChannel.lambda$serviceAsync$2(ReliableChannel.java:268)
at
org.apache.ignite.internal.client.ClientFutureUtils.doWithRetryAsync(ClientFutureUtils.java:56)
at
org.apache.ignite.internal.client.ClientFutureUtils.doWithRetryAsync(ClientFutureUtils.java:45)
at
org.apache.ignite.internal.client.ReliableChannel.serviceAsync(ReliableChannel.java:266)
at
org.apache.ignite.internal.client.ReliableChannel.serviceAsync(ReliableChannel.java:339)
at
org.apache.ignite.internal.client.table.ClientTable.loadSchema(ClientTable.java:227)
at
java.base/java.util.concurrent.ConcurrentHashMap.computeIfAbsent(ConcurrentHashMap.java:1705)
at
org.apache.ignite.internal.client.table.ClientTable.getSchema(ClientTable.java:215)
at
org.apache.ignite.internal.client.table.ClientTable.doSchemaOutInOpAsync(ClientTable.java:456)
at
org.apache.ignite.internal.client.table.ClientTable.doSchemaOutInOpAsync(ClientTable.java:423)
at
org.apache.ignite.internal.client.table.ClientRecordBinaryView.getAsync(ClientRecordBinaryView.java:80)
at
org.apache.ignite.internal.client.table.ClientRecordBinaryView.get(ClientRecordBinaryView.java:72)
at
org.apache.ignite.internal.client.table.ClientRecordBinaryView.get(ClientRecordBinaryView.java:53)
at
org.apache.ignite.internal.client.table.api.PublicApiClientRecordView.lambda$get$0(PublicApiClientRecordView.java:52)
at
org.apache.ignite.internal.thread.PublicApiThreading.executeWithRole(PublicApiThreading.java:144)
at
org.apache.ignite.internal.thread.PublicApiThreading.execUserSyncOperation(PublicApiThreading.java:102)
at
org.apache.ignite.internal.client.table.api.PublicApiClientViewBase.executeSyncOp(PublicApiClientViewBase.java:101)
at
org.apache.ignite.internal.client.table.api.PublicApiClientRecordView.get(PublicApiClientRecordView.java:52)
{noformat}
We need to improve the readSchema method and map unknown column types to
something like ColumnType.UNKNOWN (id=0?).
This way, we'll be able to work with the table until we try to access an
unknown column.
When accessing a column of an unknown type, an exception with user-friendly
text should be thrown.
Currently, DML queries can be executed via SQL (unless you specify a column of
an unknown type).
However, this behavior needs to be improved to allow querying all columns
(using '*'), and the error message seems to be improved.
For example:
{code:sql}
SELECT id FROM TEST; -- ok
SELECT * FROM TEST; -- fails
{code}
{noformat}
[2025-09-28T12:45:02,056][ERROR][ForkJoinPool.commonPool-worker-19][TcpClientChannel]
Failed to deserialize server response [remoteAddress=127.0.0.1:10800]: Invalid
column type id: 14
org.apache.ignite.lang.IgniteException: Invalid column type id: 14
at
org.apache.ignite.internal.client.proto.ColumnTypeConverter.fromIdOrThrow(ColumnTypeConverter.java:39)
at
org.apache.ignite.internal.client.sql.ClientResultSetMetadata.readColumn(ClientResultSetMetadata.java:60)
at
org.apache.ignite.internal.client.sql.ClientResultSetMetadata.read(ClientResultSetMetadata.java:47)
at
org.apache.ignite.internal.client.sql.ClientAsyncResultSet.<init>(ClientAsyncResultSet.java:112)
at
org.apache.ignite.internal.client.sql.ClientSql.lambda$payloadReader$5(ClientSql.java:345)
{noformat}
p.s. It is necessary to check the behavior of other thin clients (C#, C++) and
improve it, if necessary, separately.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
