Skip to content

H
h2database
提交 dfd877e2 authored 作者: Noel Grandin's avatar Noel Grandin
浏览文件
操作

When in cluster mode, and one of the nodes goes down, we need to log the 

problem with priority "error", not "debug"
上级 555af2d5
隐藏空白字符变更
内嵌 并排
正在显示 包含 1295 行增加 和 1293 行删除
h2/src/docsrc/html/changelog.html
浏览文件 @ dfd877e2
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<!--
Copyright 2004-2014 H2 Group. Multiple-Licensed under the MPL 2.0, Version 1.0,
and under the Eclipse Public License, Version 1.0
Initial Developer: H2 Group
-->
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">
<head><meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<title>
Change Log
</title>
<link rel="stylesheet" type="text/css" href="stylesheet.css" />
<!-- [search] { -->
<script type="text/javascript" src="navigation.js"></script>
</head><body onload="frameMe();">
<table class="content"><tr class="content"><td class="content"><div class="contentDiv">
<!-- } -->
<h1>Change Log</h1>
<h2>Next Version (unreleased)</h2>
<ul><li>Pull request #4: Creating and removing temporary tables was getting
slower and slower over time, because an internal object id was allocated but
never de-allocated.
</li><li>Issue 609: the spatial index did not support NULL with update and delete operations.
</li><li>Pull request #2: Add external metadata type support (table type "external")
</li><li>MS SQL Server: the CONVERT method did not work in views
and derrived tables.
</li><li>Java 8 compatibility for "regexp_replace".
</li></ul>
<h2>Version 1.4.187 Beta (2015-04-10)</h2>
<ul><li>MVStore: concurrent changes to the same row could result in
the exception "The transaction log might be corrupt for key ...".
This could only be reproduced with 3 or more threads.
</li><li>Results with CLOB or BLOB data are no longer reused.
</li><li>References to BLOB and CLOB objects now have a timeout.
The configuration setting is LOB_TIMEOUT (default 5 minutes).
This should avoid growing the database file if there are many queries that return BLOB or CLOB objects,
and the database is not closed for a longer time.
</li><li>MVStore: when committing a session that removed LOB values,
changes were flushed unnecessarily.
</li><li>Issue 610: possible integer overflow in WriteBuffer.grow().
</li><li>Issue 609: the spatial index did not support NULL (ClassCastException).
</li><li>MVStore: in some cases, CLOB/BLOB data blocks were removed
incorrectly when opening a database.
</li><li>MVStore: updates that affected many rows were were slow
in some cases if there was a secondary index.
</li><li>Using "runscript" with autocommit disabled could result
in a lock timeout on the internal table "SYS".
</li><li>Issue 603: there was a memory leak when using H2 in a web application.
Apache Tomcat logged an error message: "The web application ...
created a ThreadLocal with key of type [org.h2.util.DateTimeUtils$1]".
</li><li>When using the MVStore,
running a SQL script generate by the Recover tool from a PageStore file
failed with a strange error message (NullPointerException),
now a clear error message is shown.
</li><li>Issue 605: with version 1.4.186, opening a database could result in
an endless loop in LobStorageMap.init.
</li><li>Queries that use the same table alias multiple times now work.
Before, the select expression list was expanded incorrectly.
Example: "select * from a as x, b as x".
</li><li>The MySQL compatibility feature "insert ... on duplicate key update"
did not work with a non-default schema.
</li><li>Issue 599: the condition "in(x, y)" could not be used in the select list
when using "group by".
</li><li>The LIRS cache could grow larger than the allocated memory.
</li><li>A new file system implementation that re-opens the file if it was closed due
to the application calling Thread.interrupt(). File name prefix "retry:".
Please note it is strongly recommended to avoid calling Thread.interrupt;
this is a problem for various libraries, including Apache Lucene.
</li><li>MVStore: use RandomAccessFile file system if the file name starts with "file:".
</li><li>Allow DATEADD to take a long value for count when manipulating milliseconds.
</li><li>When using MV_STORE=TRUE and the SET CACHE_SIZE setting, the cache size was incorrectly set,
so that it was effectively 1024 times smaller than it should be.
</li><li>Concurrent CREATE TABLE... IF NOT EXISTS in the presence of MULTI_THREAD=TRUE could
throw an exception.
</li><li>Fix bug in MVStore when creating lots of temporary tables, where we could run out of
transaction IDs.
</li><li>Add support for PostgreSQL STRING_AGG function. Patch by Fred Aquiles.
</li><li>Fix bug in "jdbc:h2:nioMemFS" isRoot() function.
Also, the page size was increased to 64 KB.
</li></ul>
<h2>Version 1.4.186 Beta (2015-03-02)</h2>
<ul><li>The Servlet API 3.0.1 is now used, instead of 2.4.
</li><li>MVStore: old chunks no longer removed in append-only mode.
</li><li>MVStore: the cache for page references could grow far too big, resulting in out of memory in some cases.
</li><li>MVStore: orphaned lob objects were not correctly removed in some cases,
making the database grow unnecessarily.
</li><li>MVStore: the maximum cache size was artificially limited to 2 GB
(due to an integer overflow).
</li><li>MVStore / TransactionStore: concurrent updates could result in a
"Too many open transactions" exception.
</li><li>StringUtils.toUpperEnglish now has a small cache.
This should speed up reading from a ResultSet when using the column name.
</li><li>MVStore: up to 65535 open transactions are now supported.
Previously, the limit was at most 65535 transactions between the oldest open and the
newest open transaction (which was quite a strange limit).
</li><li>The default limit for in-place LOB objects was changed from 128 to 256 bytes.
This is because each read creates a reference to a LOB, and maintaining the references
is a big overhead. With the higher limit, less references are needed.
</li><li>Tables without columns didn't work.
(The use case for such tables is testing.)
</li><li>The LIRS cache now resizes the table automatically in all cases
and no longer needs the averageMemory configuration.
</li><li>Creating a linked table from an MVStore database to a non-MVStore database
created a second (non-MVStore) database file.
</li><li>In version 1.4.184, a bug was introduced that broke queries
that have both joins and wildcards, for example:
select * from dual join(select x from dual) on 1=1
</li><li>Issue 598: parser fails on timestamp "24:00:00.1234" - prevent the creation of out-of-range time values.
</li><li>Allow declaring triggers as source code (like functions). Patch by Sylvain Cuaz.
</li><li>Make the planner use indexes for sorting when doing a GROUP BY where
all of the GROUP BY columns are not mentioned in the select. Patch by Frederico (zepfred).
</li><li>PostgreSQL compatibility: generate_series (as an alias for system_range). Patch by litailang.
</li><li>Fix missing "column" type in right-hand parameter in ConditionIn. Patch by Arnaud Thimel.
</li></ul>
<h2>Version 1.4.185 Beta (2015-01-16)</h2>
<ul><li>In version 1.4.184, "group by" ignored the table name,
and could pick a select column by mistake.
Example: select 0 as x from system_range(1, 2) d group by d.x;
</li><li>New connection setting "REUSE_SPACE" (default: true). If disabled,
all changes are appended to the database file, and existing content is never overwritten.
This allows to rollback to a previous state of the database by truncating
the database file.
</li><li>Issue 587: MVStore: concurrent compaction and store operations could result in an IllegalStateException.
</li><li>Issue 594: Profiler.copyInThread does not work properly.
</li><li>Script tool: Now, SCRIPT ... TO is always used (for higher speed and lower disk space usage).
</li><li>Script tool: Fix parsing of BLOCKSIZE parameter, original patch by Ken Jorissen.
</li><li>Fix bug in PageStore#commit method - when the ignoreBigLog flag was set,
the logic that cleared the flag could never be reached, resulting in performance degradation.
Reported by Alexander Nesterov.
</li><li>Issue 552: Implement BIT_AND and BIT_OR aggregate functions.
</li></ul>
<h2>Version 1.4.184 Beta (2014-12-19)</h2>
<ul><li>In version 1.3.183, indexes were not used if the table contains
columns with a default value generated by a sequence.
This includes tables with identity and auto-increment columns.
This bug was introduced by supporting "rownum" in views and derived tables.
</li><li>MVStore: imported BLOB and CLOB data sometimes disappeared.
This was caused by a bug in the ObjectDataType comparison.
</li><li>Reading from a StreamStore now throws an
IOException if the underlying data doesn't exist.
</li><li>MVStore: if there is an exception while saving, the store is now in all cases immediately closed.
</li><li>MVStore: the dump tool could go into an endless loop for some files.
</li><li>MVStore: recovery for a database with many CLOB or BLOB entries is now much faster.
</li><li>Group by with a quoted select column name alias didn't work. Example:
select 1 "a" from dual group by "a"
</li><li>Auto-server mode: the host name is now stored in the .lock.db file.
</li></ul>
<h2>Version 1.4.183 Beta (2014-12-13)</h2>
<ul><li>MVStore: the default auto-commit buffer size is now about twice as big.
This should reduce the database file size after inserting a lot of data.
</li><li>The built-in functions "power" and "radians" now always return a double.
</li><li>Using "row_number" or "rownum" in views or derived tables had unexpected results
if the outer query contained constraints for the given view. Example:
select b.nr, b.id from (select row_number() over() as nr, a.id as id
from (select id from test order by name) as a) as b where b.id = 1
</li><li>MVStore: the Recover tool can now deal with more types of corruption in the file.
</li><li>MVStore: the TransactionStore now first needs to be initialized before it can be used.
</li><li>Views and derived tables with equality and range conditions on the same columns
did not work properly. example: select x from (select x from (select 1 as x)
where x &gt; 0 and x &lt; 2) where x = 1
</li><li>The database URL setting PAGE_SIZE setting is now also used for the MVStore.
</li><li>MVStore: the default page split size for persistent stores is now 4096
(it was 16 KB so far). This should reduce the database file size for most situations
(in some cases, less than half the size of the previous version).
</li><li>With query literals disabled, auto-analyze of a table with CLOB or BLOB did not work.
</li><li>MVStore: use a mark and sweep GC algorithm instead of reference counting,
to ensure used chunks are never overwrite, even if the reference counting
algorithm does not work properly.
</li><li>In the multi-threaded mode, updating the column selectivity ("analyze")
in the background sometimes did not work.
</li><li>In the multi-threaded mode, database metadata operations
did sometimes not work if the schema was changed at the same time
(for example, if tables were dropped).
</li><li>Some CLOB and BLOB values could no longer be read when
the original row was removed (even when using the MVCC mode).
</li><li>The MVStoreTool could throw an IllegalArgumentException.
</li><li>Improved performance for some
date / time / timestamp conversion operations.
Thanks to Sergey Evdokimov for reporting the problem.
</li><li>H2 Console: the built-in web server did not work properly
if an unknown file was requested.
</li><li>MVStore: the jar file is renamed to "h2-mvstore-*.jar" and is
deployed to Maven separately.
</li><li>MVStore: support for concurrent reads and writes is now enabled by default.
</li><li>Server mode: the transfer buffer size has been changed from 16 KB to 64 KB,
after it was found that this improves performance on Linux quite a lot.
</li><li>H2 Console and server mode: SSL is now disabled and TLS is used
to protect against the Poodle SSLv3 vulnerability.
The system property to disable secure anonymous connections is now
"h2.enableAnonymousTLS".
The default certificate is still self-signed, so you need to manually install
another one if you want to avoid man in the middle attacks.
</li><li>MVStore: the R-tree did not correctly measure the memory usage.
</li><li>MVStore: compacting a store with an R-tree did not always work.
</li><li>Issue 581: When running in LOCK_MODE=0,
JdbcDatabaseMetaData#supportsTransactionIsolationLevel(TRANSACTION_READ_UNCOMMITTED)
should return false
</li><li>Fix bug which could generate deadlocks when multiple connections accessed the same table.
</li><li>Some places in the code were not respecting the value set in the "SET MAX_MEMORY_ROWS x" command
</li><li>Fix bug which could generate a NegativeArraySizeException when performing large (>40M) row union operations
</li><li>Fix "USE schema" command for MySQL compatibility, patch by mfulton
</li><li>Parse and ignore the ROW_FORMAT=DYNAMIC MySQL syntax, patch by mfulton
</li></ul>
<h2>Version 1.4.182 Beta (2014-10-17)</h2>
<ul><li>MVStore: improved error messages and logging;
improved behavior if there is an error when serializing objects.
</li><li>OSGi: the MVStore packages are now exported.
</li><li>With the MVStore option, when using multiple threads
that concurrently create indexes or tables,
it was relatively easy to get a lock timeout on the "SYS" table.
</li><li>When using the multi-threaded option, the exception
"Unexpected code path" could be thrown, specially if the option
"analyze_auto" was set to a low value.
</li><li>In the server mode, when reading from a CLOB or BLOB, if the connection
was closed, a NullPointerException could be thrown instead of an exception saying
the connection is closed.
</li><li>DatabaseMetaData.getProcedures and getProcedureColumns
could throw an exception if a user defined class is not available.
</li><li>Issue 584: the error message for a wrong sequence definition was wrong.
</li><li>CSV tool: the rowSeparator option is no longer supported,
as the same can be achieved with the lineSeparator.
</li><li>Descending indexes on MVStore tables did not work properly.
</li><li>Issue 579: Conditions on the "_rowid_" pseudo-column didn't use an index
when using the MVStore.
</li><li>Fixed documentation that "offset" and "fetch" are also keywords since version 1.4.x.
</li><li>The Long.MIN_VALUE could not be parsed for auto-increment (identity) columns.
</li><li>Issue 573: Add implementation for Methods "isWrapperFor()" and "unwrap()" in
other JDBC classes.
</li><li>Issue 572: MySQL compatibility for "order by" in update statements.
</li><li>The change in JDBC escape processing in version 1.4.181 affects both the parser
(which is running on the server) and the JDBC API (which is running on the client).
If you (or a tool you use) use the syntax "{t 'time}", or "{ts 'timestamp'}", or "{d 'data'}",
then both the client and the server need to be upgraded to version 1.4.181 or later.
</li></ul>
<h2>Version 1.4.181 Beta (2014-08-06)</h2>
<ul><li>Improved MySQL compatibility by supporting "use schema".
Thanks a lot to Karl Pietrzak for the patch!
</li><li>Writing to the trace file is now faster, specially with the debug level.
</li><li>The database option "defrag_always=true" did not work with the MVStore.
</li><li>The JDBC escape syntax {ts 'value'} did not interpret the value as a timestamp.
The same for {d 'value'} (for date) and {t 'value'} (for time).
Thanks to Lukas Eder for reporting the issue.
The following problem was detected after version 1.4.181 was released:
The change in JDBC escape processing affects both the parser (which is running on the server)
and the JDBC API (which is running on the client).
If you (or a tool you use) use the syntax {t 'time'}, or {ts 'timestamp'}, or {d 'date'},
then both the client and the server need to be upgraded to version 1.4.181 or later.
</li><li>File system abstraction: support replacing existing files using move
(currently not for Windows).
</li><li>The statement "shutdown defrag" now compresses the database (with the MVStore).
This command can greatly reduce the file size, and is relatively fast,
but is not incremental.
</li><li>The MVStore now automatically compacts the store in the background if there is no read or write activity,
which should (after some time; sometimes about one minute) reduce the file size.
This is still work in progress, feedback is welcome!
</li><li>Change default value of PAGE_SIZE from 2048 to 4096 to more closely match most file systems block size
(PageStore only; the MVStore already used 4096).
</li><li>Auto-scale MAX_MEMORY_ROWS and CACHE_SIZE settings by the amount of available RAM. Gives a better
out of box experience for people with more powerful machines.
</li><li>Handle tabs like 4 spaces in web console, patch by Martin Grajcar.
</li><li>Issue 573: Add implementation for Methods "isWrapperFor()" and "unwrap()" in JdbcConnection.java,
patch by BigMichi1.
</li></ul>
<h2>Version 1.4.180 Beta (2014-07-13)</h2>
<ul><li>MVStore: the store is now auto-compacted automatically up to some point,
to avoid very large file sizes. This area is still work in progress.
</li><li>Sequences of temporary tables (auto-increment or identity columns)
were persisted unnecessarily in the database file, and were not removed
when re-opening the database.
</li><li>MVStore: an IndexOutOfBoundsException could sometimes
occur MVMap.openVersion when concurrently accessing the store.
</li><li>The LIRS cache now re-sizes the internal hash map if needed.
</li><li>Optionally persist session history in the H2 console. (patch from Martin Grajcar)
</li><li>Add client-info property to get the number of servers currently in the cluster
and which servers that are available. (patch from Nikolaj Fogh)
</li><li>Fix bug in changing encrypted DB password that kept the file handle
open when the wrong password was supplied. (test case from Jens Hohmuth).
</li><li>Issue 567: H2 hangs for a long time then (sometimes) recovers.
Introduce a queue when doing table locking to prevent session starvation.
</li></ul>
<h2>Version 1.4.179 Beta (2014-06-23)</h2>
<ul><li>The license was changed to MPL 2.0 (from 1.0) and EPL 1.0.
</li><li>Issue 565: MVStore: concurrently adding LOB objects
(with MULTI_THREADED option) resulted in a NullPointerException.
</li><li>MVStore: reduced dependencies to other H2 classes.
</li><li>There was a way to prevent a database from being re-opened,
by creating a column constraint that references a table with a higher id,
for example with "check" constraints that contains queries.
This is now detected, and creating the table is prohibited.
In future versions of H2, most likely creating references to other
tables will no longer be supported because of such problems.
</li><li>MVStore: descending indexes with "nulls first" did not work as expected
(null was ordered last).
</li><li>Large result sets now always create temporary tables instead of temporary files.
</li><li>When using the PageStore, opening a database failed in some cases with a NullPointerException
if temporary tables were used (explicitly, or implicitly when using large result sets).
</li><li>If a database file in the PageStore file format exists, this file and this mode
is now used, even if the database URL does not contain "MV_STORE=FALSE".
If a MVStore file exists, it is used.
</li><li>Databases created with version 1.3.175 and earlier
that contained foreign keys in combination with multi-column indexes
could not be opened in some cases.
This was due to a bugfix in version 1.3.176:
Referential integrity constraints sometimes used the wrong index.
</li><li>MVStore: the ObjectDataType comparison method was incorrect if one
key was Serializable and the other was of a common class.
</li><li>Recursive queries with many result rows (more than the setting "max_memory_rows")
did not work correctly.
</li><li>The license has changed to MPL 2.0 + EPL 1.0.
</li><li>MVStore: temporary tables from result sets could survive re-opening a database,
which could result in a ClassCastException.
</li><li>Issue 566: MVStore: unique indexes that were created later on did not work correctly
if there were over 5000 rows in the table.
Existing databases need to be re-created (at least the broken index need to be re-built).
</li><li>MVStore: creating secondary indexes on large tables
results in missing rows in the index.
</li><li>Metadata: the password of linked tables is now only visible for admin users.
</li><li>For Windows, database URLs of the form "jdbc:h2:/test" where considered
relative and did not work unless the system property "h2.implicitRelativePath" was used.
</li><li>Windows: using a base directory of "C:/" and similar did not work as expected.
</li><li>Follow JDBC specification on Procedures MetaData, use P0 as
return type of procedure.
</li><li>Issue 531: IDENTITY ignored for added column.
</li><li>FileSystem: improve exception throwing compatibility with JDK
</li><li>Spatial Index: adjust costs so we do not use the spatial index if the
query does not contain an intersects operator.
</li><li>Fix multi-threaded deadlock when using a View that includes a TableFunction.
</li><li>Fix bug in dividing very-small BigDecimal numbers.
</li></ul>
<h2>Version 1.4.178 Beta (2014-05-02)</h2>
<ul><li>Issue 559: Make dependency on org.osgi.service.jdbc optional.
</li><li>Improve error message when the user specifies an unsupported combination of database settings.
</li><li>MVStore: in the multi-threaded mode, NullPointerException and other exceptions could occur.
</li><li>MVStore: some database file could not be compacted due to a bug in
the bookkeeping of the fill rate. Also, database file were compacted quite slowly.
This has been improved; but more changes in this area are expected.
</li><li>MVStore: support for volatile maps (that don't store changes).
</li><li>MVStore mode: in-memory databases now also use the MVStore.
</li><li>In server mode, appending ";autocommit=false" to the database URL was working,
but the return value of Connection.getAutoCommit() was wrong.
</li><li>Issue 561: OSGi: the import package declaration of org.h2 excluded version 1.4.
</li><li>Issue 558: with the MVStore, a NullPointerException could occur when using LOBs
at session commit (LobStorageMap.removeLob).
</li><li>Remove the "h2.MAX_MEMORY_ROWS_DISTINCT" system property to reduce confusion.
We already have the MAX_MEMORY_ROWS setting which does a very similar thing, and is better documented.
</li><li>Issue 554: Web Console in an IFrame was not fully supported.
</li></ul>
<h2>Version 1.4.177 Beta (2014-04-12)</h2>
<ul><li>By default, the MV_STORE option is enabled, so it is using the new MVStore
storage. The MVCC setting is by default set to the same values as the MV_STORE setting,
so it is also enabled by default. For testing, both settings can be disabled by appending
";MV_STORE=FALSE" and/or ";MVCC=FALSE" to the database URL.
</li><li>The file locking method 'serialized' is no longer supported.
This mode might return in a future version,
however this is not clear right now.
A new implementation and new tests would be needed.
</li><li>Enable the new storage format for dates (system property "h2.storeLocalTime").
For the MVStore mode, this is always enabled, but with version 1.4
this is even enabled in the PageStore mode.
</li><li>Implicit relative paths are disabled (system property "h2.implicitRelativePath"),
so that the database URL jdbc:h2:test now needs to be written as jdbc:h2:./test.
</li><li>"select ... fetch first 1 row only" is supported with the regular mode.
This was disabled so far because "fetch" and "offset" are now keywords.
See also Mode.supportOffsetFetch.
</li><li>Byte arrays are now sorted in unsigned mode
(x'99' is larger than x'09').
(System property "h2.sortBinaryUnsigned", Mode.binaryUnsigned, setting "binary_collation").
</li><li>Csv.getInstance will be removed in future versions of 1.4.
Use the public constructor instead.
</li><li>Remove support for the limited old-style outer join syntax using "(+)".
Use "outer join" instead.
System property "h2.oldStyleOuterJoin".
</li><li>Support the data type "DATETIME2" as an alias for "DATETIME", for MS SQL Server compatibility.
</li><li>Add Oracle-compatible TRANSLATE function, patch by Eric Chatellier.
</li></ul>
<h2>Version 1.3.176 (2014-04-05)</h2>
<ul><li>The file locking method 'serialized' is no longer documented,
as it will not be available in version 1.4.
</li><li>The static method Csv.getInstance() was removed.
Use the public constructor instead.
</li><li>The default user name for the Script, RunScript, Shell,
and CreateCluster tools are no longer "sa" but an empty string.
</li><li>The stack trace of the exception "The object is already closed" is no longer logged by default.
</li><li>If a value of a result set was itself a result set, the result
could only be read once.
</li><li>Column constraints are also visible in views (patch from Nicolas Fortin for H2GIS).
</li><li>Granting a additional right to a role that already had a right for that table was not working.
</li><li>Spatial index: a few bugs have been fixed (using spatial constraints in views,
transferring geometry objects over TCP/IP, the returned geometry object is copied when needed).
</li><li>Issue 551: the datatype documentation was incorrect (found by Bernd Eckenfels).
</li><li>Issue 368: ON DUPLICATE KEY UPDATE did not work for multi-row inserts.
Test case from Angus Macdonald.
</li><li>OSGi: the package javax.tools is now imported (as an optional).
</li><li>H2 Console: auto-complete is now disabled by default, but there is a hot-key (Ctrl+Space).
</li><li>H2 Console: auto-complete did not work with multi-line statements.
</li><li>CLOB and BLOB data was not immediately removed after a rollback.
</li><li>There is a new Aggregate API that supports the internal H2 data types
(GEOMETRY for example). Thanks a lot to Nicolas Fortin for the patch!
</li><li>Referential integrity constraints sometimes used the wrong index,
such that updating a row in the referenced table incorrectly failed with
a constraint violation.
</li><li>The Polish translation was completed and corrected by Wojtek Jurczyk. Thanks a lot!
</li><li>Issue 545: Unnecessary duplicate code was removed.
</li><li>The profiler tool can now process files with full thread dumps.
</li><li>MVStore: the file format was changed slightly.
</li><li>MVStore mode: the CLOB and BLOB storage was re-implemented and is
now much faster than with the PageStore (which is still the default storage).
</li><li>MVStore mode: creating indexes is now much faster
(in many cases faster than with the default PageStore).
</li><li>Various bugs in the MVStore storage and have been fixed,
including a bug in the R-tree implementation.
The database could get corrupt if there were transient IO exceptions while storing.
</li><li>The method org.h2.expression.Function.getCost could throw a NullPointException.
</li><li>Storing LOBs in separate files (outside of the main database file)
is no longer supported for new databases.
</li><li>Lucene 2 is no longer supported.
</li><li>Fix bug in calculating default MIN and MAX values for SEQUENCE.
</li><li>Fix bug in performing IN queries with multiple values when IGNORECASE=TRUE
</li><li>Add entry-point to org.h2.tools.Shell so it can be called from inside an application.
patch by Thomas Gillet.
</li><li>Fix bug that prevented the PgServer from being stopped and started multiple times.
</li><li>Support some more DDL syntax for MySQL, patch from Peter Jentsch.
</li><li>Issue 548: TO_CHAR does not format MM and DD correctly when the month or day of
the month is 1 digit, patch from "the.tucc"
</li><li>Fix bug in varargs support in ALIAS's, patch from Nicolas Fortin
</li></ul>
<!-- [close] { --></div></td></tr></table><!-- } --><!-- analytics --></body></html>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<!--
Copyright 2004-2014 H2 Group. Multiple-Licensed under the MPL 2.0, Version 1.0,
and under the Eclipse Public License, Version 1.0
Initial Developer: H2 Group
-->
<html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en">
<head><meta http-equiv="Content-Type" content="text/html;charset=utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<title>
Change Log
</title>
<link rel="stylesheet" type="text/css" href="stylesheet.css" />
<!-- [search] { -->
<script type="text/javascript" src="navigation.js"></script>
</head><body onload="frameMe();">
<table class="content"><tr class="content"><td class="content"><div class="contentDiv">
<!-- } -->
<h1>Change Log</h1>
<h2>Next Version (unreleased)</h2>
<ul><li>Pull request #4: Creating and removing temporary tables was getting
slower and slower over time, because an internal object id was allocated but
never de-allocated.
</li><li>Issue 609: the spatial index did not support NULL with update and delete operations.
</li><li>Pull request #2: Add external metadata type support (table type "external")
</li><li>MS SQL Server: the CONVERT method did not work in views
and derrived tables.
</li><li>Java 8 compatibility for "regexp_replace".
</li><li>When in cluster mode, and one of the nodes goes down,
we need to log the problem with priority "error", not "debug"
</li></ul>
<h2>Version 1.4.187 Beta (2015-04-10)</h2>
<ul><li>MVStore: concurrent changes to the same row could result in
the exception "The transaction log might be corrupt for key ...".
This could only be reproduced with 3 or more threads.
</li><li>Results with CLOB or BLOB data are no longer reused.
</li><li>References to BLOB and CLOB objects now have a timeout.
The configuration setting is LOB_TIMEOUT (default 5 minutes).
This should avoid growing the database file if there are many queries that return BLOB or CLOB objects,
and the database is not closed for a longer time.
</li><li>MVStore: when committing a session that removed LOB values,
changes were flushed unnecessarily.
</li><li>Issue 610: possible integer overflow in WriteBuffer.grow().
</li><li>Issue 609: the spatial index did not support NULL (ClassCastException).
</li><li>MVStore: in some cases, CLOB/BLOB data blocks were removed
incorrectly when opening a database.
</li><li>MVStore: updates that affected many rows were were slow
in some cases if there was a secondary index.
</li><li>Using "runscript" with autocommit disabled could result
in a lock timeout on the internal table "SYS".
</li><li>Issue 603: there was a memory leak when using H2 in a web application.
Apache Tomcat logged an error message: "The web application ...
created a ThreadLocal with key of type [org.h2.util.DateTimeUtils$1]".
</li><li>When using the MVStore,
running a SQL script generate by the Recover tool from a PageStore file
failed with a strange error message (NullPointerException),
now a clear error message is shown.
</li><li>Issue 605: with version 1.4.186, opening a database could result in
an endless loop in LobStorageMap.init.
</li><li>Queries that use the same table alias multiple times now work.
Before, the select expression list was expanded incorrectly.
Example: "select * from a as x, b as x".
</li><li>The MySQL compatibility feature "insert ... on duplicate key update"
did not work with a non-default schema.
</li><li>Issue 599: the condition "in(x, y)" could not be used in the select list
when using "group by".
</li><li>The LIRS cache could grow larger than the allocated memory.
</li><li>A new file system implementation that re-opens the file if it was closed due
to the application calling Thread.interrupt(). File name prefix "retry:".
Please note it is strongly recommended to avoid calling Thread.interrupt;
this is a problem for various libraries, including Apache Lucene.
</li><li>MVStore: use RandomAccessFile file system if the file name starts with "file:".
</li><li>Allow DATEADD to take a long value for count when manipulating milliseconds.
</li><li>When using MV_STORE=TRUE and the SET CACHE_SIZE setting, the cache size was incorrectly set,
so that it was effectively 1024 times smaller than it should be.
</li><li>Concurrent CREATE TABLE... IF NOT EXISTS in the presence of MULTI_THREAD=TRUE could
throw an exception.
</li><li>Fix bug in MVStore when creating lots of temporary tables, where we could run out of
transaction IDs.
</li><li>Add support for PostgreSQL STRING_AGG function. Patch by Fred Aquiles.
</li><li>Fix bug in "jdbc:h2:nioMemFS" isRoot() function.
Also, the page size was increased to 64 KB.
</li></ul>
<h2>Version 1.4.186 Beta (2015-03-02)</h2>
<ul><li>The Servlet API 3.0.1 is now used, instead of 2.4.
</li><li>MVStore: old chunks no longer removed in append-only mode.
</li><li>MVStore: the cache for page references could grow far too big, resulting in out of memory in some cases.
</li><li>MVStore: orphaned lob objects were not correctly removed in some cases,
making the database grow unnecessarily.
</li><li>MVStore: the maximum cache size was artificially limited to 2 GB
(due to an integer overflow).
</li><li>MVStore / TransactionStore: concurrent updates could result in a
"Too many open transactions" exception.
</li><li>StringUtils.toUpperEnglish now has a small cache.
This should speed up reading from a ResultSet when using the column name.
</li><li>MVStore: up to 65535 open transactions are now supported.
Previously, the limit was at most 65535 transactions between the oldest open and the
newest open transaction (which was quite a strange limit).
</li><li>The default limit for in-place LOB objects was changed from 128 to 256 bytes.
This is because each read creates a reference to a LOB, and maintaining the references
is a big overhead. With the higher limit, less references are needed.
</li><li>Tables without columns didn't work.
(The use case for such tables is testing.)
</li><li>The LIRS cache now resizes the table automatically in all cases
and no longer needs the averageMemory configuration.
</li><li>Creating a linked table from an MVStore database to a non-MVStore database
created a second (non-MVStore) database file.
</li><li>In version 1.4.184, a bug was introduced that broke queries
that have both joins and wildcards, for example:
select * from dual join(select x from dual) on 1=1
</li><li>Issue 598: parser fails on timestamp "24:00:00.1234" - prevent the creation of out-of-range time values.
</li><li>Allow declaring triggers as source code (like functions). Patch by Sylvain Cuaz.
</li><li>Make the planner use indexes for sorting when doing a GROUP BY where
all of the GROUP BY columns are not mentioned in the select. Patch by Frederico (zepfred).
</li><li>PostgreSQL compatibility: generate_series (as an alias for system_range). Patch by litailang.
</li><li>Fix missing "column" type in right-hand parameter in ConditionIn. Patch by Arnaud Thimel.
</li></ul>
<h2>Version 1.4.185 Beta (2015-01-16)</h2>
<ul><li>In version 1.4.184, "group by" ignored the table name,
and could pick a select column by mistake.
Example: select 0 as x from system_range(1, 2) d group by d.x;
</li><li>New connection setting "REUSE_SPACE" (default: true). If disabled,
all changes are appended to the database file, and existing content is never overwritten.
This allows to rollback to a previous state of the database by truncating
the database file.
</li><li>Issue 587: MVStore: concurrent compaction and store operations could result in an IllegalStateException.
</li><li>Issue 594: Profiler.copyInThread does not work properly.
</li><li>Script tool: Now, SCRIPT ... TO is always used (for higher speed and lower disk space usage).
</li><li>Script tool: Fix parsing of BLOCKSIZE parameter, original patch by Ken Jorissen.
</li><li>Fix bug in PageStore#commit method - when the ignoreBigLog flag was set,
the logic that cleared the flag could never be reached, resulting in performance degradation.
Reported by Alexander Nesterov.
</li><li>Issue 552: Implement BIT_AND and BIT_OR aggregate functions.
</li></ul>
<h2>Version 1.4.184 Beta (2014-12-19)</h2>
<ul><li>In version 1.3.183, indexes were not used if the table contains
columns with a default value generated by a sequence.
This includes tables with identity and auto-increment columns.
This bug was introduced by supporting "rownum" in views and derived tables.
</li><li>MVStore: imported BLOB and CLOB data sometimes disappeared.
This was caused by a bug in the ObjectDataType comparison.
</li><li>Reading from a StreamStore now throws an
IOException if the underlying data doesn't exist.
</li><li>MVStore: if there is an exception while saving, the store is now in all cases immediately closed.
</li><li>MVStore: the dump tool could go into an endless loop for some files.
</li><li>MVStore: recovery for a database with many CLOB or BLOB entries is now much faster.
</li><li>Group by with a quoted select column name alias didn't work. Example:
select 1 "a" from dual group by "a"
</li><li>Auto-server mode: the host name is now stored in the .lock.db file.
</li></ul>
<h2>Version 1.4.183 Beta (2014-12-13)</h2>
<ul><li>MVStore: the default auto-commit buffer size is now about twice as big.
This should reduce the database file size after inserting a lot of data.
</li><li>The built-in functions "power" and "radians" now always return a double.
</li><li>Using "row_number" or "rownum" in views or derived tables had unexpected results
if the outer query contained constraints for the given view. Example:
select b.nr, b.id from (select row_number() over() as nr, a.id as id
from (select id from test order by name) as a) as b where b.id = 1
</li><li>MVStore: the Recover tool can now deal with more types of corruption in the file.
</li><li>MVStore: the TransactionStore now first needs to be initialized before it can be used.
</li><li>Views and derived tables with equality and range conditions on the same columns
did not work properly. example: select x from (select x from (select 1 as x)
where x &gt; 0 and x &lt; 2) where x = 1
</li><li>The database URL setting PAGE_SIZE setting is now also used for the MVStore.
</li><li>MVStore: the default page split size for persistent stores is now 4096
(it was 16 KB so far). This should reduce the database file size for most situations
(in some cases, less than half the size of the previous version).
</li><li>With query literals disabled, auto-analyze of a table with CLOB or BLOB did not work.
</li><li>MVStore: use a mark and sweep GC algorithm instead of reference counting,
to ensure used chunks are never overwrite, even if the reference counting
algorithm does not work properly.
</li><li>In the multi-threaded mode, updating the column selectivity ("analyze")
in the background sometimes did not work.
</li><li>In the multi-threaded mode, database metadata operations
did sometimes not work if the schema was changed at the same time
(for example, if tables were dropped).
</li><li>Some CLOB and BLOB values could no longer be read when
the original row was removed (even when using the MVCC mode).
</li><li>The MVStoreTool could throw an IllegalArgumentException.
</li><li>Improved performance for some
date / time / timestamp conversion operations.
Thanks to Sergey Evdokimov for reporting the problem.
</li><li>H2 Console: the built-in web server did not work properly
if an unknown file was requested.
</li><li>MVStore: the jar file is renamed to "h2-mvstore-*.jar" and is
deployed to Maven separately.
</li><li>MVStore: support for concurrent reads and writes is now enabled by default.
</li><li>Server mode: the transfer buffer size has been changed from 16 KB to 64 KB,
after it was found that this improves performance on Linux quite a lot.
</li><li>H2 Console and server mode: SSL is now disabled and TLS is used
to protect against the Poodle SSLv3 vulnerability.
The system property to disable secure anonymous connections is now
"h2.enableAnonymousTLS".
The default certificate is still self-signed, so you need to manually install
another one if you want to avoid man in the middle attacks.
</li><li>MVStore: the R-tree did not correctly measure the memory usage.
</li><li>MVStore: compacting a store with an R-tree did not always work.
</li><li>Issue 581: When running in LOCK_MODE=0,
JdbcDatabaseMetaData#supportsTransactionIsolationLevel(TRANSACTION_READ_UNCOMMITTED)
should return false
</li><li>Fix bug which could generate deadlocks when multiple connections accessed the same table.
</li><li>Some places in the code were not respecting the value set in the "SET MAX_MEMORY_ROWS x" command
</li><li>Fix bug which could generate a NegativeArraySizeException when performing large (>40M) row union operations
</li><li>Fix "USE schema" command for MySQL compatibility, patch by mfulton
</li><li>Parse and ignore the ROW_FORMAT=DYNAMIC MySQL syntax, patch by mfulton
</li></ul>
<h2>Version 1.4.182 Beta (2014-10-17)</h2>
<ul><li>MVStore: improved error messages and logging;
improved behavior if there is an error when serializing objects.
</li><li>OSGi: the MVStore packages are now exported.
</li><li>With the MVStore option, when using multiple threads
that concurrently create indexes or tables,
it was relatively easy to get a lock timeout on the "SYS" table.
</li><li>When using the multi-threaded option, the exception
"Unexpected code path" could be thrown, specially if the option
"analyze_auto" was set to a low value.
</li><li>In the server mode, when reading from a CLOB or BLOB, if the connection
was closed, a NullPointerException could be thrown instead of an exception saying
the connection is closed.
</li><li>DatabaseMetaData.getProcedures and getProcedureColumns
could throw an exception if a user defined class is not available.
</li><li>Issue 584: the error message for a wrong sequence definition was wrong.
</li><li>CSV tool: the rowSeparator option is no longer supported,
as the same can be achieved with the lineSeparator.
</li><li>Descending indexes on MVStore tables did not work properly.
</li><li>Issue 579: Conditions on the "_rowid_" pseudo-column didn't use an index
when using the MVStore.
</li><li>Fixed documentation that "offset" and "fetch" are also keywords since version 1.4.x.
</li><li>The Long.MIN_VALUE could not be parsed for auto-increment (identity) columns.
</li><li>Issue 573: Add implementation for Methods "isWrapperFor()" and "unwrap()" in
other JDBC classes.
</li><li>Issue 572: MySQL compatibility for "order by" in update statements.
</li><li>The change in JDBC escape processing in version 1.4.181 affects both the parser
(which is running on the server) and the JDBC API (which is running on the client).
If you (or a tool you use) use the syntax "{t 'time}", or "{ts 'timestamp'}", or "{d 'data'}",
then both the client and the server need to be upgraded to version 1.4.181 or later.
</li></ul>
<h2>Version 1.4.181 Beta (2014-08-06)</h2>
<ul><li>Improved MySQL compatibility by supporting "use schema".
Thanks a lot to Karl Pietrzak for the patch!
</li><li>Writing to the trace file is now faster, specially with the debug level.
</li><li>The database option "defrag_always=true" did not work with the MVStore.
</li><li>The JDBC escape syntax {ts 'value'} did not interpret the value as a timestamp.
The same for {d 'value'} (for date) and {t 'value'} (for time).
Thanks to Lukas Eder for reporting the issue.
The following problem was detected after version 1.4.181 was released:
The change in JDBC escape processing affects both the parser (which is running on the server)
and the JDBC API (which is running on the client).
If you (or a tool you use) use the syntax {t 'time'}, or {ts 'timestamp'}, or {d 'date'},
then both the client and the server need to be upgraded to version 1.4.181 or later.
</li><li>File system abstraction: support replacing existing files using move
(currently not for Windows).
</li><li>The statement "shutdown defrag" now compresses the database (with the MVStore).
This command can greatly reduce the file size, and is relatively fast,
but is not incremental.
</li><li>The MVStore now automatically compacts the store in the background if there is no read or write activity,
which should (after some time; sometimes about one minute) reduce the file size.
This is still work in progress, feedback is welcome!
</li><li>Change default value of PAGE_SIZE from 2048 to 4096 to more closely match most file systems block size
(PageStore only; the MVStore already used 4096).
</li><li>Auto-scale MAX_MEMORY_ROWS and CACHE_SIZE settings by the amount of available RAM. Gives a better
out of box experience for people with more powerful machines.
</li><li>Handle tabs like 4 spaces in web console, patch by Martin Grajcar.
</li><li>Issue 573: Add implementation for Methods "isWrapperFor()" and "unwrap()" in JdbcConnection.java,
patch by BigMichi1.
</li></ul>
<h2>Version 1.4.180 Beta (2014-07-13)</h2>
<ul><li>MVStore: the store is now auto-compacted automatically up to some point,
to avoid very large file sizes. This area is still work in progress.
</li><li>Sequences of temporary tables (auto-increment or identity columns)
were persisted unnecessarily in the database file, and were not removed
when re-opening the database.
</li><li>MVStore: an IndexOutOfBoundsException could sometimes
occur MVMap.openVersion when concurrently accessing the store.
</li><li>The LIRS cache now re-sizes the internal hash map if needed.
</li><li>Optionally persist session history in the H2 console. (patch from Martin Grajcar)
</li><li>Add client-info property to get the number of servers currently in the cluster
and which servers that are available. (patch from Nikolaj Fogh)
</li><li>Fix bug in changing encrypted DB password that kept the file handle
open when the wrong password was supplied. (test case from Jens Hohmuth).
</li><li>Issue 567: H2 hangs for a long time then (sometimes) recovers.
Introduce a queue when doing table locking to prevent session starvation.
</li></ul>
<h2>Version 1.4.179 Beta (2014-06-23)</h2>
<ul><li>The license was changed to MPL 2.0 (from 1.0) and EPL 1.0.
</li><li>Issue 565: MVStore: concurrently adding LOB objects
(with MULTI_THREADED option) resulted in a NullPointerException.
</li><li>MVStore: reduced dependencies to other H2 classes.
</li><li>There was a way to prevent a database from being re-opened,
by creating a column constraint that references a table with a higher id,
for example with "check" constraints that contains queries.
This is now detected, and creating the table is prohibited.
In future versions of H2, most likely creating references to other
tables will no longer be supported because of such problems.
</li><li>MVStore: descending indexes with "nulls first" did not work as expected
(null was ordered last).
</li><li>Large result sets now always create temporary tables instead of temporary files.
</li><li>When using the PageStore, opening a database failed in some cases with a NullPointerException
if temporary tables were used (explicitly, or implicitly when using large result sets).
</li><li>If a database file in the PageStore file format exists, this file and this mode
is now used, even if the database URL does not contain "MV_STORE=FALSE".
If a MVStore file exists, it is used.
</li><li>Databases created with version 1.3.175 and earlier
that contained foreign keys in combination with multi-column indexes
could not be opened in some cases.
This was due to a bugfix in version 1.3.176:
Referential integrity constraints sometimes used the wrong index.
</li><li>MVStore: the ObjectDataType comparison method was incorrect if one
key was Serializable and the other was of a common class.
</li><li>Recursive queries with many result rows (more than the setting "max_memory_rows")
did not work correctly.
</li><li>The license has changed to MPL 2.0 + EPL 1.0.
</li><li>MVStore: temporary tables from result sets could survive re-opening a database,
which could result in a ClassCastException.
</li><li>Issue 566: MVStore: unique indexes that were created later on did not work correctly
if there were over 5000 rows in the table.
Existing databases need to be re-created (at least the broken index need to be re-built).
</li><li>MVStore: creating secondary indexes on large tables
results in missing rows in the index.
</li><li>Metadata: the password of linked tables is now only visible for admin users.
</li><li>For Windows, database URLs of the form "jdbc:h2:/test" where considered
relative and did not work unless the system property "h2.implicitRelativePath" was used.
</li><li>Windows: using a base directory of "C:/" and similar did not work as expected.
</li><li>Follow JDBC specification on Procedures MetaData, use P0 as
return type of procedure.
</li><li>Issue 531: IDENTITY ignored for added column.
</li><li>FileSystem: improve exception throwing compatibility with JDK
</li><li>Spatial Index: adjust costs so we do not use the spatial index if the
query does not contain an intersects operator.
</li><li>Fix multi-threaded deadlock when using a View that includes a TableFunction.
</li><li>Fix bug in dividing very-small BigDecimal numbers.
</li></ul>
<h2>Version 1.4.178 Beta (2014-05-02)</h2>
<ul><li>Issue 559: Make dependency on org.osgi.service.jdbc optional.
</li><li>Improve error message when the user specifies an unsupported combination of database settings.
</li><li>MVStore: in the multi-threaded mode, NullPointerException and other exceptions could occur.
</li><li>MVStore: some database file could not be compacted due to a bug in
the bookkeeping of the fill rate. Also, database file were compacted quite slowly.
This has been improved; but more changes in this area are expected.
</li><li>MVStore: support for volatile maps (that don't store changes).
</li><li>MVStore mode: in-memory databases now also use the MVStore.
</li><li>In server mode, appending ";autocommit=false" to the database URL was working,
but the return value of Connection.getAutoCommit() was wrong.
</li><li>Issue 561: OSGi: the import package declaration of org.h2 excluded version 1.4.
</li><li>Issue 558: with the MVStore, a NullPointerException could occur when using LOBs
at session commit (LobStorageMap.removeLob).
</li><li>Remove the "h2.MAX_MEMORY_ROWS_DISTINCT" system property to reduce confusion.
We already have the MAX_MEMORY_ROWS setting which does a very similar thing, and is better documented.
</li><li>Issue 554: Web Console in an IFrame was not fully supported.
</li></ul>
<h2>Version 1.4.177 Beta (2014-04-12)</h2>
<ul><li>By default, the MV_STORE option is enabled, so it is using the new MVStore
storage. The MVCC setting is by default set to the same values as the MV_STORE setting,
so it is also enabled by default. For testing, both settings can be disabled by appending
";MV_STORE=FALSE" and/or ";MVCC=FALSE" to the database URL.
</li><li>The file locking method 'serialized' is no longer supported.
This mode might return in a future version,
however this is not clear right now.
A new implementation and new tests would be needed.
</li><li>Enable the new storage format for dates (system property "h2.storeLocalTime").
For the MVStore mode, this is always enabled, but with version 1.4
this is even enabled in the PageStore mode.
</li><li>Implicit relative paths are disabled (system property "h2.implicitRelativePath"),
so that the database URL jdbc:h2:test now needs to be written as jdbc:h2:./test.
</li><li>"select ... fetch first 1 row only" is supported with the regular mode.
This was disabled so far because "fetch" and "offset" are now keywords.
See also Mode.supportOffsetFetch.
</li><li>Byte arrays are now sorted in unsigned mode
(x'99' is larger than x'09').
(System property "h2.sortBinaryUnsigned", Mode.binaryUnsigned, setting "binary_collation").
</li><li>Csv.getInstance will be removed in future versions of 1.4.
Use the public constructor instead.
</li><li>Remove support for the limited old-style outer join syntax using "(+)".
Use "outer join" instead.
System property "h2.oldStyleOuterJoin".
</li><li>Support the data type "DATETIME2" as an alias for "DATETIME", for MS SQL Server compatibility.
</li><li>Add Oracle-compatible TRANSLATE function, patch by Eric Chatellier.
</li></ul>
<h2>Version 1.3.176 (2014-04-05)</h2>
<ul><li>The file locking method 'serialized' is no longer documented,
as it will not be available in version 1.4.
</li><li>The static method Csv.getInstance() was removed.
Use the public constructor instead.
</li><li>The default user name for the Script, RunScript, Shell,
and CreateCluster tools are no longer "sa" but an empty string.
</li><li>The stack trace of the exception "The object is already closed" is no longer logged by default.
</li><li>If a value of a result set was itself a result set, the result
could only be read once.
</li><li>Column constraints are also visible in views (patch from Nicolas Fortin for H2GIS).
</li><li>Granting a additional right to a role that already had a right for that table was not working.
</li><li>Spatial index: a few bugs have been fixed (using spatial constraints in views,
transferring geometry objects over TCP/IP, the returned geometry object is copied when needed).
</li><li>Issue 551: the datatype documentation was incorrect (found by Bernd Eckenfels).
</li><li>Issue 368: ON DUPLICATE KEY UPDATE did not work for multi-row inserts.
Test case from Angus Macdonald.
</li><li>OSGi: the package javax.tools is now imported (as an optional).
</li><li>H2 Console: auto-complete is now disabled by default, but there is a hot-key (Ctrl+Space).
</li><li>H2 Console: auto-complete did not work with multi-line statements.
</li><li>CLOB and BLOB data was not immediately removed after a rollback.
</li><li>There is a new Aggregate API that supports the internal H2 data types
(GEOMETRY for example). Thanks a lot to Nicolas Fortin for the patch!
</li><li>Referential integrity constraints sometimes used the wrong index,
such that updating a row in the referenced table incorrectly failed with
a constraint violation.
</li><li>The Polish translation was completed and corrected by Wojtek Jurczyk. Thanks a lot!
</li><li>Issue 545: Unnecessary duplicate code was removed.
</li><li>The profiler tool can now process files with full thread dumps.
</li><li>MVStore: the file format was changed slightly.
</li><li>MVStore mode: the CLOB and BLOB storage was re-implemented and is
now much faster than with the PageStore (which is still the default storage).
</li><li>MVStore mode: creating indexes is now much faster
(in many cases faster than with the default PageStore).
</li><li>Various bugs in the MVStore storage and have been fixed,
including a bug in the R-tree implementation.
The database could get corrupt if there were transient IO exceptions while storing.
</li><li>The method org.h2.expression.Function.getCost could throw a NullPointException.
</li><li>Storing LOBs in separate files (outside of the main database file)
is no longer supported for new databases.
</li><li>Lucene 2 is no longer supported.
</li><li>Fix bug in calculating default MIN and MAX values for SEQUENCE.
</li><li>Fix bug in performing IN queries with multiple values when IGNORECASE=TRUE
</li><li>Add entry-point to org.h2.tools.Shell so it can be called from inside an application.
patch by Thomas Gillet.
</li><li>Fix bug that prevented the PgServer from being stopped and started multiple times.
</li><li>Support some more DDL syntax for MySQL, patch from Peter Jentsch.
</li><li>Issue 548: TO_CHAR does not format MM and DD correctly when the month or day of
the month is 1 digit, patch from "the.tucc"
</li><li>Fix bug in varargs support in ALIAS's, patch from Nicolas Fortin
</li></ul>
<!-- [close] { --></div></td></tr></table><!-- } --><!-- analytics --></body></html>
h2/src/main/org/h2/engine/SessionRemote.java
浏览文件 @ dfd877e2
/*
* Copyright 2004-2014 H2 Group. Multiple-Licensed under the MPL 2.0,
* and the EPL 1.0 (http://h2database.com/html/license.html).
* Initial Developer: H2 Group
*/
package org.h2.engine;
import java.io.IOException;
import java.net.Socket;
import java.util.ArrayList;
import org.h2.api.DatabaseEventListener;
import org.h2.api.ErrorCode;
import org.h2.api.JavaObjectSerializer;
import org.h2.command.CommandInterface;
import org.h2.command.CommandRemote;
import org.h2.command.dml.SetTypes;
import org.h2.jdbc.JdbcSQLException;
import org.h2.message.DbException;
import org.h2.message.Trace;
import org.h2.message.TraceSystem;
import org.h2.result.ResultInterface;
import org.h2.store.DataHandler;
import org.h2.store.FileStore;
import org.h2.store.LobStorageFrontend;
import org.h2.store.LobStorageInterface;
import org.h2.store.fs.FileUtils;
import org.h2.util.JdbcUtils;
import org.h2.util.MathUtils;
import org.h2.util.NetUtils;
import org.h2.util.New;
import org.h2.util.SmallLRUCache;
import org.h2.util.StringUtils;
import org.h2.util.TempFileDeleter;
import org.h2.value.Transfer;
import org.h2.value.Value;
/**
* The client side part of a session when using the server mode. This object
* communicates with a Session on the server side.
*/
public class SessionRemote extends SessionWithState implements DataHandler {
public static final int SESSION_PREPARE = 0;
public static final int SESSION_CLOSE = 1;
public static final int COMMAND_EXECUTE_QUERY = 2;
public static final int COMMAND_EXECUTE_UPDATE = 3;
public static final int COMMAND_CLOSE = 4;
public static final int RESULT_FETCH_ROWS = 5;
public static final int RESULT_RESET = 6;
public static final int RESULT_CLOSE = 7;
public static final int COMMAND_COMMIT = 8;
public static final int CHANGE_ID = 9;
public static final int COMMAND_GET_META_DATA = 10;
public static final int SESSION_PREPARE_READ_PARAMS = 11;
public static final int SESSION_SET_ID = 12;
public static final int SESSION_CANCEL_STATEMENT = 13;
public static final int SESSION_CHECK_KEY = 14;
public static final int SESSION_SET_AUTOCOMMIT = 15;
public static final int SESSION_HAS_PENDING_TRANSACTION = 16;
public static final int LOB_READ = 17;
public static final int STATUS_ERROR = 0;
public static final int STATUS_OK = 1;
public static final int STATUS_CLOSED = 2;
public static final int STATUS_OK_STATE_CHANGED = 3;
private static SessionFactory sessionFactory;
private TraceSystem traceSystem;
private Trace trace;
private ArrayList<Transfer> transferList = New.arrayList();
private int nextId;
private boolean autoCommit = true;
private CommandInterface autoCommitFalse, autoCommitTrue;
private ConnectionInfo connectionInfo;
private String databaseName;
private String cipher;
private byte[] fileEncryptionKey;
private final Object lobSyncObject = new Object();
private String sessionId;
private int clientVersion;
private boolean autoReconnect;
private int lastReconnect;
private SessionInterface embedded;
private DatabaseEventListener eventListener;
private LobStorageFrontend lobStorage;
private boolean cluster;
private TempFileDeleter tempFileDeleter;
private JavaObjectSerializer javaObjectSerializer;
private volatile boolean javaObjectSerializerInitialized;
public SessionRemote(ConnectionInfo ci) {
this.connectionInfo = ci;
}
@Override
public ArrayList<String> getClusterServers() {
ArrayList<String> serverList = new ArrayList<String>();
for (int i = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
serverList.add(transfer.getSocket().getInetAddress().
getHostAddress() + ":" +
transfer.getSocket().getPort());
}
return serverList;
}
private Transfer initTransfer(ConnectionInfo ci, String db, String server)
throws IOException {
Socket socket = NetUtils.createSocket(server,
Constants.DEFAULT_TCP_PORT, ci.isSSL());
Transfer trans = new Transfer(this);
trans.setSocket(socket);
trans.setSSL(ci.isSSL());
trans.init();
trans.writeInt(Constants.TCP_PROTOCOL_VERSION_6);
trans.writeInt(Constants.TCP_PROTOCOL_VERSION_15);
trans.writeString(db);
trans.writeString(ci.getOriginalURL());
trans.writeString(ci.getUserName());
trans.writeBytes(ci.getUserPasswordHash());
trans.writeBytes(ci.getFilePasswordHash());
String[] keys = ci.getKeys();
trans.writeInt(keys.length);
for (String key : keys) {
trans.writeString(key).writeString(ci.getProperty(key));
}
try {
done(trans);
clientVersion = trans.readInt();
trans.setVersion(clientVersion);
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_14) {
if (ci.getFileEncryptionKey() != null) {
trans.writeBytes(ci.getFileEncryptionKey());
}
}
trans.writeInt(SessionRemote.SESSION_SET_ID);
trans.writeString(sessionId);
done(trans);
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_15) {
autoCommit = trans.readBoolean();
} else {
autoCommit = true;
}
return trans;
} catch (DbException e) {
trans.close();
throw e;
}
}
@Override
public boolean hasPendingTransaction() {
if (clientVersion < Constants.TCP_PROTOCOL_VERSION_10) {
return true;
}
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("SESSION_HAS_PENDING_TRANSACTION", 0);
transfer.writeInt(
SessionRemote.SESSION_HAS_PENDING_TRANSACTION);
done(transfer);
return transfer.readInt() != 0;
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
return true;
}
@Override
public void cancel() {
// this method is called when closing the connection
// the statement that is currently running is not canceled in this case
// however Statement.cancel is supported
}
/**
* Cancel the statement with the given id.
*
* @param id the statement id
*/
public void cancelStatement(int id) {
for (Transfer transfer : transferList) {
try {
Transfer trans = transfer.openNewConnection();
trans.init();
trans.writeInt(clientVersion);
trans.writeInt(clientVersion);
trans.writeString(null);
trans.writeString(null);
trans.writeString(sessionId);
trans.writeInt(SessionRemote.SESSION_CANCEL_STATEMENT);
trans.writeInt(id);
trans.close();
} catch (IOException e) {
trace.debug(e, "could not cancel statement");
}
}
}
private void checkClusterDisableAutoCommit(String serverList) {
if (autoCommit && transferList.size() > 1) {
setAutoCommitSend(false);
CommandInterface c = prepareCommand(
"SET CLUSTER " + serverList, Integer.MAX_VALUE);
// this will set autoCommit to false
c.executeUpdate();
// so we need to switch it on
autoCommit = true;
cluster = true;
}
}
@Override
public boolean getAutoCommit() {
return autoCommit;
}
@Override
public void setAutoCommit(boolean autoCommit) {
if (!cluster) {
setAutoCommitSend(autoCommit);
}
this.autoCommit = autoCommit;
}
public void setAutoCommitFromServer(boolean autoCommit) {
if (cluster) {
if (autoCommit) {
// the user executed SET AUTOCOMMIT TRUE
setAutoCommitSend(false);
this.autoCommit = true;
}
} else {
this.autoCommit = autoCommit;
}
}
private void setAutoCommitSend(boolean autoCommit) {
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_8) {
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("SESSION_SET_AUTOCOMMIT", autoCommit ? 1 : 0);
transfer.writeInt(SessionRemote.SESSION_SET_AUTOCOMMIT).
writeBoolean(autoCommit);
done(transfer);
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
} else {
if (autoCommit) {
if (autoCommitTrue == null) {
autoCommitTrue = prepareCommand(
"SET AUTOCOMMIT TRUE", Integer.MAX_VALUE);
}
autoCommitTrue.executeUpdate();
} else {
if (autoCommitFalse == null) {
autoCommitFalse = prepareCommand(
"SET AUTOCOMMIT FALSE", Integer.MAX_VALUE);
}
autoCommitFalse.executeUpdate();
}
}
}
/**
* Calls COMMIT if the session is in cluster mode.
*/
public void autoCommitIfCluster() {
if (autoCommit && cluster) {
// server side auto commit is off because of race conditions
// (update set id=1 where id=0, but update set id=2 where id=0 is
// faster)
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("COMMAND_COMMIT", 0);
transfer.writeInt(SessionRemote.COMMAND_COMMIT);
done(transfer);
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
}
}
private String getFilePrefix(String dir) {
StringBuilder buff = new StringBuilder(dir);
buff.append('/');
for (int i = 0; i < databaseName.length(); i++) {
char ch = databaseName.charAt(i);
if (Character.isLetterOrDigit(ch)) {
buff.append(ch);
} else {
buff.append('_');
}
}
return buff.toString();
}
@Override
public int getPowerOffCount() {
return 0;
}
@Override
public void setPowerOffCount(int count) {
throw DbException.getUnsupportedException("remote");
}
/**
* Open a new (remote or embedded) session.
*
* @param openNew whether to open a new session in any case
* @return the session
*/
public SessionInterface connectEmbeddedOrServer(boolean openNew) {
ConnectionInfo ci = connectionInfo;
if (ci.isRemote()) {
connectServer(ci);
return this;
}
// create the session using reflection,
// so that the JDBC layer can be compiled without it
boolean autoServerMode = Boolean.parseBoolean(
ci.getProperty("AUTO_SERVER", "false"));
ConnectionInfo backup = null;
try {
if (autoServerMode) {
backup = ci.clone();
connectionInfo = ci.clone();
}
if (openNew) {
ci.setProperty("OPEN_NEW", "true");
}
if (sessionFactory == null) {
sessionFactory = (SessionFactory) Class.forName(
"org.h2.engine.Engine").getMethod("getInstance").invoke(null);
}
return sessionFactory.createSession(ci);
} catch (Exception re) {
DbException e = DbException.convert(re);
if (e.getErrorCode() == ErrorCode.DATABASE_ALREADY_OPEN_1) {
if (autoServerMode) {
String serverKey = ((JdbcSQLException) e.getSQLException()).
getSQL();
if (serverKey != null) {
backup.setServerKey(serverKey);
// OPEN_NEW must be removed now, otherwise
// opening a session with AUTO_SERVER fails
// if another connection is already open
backup.removeProperty("OPEN_NEW", null);
connectServer(backup);
return this;
}
}
}
throw e;
}
}
private void connectServer(ConnectionInfo ci) {
String name = ci.getName();
if (name.startsWith("//")) {
name = name.substring("//".length());
}
int idx = name.indexOf('/');
if (idx < 0) {
throw ci.getFormatException();
}
databaseName = name.substring(idx + 1);
String server = name.substring(0, idx);
traceSystem = new TraceSystem(null);
String traceLevelFile = ci.getProperty(
SetTypes.TRACE_LEVEL_FILE, null);
if (traceLevelFile != null) {
int level = Integer.parseInt(traceLevelFile);
String prefix = getFilePrefix(
SysProperties.CLIENT_TRACE_DIRECTORY);
try {
traceSystem.setLevelFile(level);
if (level > 0 && level < 4) {
String file = FileUtils.createTempFile(prefix,
Constants.SUFFIX_TRACE_FILE, false, false);
traceSystem.setFileName(file);
}
} catch (IOException e) {
throw DbException.convertIOException(e, prefix);
}
}
String traceLevelSystemOut = ci.getProperty(
SetTypes.TRACE_LEVEL_SYSTEM_OUT, null);
if (traceLevelSystemOut != null) {
int level = Integer.parseInt(traceLevelSystemOut);
traceSystem.setLevelSystemOut(level);
}
trace = traceSystem.getTrace(Trace.JDBC);
String serverList = null;
if (server.indexOf(',') >= 0) {
serverList = StringUtils.quoteStringSQL(server);
ci.setProperty("CLUSTER", Constants.CLUSTERING_ENABLED);
}
autoReconnect = Boolean.parseBoolean(ci.getProperty(
"AUTO_RECONNECT", "false"));
// AUTO_SERVER implies AUTO_RECONNECT
boolean autoServer = Boolean.parseBoolean(ci.getProperty(
"AUTO_SERVER", "false"));
if (autoServer && serverList != null) {
throw DbException
.getUnsupportedException("autoServer && serverList != null");
}
autoReconnect |= autoServer;
if (autoReconnect) {
String className = ci.getProperty("DATABASE_EVENT_LISTENER");
if (className != null) {
className = StringUtils.trim(className, true, true, "'");
try {
eventListener = (DatabaseEventListener) JdbcUtils
.loadUserClass(className).newInstance();
} catch (Throwable e) {
throw DbException.convert(e);
}
}
}
cipher = ci.getProperty("CIPHER");
if (cipher != null) {
fileEncryptionKey = MathUtils.secureRandomBytes(32);
}
String[] servers = StringUtils.arraySplit(server, ',', true);
int len = servers.length;
transferList.clear();
sessionId = StringUtils.convertBytesToHex(MathUtils.secureRandomBytes(32));
// TODO cluster: support more than 2 connections
boolean switchOffCluster = false;
try {
for (int i = 0; i < len; i++) {
String s = servers[i];
try {
Transfer trans = initTransfer(ci, databaseName, s);
transferList.add(trans);
} catch (IOException e) {
if (len == 1) {
throw DbException.get(ErrorCode.CONNECTION_BROKEN_1, e, e + ": " + s);
}
switchOffCluster = true;
}
}
checkClosed();
if (switchOffCluster) {
switchOffCluster();
}
checkClusterDisableAutoCommit(serverList);
} catch (DbException e) {
traceSystem.close();
throw e;
}
}
private void switchOffCluster() {
CommandInterface ci = prepareCommand("SET CLUSTER ''", Integer.MAX_VALUE);
ci.executeUpdate();
}
/**
* Remove a server from the list of cluster nodes and disables the cluster
* mode.
*
* @param e the exception (used for debugging)
* @param i the index of the server to remove
* @param count the retry count index
*/
public void removeServer(IOException e, int i, int count) {
trace.debug(e, "removing server because of exception");
transferList.remove(i);
if (transferList.size() == 0 && autoReconnect(count)) {
return;
}
checkClosed();
switchOffCluster();
}
@Override
public synchronized CommandInterface prepareCommand(String sql, int fetchSize) {
checkClosed();
return new CommandRemote(this, transferList, sql, fetchSize);
}
/**
* Automatically re-connect if necessary and if configured to do so.
*
* @param count the retry count index
* @return true if reconnected
*/
private boolean autoReconnect(int count) {
if (!isClosed()) {
return false;
}
if (!autoReconnect) {
return false;
}
if (!cluster && !autoCommit) {
return false;
}
if (count > SysProperties.MAX_RECONNECT) {
return false;
}
lastReconnect++;
while (true) {
try {
embedded = connectEmbeddedOrServer(false);
break;
} catch (DbException e) {
if (e.getErrorCode() != ErrorCode.DATABASE_IS_IN_EXCLUSIVE_MODE) {
throw e;
}
// exclusive mode: re-try endlessly
try {
Thread.sleep(500);
} catch (Exception e2) {
// ignore
}
}
}
if (embedded == this) {
// connected to a server somewhere else
embedded = null;
} else {
// opened an embedded connection now -
// must connect to this database in server mode
// unfortunately
connectEmbeddedOrServer(true);
}
recreateSessionState();
if (eventListener != null) {
eventListener.setProgress(DatabaseEventListener.STATE_RECONNECTED,
databaseName, count, SysProperties.MAX_RECONNECT);
}
return true;
}
/**
* Check if this session is closed and throws an exception if so.
*
* @throws DbException if the session is closed
*/
public void checkClosed() {
if (isClosed()) {
throw DbException.get(ErrorCode.CONNECTION_BROKEN_1, "session closed");
}
}
@Override
public void close() {
RuntimeException closeError = null;
if (transferList != null) {
synchronized (this) {
for (Transfer transfer : transferList) {
try {
traceOperation("SESSION_CLOSE", 0);
transfer.writeInt(SessionRemote.SESSION_CLOSE);
done(transfer);
transfer.close();
} catch (RuntimeException e) {
trace.error(e, "close");
closeError = e;
} catch (Exception e) {
trace.error(e, "close");
}
}
}
transferList = null;
}
traceSystem.close();
if (embedded != null) {
embedded.close();
embedded = null;
}
if (closeError != null) {
throw closeError;
}
}
@Override
public Trace getTrace() {
return traceSystem.getTrace(Trace.JDBC);
}
public int getNextId() {
return nextId++;
}
public int getCurrentId() {
return nextId;
}
/**
* Called to flush the output after data has been sent to the server and
* just before receiving data. This method also reads the status code from
* the server and throws any exception the server sent.
*
* @param transfer the transfer object
* @throws DbException if the server sent an exception
* @throws IOException if there is a communication problem between client
* and server
*/
public void done(Transfer transfer) throws IOException {
transfer.flush();
int status = transfer.readInt();
if (status == STATUS_ERROR) {
String sqlstate = transfer.readString();
String message = transfer.readString();
String sql = transfer.readString();
int errorCode = transfer.readInt();
String stackTrace = transfer.readString();
JdbcSQLException s = new JdbcSQLException(message, sql, sqlstate,
errorCode, null, stackTrace);
if (errorCode == ErrorCode.CONNECTION_BROKEN_1) {
// allow re-connect
IOException e = new IOException(s.toString(), s);
throw e;
}
throw DbException.convert(s);
} else if (status == STATUS_CLOSED) {
transferList = null;
} else if (status == STATUS_OK_STATE_CHANGED) {
sessionStateChanged = true;
} else if (status == STATUS_OK) {
// ok
} else {
throw DbException.get(ErrorCode.CONNECTION_BROKEN_1,
"unexpected status " + status);
}
}
/**
* Returns true if the connection was opened in cluster mode.
*
* @return true if it is
*/
public boolean isClustered() {
return cluster;
}
@Override
public boolean isClosed() {
return transferList == null || transferList.size() == 0;
}
/**
* Write the operation to the trace system if debug trace is enabled.
*
* @param operation the operation performed
* @param id the id of the operation
*/
public void traceOperation(String operation, int id) {
if (trace.isDebugEnabled()) {
trace.debug("{0} {1}", operation, id);
}
}
@Override
public void checkPowerOff() {
// ok
}
@Override
public void checkWritingAllowed() {
// ok
}
@Override
public String getDatabasePath() {
return "";
}
@Override
public String getLobCompressionAlgorithm(int type) {
return null;
}
@Override
public int getMaxLengthInplaceLob() {
return SysProperties.LOB_CLIENT_MAX_SIZE_MEMORY;
}
@Override
public FileStore openFile(String name, String mode, boolean mustExist) {
if (mustExist && !FileUtils.exists(name)) {
throw DbException.get(ErrorCode.FILE_NOT_FOUND_1, name);
}
FileStore store;
if (cipher == null) {
store = FileStore.open(this, name, mode);
} else {
store = FileStore.open(this, name, mode, cipher, fileEncryptionKey, 0);
}
store.setCheckedWriting(false);
try {
store.init();
} catch (DbException e) {
store.closeSilently();
throw e;
}
return store;
}
@Override
public DataHandler getDataHandler() {
return this;
}
@Override
public Object getLobSyncObject() {
return lobSyncObject;
}
@Override
public SmallLRUCache<String, String[]> getLobFileListCache() {
return null;
}
public int getLastReconnect() {
return lastReconnect;
}
@Override
public TempFileDeleter getTempFileDeleter() {
if (tempFileDeleter == null) {
tempFileDeleter = TempFileDeleter.getInstance();
}
return tempFileDeleter;
}
@Override
public boolean isReconnectNeeded(boolean write) {
return false;
}
@Override
public SessionInterface reconnect(boolean write) {
return this;
}
@Override
public void afterWriting() {
// nothing to do
}
@Override
public LobStorageInterface getLobStorage() {
if (lobStorage == null) {
lobStorage = new LobStorageFrontend(this);
}
return lobStorage;
}
@Override
public synchronized int readLob(long lobId, byte[] hmac, long offset,
byte[] buff, int off, int length) {
checkClosed();
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("LOB_READ", (int) lobId);
transfer.writeInt(SessionRemote.LOB_READ);
transfer.writeLong(lobId);
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_12) {
transfer.writeBytes(hmac);
}
transfer.writeLong(offset);
transfer.writeInt(length);
done(transfer);
length = transfer.readInt();
if (length <= 0) {
return length;
}
transfer.readBytes(buff, off, length);
return length;
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
return 1;
}
@Override
public JavaObjectSerializer getJavaObjectSerializer() {
initJavaObjectSerializer();
return javaObjectSerializer;
}
private void initJavaObjectSerializer() {
if (javaObjectSerializerInitialized) {
return;
}
synchronized (this) {
if (javaObjectSerializerInitialized) {
return;
}
String serializerFQN = readSerializationSettings();
if (serializerFQN != null) {
serializerFQN = serializerFQN.trim();
if (!serializerFQN.isEmpty() && !serializerFQN.equals("null")) {
try {
javaObjectSerializer = (JavaObjectSerializer) JdbcUtils
.loadUserClass(serializerFQN).newInstance();
} catch (Exception e) {
throw DbException.convert(e);
}
}
}
javaObjectSerializerInitialized = true;
}
}
/**
* Read the serializer name from the persistent database settings.
*
* @return the serializer
*/
private String readSerializationSettings() {
String javaObjectSerializerFQN = null;
CommandInterface ci = prepareCommand(
"SELECT VALUE FROM INFORMATION_SCHEMA.SETTINGS "+
" WHERE NAME='JAVA_OBJECT_SERIALIZER'", Integer.MAX_VALUE);
try {
ResultInterface result = ci.executeQuery(0, false);
if (result.next()) {
Value[] row = result.currentRow();
javaObjectSerializerFQN = row[0].getString();
}
} finally {
ci.close();
}
return javaObjectSerializerFQN;
}
@Override
public void addTemporaryLob(Value v) {
// do nothing
}
}
/*
* Copyright 2004-2014 H2 Group. Multiple-Licensed under the MPL 2.0,
* and the EPL 1.0 (http://h2database.com/html/license.html).
* Initial Developer: H2 Group
*/
package org.h2.engine;
import java.io.IOException;
import java.net.Socket;
import java.util.ArrayList;
import org.h2.api.DatabaseEventListener;
import org.h2.api.ErrorCode;
import org.h2.api.JavaObjectSerializer;
import org.h2.command.CommandInterface;
import org.h2.command.CommandRemote;
import org.h2.command.dml.SetTypes;
import org.h2.jdbc.JdbcSQLException;
import org.h2.message.DbException;
import org.h2.message.Trace;
import org.h2.message.TraceSystem;
import org.h2.result.ResultInterface;
import org.h2.store.DataHandler;
import org.h2.store.FileStore;
import org.h2.store.LobStorageFrontend;
import org.h2.store.LobStorageInterface;
import org.h2.store.fs.FileUtils;
import org.h2.util.JdbcUtils;
import org.h2.util.MathUtils;
import org.h2.util.NetUtils;
import org.h2.util.New;
import org.h2.util.SmallLRUCache;
import org.h2.util.StringUtils;
import org.h2.util.TempFileDeleter;
import org.h2.value.Transfer;
import org.h2.value.Value;
/**
* The client side part of a session when using the server mode. This object
* communicates with a Session on the server side.
*/
public class SessionRemote extends SessionWithState implements DataHandler {
public static final int SESSION_PREPARE = 0;
public static final int SESSION_CLOSE = 1;
public static final int COMMAND_EXECUTE_QUERY = 2;
public static final int COMMAND_EXECUTE_UPDATE = 3;
public static final int COMMAND_CLOSE = 4;
public static final int RESULT_FETCH_ROWS = 5;
public static final int RESULT_RESET = 6;
public static final int RESULT_CLOSE = 7;
public static final int COMMAND_COMMIT = 8;
public static final int CHANGE_ID = 9;
public static final int COMMAND_GET_META_DATA = 10;
public static final int SESSION_PREPARE_READ_PARAMS = 11;
public static final int SESSION_SET_ID = 12;
public static final int SESSION_CANCEL_STATEMENT = 13;
public static final int SESSION_CHECK_KEY = 14;
public static final int SESSION_SET_AUTOCOMMIT = 15;
public static final int SESSION_HAS_PENDING_TRANSACTION = 16;
public static final int LOB_READ = 17;
public static final int STATUS_ERROR = 0;
public static final int STATUS_OK = 1;
public static final int STATUS_CLOSED = 2;
public static final int STATUS_OK_STATE_CHANGED = 3;
private static SessionFactory sessionFactory;
private TraceSystem traceSystem;
private Trace trace;
private ArrayList<Transfer> transferList = New.arrayList();
private int nextId;
private boolean autoCommit = true;
private CommandInterface autoCommitFalse, autoCommitTrue;
private ConnectionInfo connectionInfo;
private String databaseName;
private String cipher;
private byte[] fileEncryptionKey;
private final Object lobSyncObject = new Object();
private String sessionId;
private int clientVersion;
private boolean autoReconnect;
private int lastReconnect;
private SessionInterface embedded;
private DatabaseEventListener eventListener;
private LobStorageFrontend lobStorage;
private boolean cluster;
private TempFileDeleter tempFileDeleter;
private JavaObjectSerializer javaObjectSerializer;
private volatile boolean javaObjectSerializerInitialized;
public SessionRemote(ConnectionInfo ci) {
this.connectionInfo = ci;
}
@Override
public ArrayList<String> getClusterServers() {
ArrayList<String> serverList = new ArrayList<String>();
for (int i = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
serverList.add(transfer.getSocket().getInetAddress().
getHostAddress() + ":" +
transfer.getSocket().getPort());
}
return serverList;
}
private Transfer initTransfer(ConnectionInfo ci, String db, String server)
throws IOException {
Socket socket = NetUtils.createSocket(server,
Constants.DEFAULT_TCP_PORT, ci.isSSL());
Transfer trans = new Transfer(this);
trans.setSocket(socket);
trans.setSSL(ci.isSSL());
trans.init();
trans.writeInt(Constants.TCP_PROTOCOL_VERSION_6);
trans.writeInt(Constants.TCP_PROTOCOL_VERSION_15);
trans.writeString(db);
trans.writeString(ci.getOriginalURL());
trans.writeString(ci.getUserName());
trans.writeBytes(ci.getUserPasswordHash());
trans.writeBytes(ci.getFilePasswordHash());
String[] keys = ci.getKeys();
trans.writeInt(keys.length);
for (String key : keys) {
trans.writeString(key).writeString(ci.getProperty(key));
}
try {
done(trans);
clientVersion = trans.readInt();
trans.setVersion(clientVersion);
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_14) {
if (ci.getFileEncryptionKey() != null) {
trans.writeBytes(ci.getFileEncryptionKey());
}
}
trans.writeInt(SessionRemote.SESSION_SET_ID);
trans.writeString(sessionId);
done(trans);
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_15) {
autoCommit = trans.readBoolean();
} else {
autoCommit = true;
}
return trans;
} catch (DbException e) {
trans.close();
throw e;
}
}
@Override
public boolean hasPendingTransaction() {
if (clientVersion < Constants.TCP_PROTOCOL_VERSION_10) {
return true;
}
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("SESSION_HAS_PENDING_TRANSACTION", 0);
transfer.writeInt(
SessionRemote.SESSION_HAS_PENDING_TRANSACTION);
done(transfer);
return transfer.readInt() != 0;
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
return true;
}
@Override
public void cancel() {
// this method is called when closing the connection
// the statement that is currently running is not canceled in this case
// however Statement.cancel is supported
}
/**
* Cancel the statement with the given id.
*
* @param id the statement id
*/
public void cancelStatement(int id) {
for (Transfer transfer : transferList) {
try {
Transfer trans = transfer.openNewConnection();
trans.init();
trans.writeInt(clientVersion);
trans.writeInt(clientVersion);
trans.writeString(null);
trans.writeString(null);
trans.writeString(sessionId);
trans.writeInt(SessionRemote.SESSION_CANCEL_STATEMENT);
trans.writeInt(id);
trans.close();
} catch (IOException e) {
trace.debug(e, "could not cancel statement");
}
}
}
private void checkClusterDisableAutoCommit(String serverList) {
if (autoCommit && transferList.size() > 1) {
setAutoCommitSend(false);
CommandInterface c = prepareCommand(
"SET CLUSTER " + serverList, Integer.MAX_VALUE);
// this will set autoCommit to false
c.executeUpdate();
// so we need to switch it on
autoCommit = true;
cluster = true;
}
}
@Override
public boolean getAutoCommit() {
return autoCommit;
}
@Override
public void setAutoCommit(boolean autoCommit) {
if (!cluster) {
setAutoCommitSend(autoCommit);
}
this.autoCommit = autoCommit;
}
public void setAutoCommitFromServer(boolean autoCommit) {
if (cluster) {
if (autoCommit) {
// the user executed SET AUTOCOMMIT TRUE
setAutoCommitSend(false);
this.autoCommit = true;
}
} else {
this.autoCommit = autoCommit;
}
}
private void setAutoCommitSend(boolean autoCommit) {
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_8) {
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("SESSION_SET_AUTOCOMMIT", autoCommit ? 1 : 0);
transfer.writeInt(SessionRemote.SESSION_SET_AUTOCOMMIT).
writeBoolean(autoCommit);
done(transfer);
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
} else {
if (autoCommit) {
if (autoCommitTrue == null) {
autoCommitTrue = prepareCommand(
"SET AUTOCOMMIT TRUE", Integer.MAX_VALUE);
}
autoCommitTrue.executeUpdate();
} else {
if (autoCommitFalse == null) {
autoCommitFalse = prepareCommand(
"SET AUTOCOMMIT FALSE", Integer.MAX_VALUE);
}
autoCommitFalse.executeUpdate();
}
}
}
/**
* Calls COMMIT if the session is in cluster mode.
*/
public void autoCommitIfCluster() {
if (autoCommit && cluster) {
// server side auto commit is off because of race conditions
// (update set id=1 where id=0, but update set id=2 where id=0 is
// faster)
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("COMMAND_COMMIT", 0);
transfer.writeInt(SessionRemote.COMMAND_COMMIT);
done(transfer);
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
}
}
private String getFilePrefix(String dir) {
StringBuilder buff = new StringBuilder(dir);
buff.append('/');
for (int i = 0; i < databaseName.length(); i++) {
char ch = databaseName.charAt(i);
if (Character.isLetterOrDigit(ch)) {
buff.append(ch);
} else {
buff.append('_');
}
}
return buff.toString();
}
@Override
public int getPowerOffCount() {
return 0;
}
@Override
public void setPowerOffCount(int count) {
throw DbException.getUnsupportedException("remote");
}
/**
* Open a new (remote or embedded) session.
*
* @param openNew whether to open a new session in any case
* @return the session
*/
public SessionInterface connectEmbeddedOrServer(boolean openNew) {
ConnectionInfo ci = connectionInfo;
if (ci.isRemote()) {
connectServer(ci);
return this;
}
// create the session using reflection,
// so that the JDBC layer can be compiled without it
boolean autoServerMode = Boolean.parseBoolean(
ci.getProperty("AUTO_SERVER", "false"));
ConnectionInfo backup = null;
try {
if (autoServerMode) {
backup = ci.clone();
connectionInfo = ci.clone();
}
if (openNew) {
ci.setProperty("OPEN_NEW", "true");
}
if (sessionFactory == null) {
sessionFactory = (SessionFactory) Class.forName(
"org.h2.engine.Engine").getMethod("getInstance").invoke(null);
}
return sessionFactory.createSession(ci);
} catch (Exception re) {
DbException e = DbException.convert(re);
if (e.getErrorCode() == ErrorCode.DATABASE_ALREADY_OPEN_1) {
if (autoServerMode) {
String serverKey = ((JdbcSQLException) e.getSQLException()).
getSQL();
if (serverKey != null) {
backup.setServerKey(serverKey);
// OPEN_NEW must be removed now, otherwise
// opening a session with AUTO_SERVER fails
// if another connection is already open
backup.removeProperty("OPEN_NEW", null);
connectServer(backup);
return this;
}
}
}
throw e;
}
}
private void connectServer(ConnectionInfo ci) {
String name = ci.getName();
if (name.startsWith("//")) {
name = name.substring("//".length());
}
int idx = name.indexOf('/');
if (idx < 0) {
throw ci.getFormatException();
}
databaseName = name.substring(idx + 1);
String server = name.substring(0, idx);
traceSystem = new TraceSystem(null);
String traceLevelFile = ci.getProperty(
SetTypes.TRACE_LEVEL_FILE, null);
if (traceLevelFile != null) {
int level = Integer.parseInt(traceLevelFile);
String prefix = getFilePrefix(
SysProperties.CLIENT_TRACE_DIRECTORY);
try {
traceSystem.setLevelFile(level);
if (level > 0 && level < 4) {
String file = FileUtils.createTempFile(prefix,
Constants.SUFFIX_TRACE_FILE, false, false);
traceSystem.setFileName(file);
}
} catch (IOException e) {
throw DbException.convertIOException(e, prefix);
}
}
String traceLevelSystemOut = ci.getProperty(
SetTypes.TRACE_LEVEL_SYSTEM_OUT, null);
if (traceLevelSystemOut != null) {
int level = Integer.parseInt(traceLevelSystemOut);
traceSystem.setLevelSystemOut(level);
}
trace = traceSystem.getTrace(Trace.JDBC);
String serverList = null;
if (server.indexOf(',') >= 0) {
serverList = StringUtils.quoteStringSQL(server);
ci.setProperty("CLUSTER", Constants.CLUSTERING_ENABLED);
}
autoReconnect = Boolean.parseBoolean(ci.getProperty(
"AUTO_RECONNECT", "false"));
// AUTO_SERVER implies AUTO_RECONNECT
boolean autoServer = Boolean.parseBoolean(ci.getProperty(
"AUTO_SERVER", "false"));
if (autoServer && serverList != null) {
throw DbException
.getUnsupportedException("autoServer && serverList != null");
}
autoReconnect |= autoServer;
if (autoReconnect) {
String className = ci.getProperty("DATABASE_EVENT_LISTENER");
if (className != null) {
className = StringUtils.trim(className, true, true, "'");
try {
eventListener = (DatabaseEventListener) JdbcUtils
.loadUserClass(className).newInstance();
} catch (Throwable e) {
throw DbException.convert(e);
}
}
}
cipher = ci.getProperty("CIPHER");
if (cipher != null) {
fileEncryptionKey = MathUtils.secureRandomBytes(32);
}
String[] servers = StringUtils.arraySplit(server, ',', true);
int len = servers.length;
transferList.clear();
sessionId = StringUtils.convertBytesToHex(MathUtils.secureRandomBytes(32));
// TODO cluster: support more than 2 connections
boolean switchOffCluster = false;
try {
for (int i = 0; i < len; i++) {
String s = servers[i];
try {
Transfer trans = initTransfer(ci, databaseName, s);
transferList.add(trans);
} catch (IOException e) {
if (len == 1) {
throw DbException.get(ErrorCode.CONNECTION_BROKEN_1, e, e + ": " + s);
}
switchOffCluster = true;
}
}
checkClosed();
if (switchOffCluster) {
switchOffCluster();
}
checkClusterDisableAutoCommit(serverList);
} catch (DbException e) {
traceSystem.close();
throw e;
}
}
private void switchOffCluster() {
CommandInterface ci = prepareCommand("SET CLUSTER ''", Integer.MAX_VALUE);
ci.executeUpdate();
}
/**
* Remove a server from the list of cluster nodes and disables the cluster
* mode.
*
* @param e the exception (used for debugging)
* @param i the index of the server to remove
* @param count the retry count index
*/
public void removeServer(IOException e, int i, int count) {
trace.error(e, "removing server because of exception");
transferList.remove(i);
if (transferList.size() == 0 && autoReconnect(count)) {
return;
}
checkClosed();
switchOffCluster();
}
@Override
public synchronized CommandInterface prepareCommand(String sql, int fetchSize) {
checkClosed();
return new CommandRemote(this, transferList, sql, fetchSize);
}
/**
* Automatically re-connect if necessary and if configured to do so.
*
* @param count the retry count index
* @return true if reconnected
*/
private boolean autoReconnect(int count) {
if (!isClosed()) {
return false;
}
if (!autoReconnect) {
return false;
}
if (!cluster && !autoCommit) {
return false;
}
if (count > SysProperties.MAX_RECONNECT) {
return false;
}
lastReconnect++;
while (true) {
try {
embedded = connectEmbeddedOrServer(false);
break;
} catch (DbException e) {
if (e.getErrorCode() != ErrorCode.DATABASE_IS_IN_EXCLUSIVE_MODE) {
throw e;
}
// exclusive mode: re-try endlessly
try {
Thread.sleep(500);
} catch (Exception e2) {
// ignore
}
}
}
if (embedded == this) {
// connected to a server somewhere else
embedded = null;
} else {
// opened an embedded connection now -
// must connect to this database in server mode
// unfortunately
connectEmbeddedOrServer(true);
}
recreateSessionState();
if (eventListener != null) {
eventListener.setProgress(DatabaseEventListener.STATE_RECONNECTED,
databaseName, count, SysProperties.MAX_RECONNECT);
}
return true;
}
/**
* Check if this session is closed and throws an exception if so.
*
* @throws DbException if the session is closed
*/
public void checkClosed() {
if (isClosed()) {
throw DbException.get(ErrorCode.CONNECTION_BROKEN_1, "session closed");
}
}
@Override
public void close() {
RuntimeException closeError = null;
if (transferList != null) {
synchronized (this) {
for (Transfer transfer : transferList) {
try {
traceOperation("SESSION_CLOSE", 0);
transfer.writeInt(SessionRemote.SESSION_CLOSE);
done(transfer);
transfer.close();
} catch (RuntimeException e) {
trace.error(e, "close");
closeError = e;
} catch (Exception e) {
trace.error(e, "close");
}
}
}
transferList = null;
}
traceSystem.close();
if (embedded != null) {
embedded.close();
embedded = null;
}
if (closeError != null) {
throw closeError;
}
}
@Override
public Trace getTrace() {
return traceSystem.getTrace(Trace.JDBC);
}
public int getNextId() {
return nextId++;
}
public int getCurrentId() {
return nextId;
}
/**
* Called to flush the output after data has been sent to the server and
* just before receiving data. This method also reads the status code from
* the server and throws any exception the server sent.
*
* @param transfer the transfer object
* @throws DbException if the server sent an exception
* @throws IOException if there is a communication problem between client
* and server
*/
public void done(Transfer transfer) throws IOException {
transfer.flush();
int status = transfer.readInt();
if (status == STATUS_ERROR) {
String sqlstate = transfer.readString();
String message = transfer.readString();
String sql = transfer.readString();
int errorCode = transfer.readInt();
String stackTrace = transfer.readString();
JdbcSQLException s = new JdbcSQLException(message, sql, sqlstate,
errorCode, null, stackTrace);
if (errorCode == ErrorCode.CONNECTION_BROKEN_1) {
// allow re-connect
IOException e = new IOException(s.toString(), s);
throw e;
}
throw DbException.convert(s);
} else if (status == STATUS_CLOSED) {
transferList = null;
} else if (status == STATUS_OK_STATE_CHANGED) {
sessionStateChanged = true;
} else if (status == STATUS_OK) {
// ok
} else {
throw DbException.get(ErrorCode.CONNECTION_BROKEN_1,
"unexpected status " + status);
}
}
/**
* Returns true if the connection was opened in cluster mode.
*
* @return true if it is
*/
public boolean isClustered() {
return cluster;
}
@Override
public boolean isClosed() {
return transferList == null || transferList.size() == 0;
}
/**
* Write the operation to the trace system if debug trace is enabled.
*
* @param operation the operation performed
* @param id the id of the operation
*/
public void traceOperation(String operation, int id) {
if (trace.isDebugEnabled()) {
trace.debug("{0} {1}", operation, id);
}
}
@Override
public void checkPowerOff() {
// ok
}
@Override
public void checkWritingAllowed() {
// ok
}
@Override
public String getDatabasePath() {
return "";
}
@Override
public String getLobCompressionAlgorithm(int type) {
return null;
}
@Override
public int getMaxLengthInplaceLob() {
return SysProperties.LOB_CLIENT_MAX_SIZE_MEMORY;
}
@Override
public FileStore openFile(String name, String mode, boolean mustExist) {
if (mustExist && !FileUtils.exists(name)) {
throw DbException.get(ErrorCode.FILE_NOT_FOUND_1, name);
}
FileStore store;
if (cipher == null) {
store = FileStore.open(this, name, mode);
} else {
store = FileStore.open(this, name, mode, cipher, fileEncryptionKey, 0);
}
store.setCheckedWriting(false);
try {
store.init();
} catch (DbException e) {
store.closeSilently();
throw e;
}
return store;
}
@Override
public DataHandler getDataHandler() {
return this;
}
@Override
public Object getLobSyncObject() {
return lobSyncObject;
}
@Override
public SmallLRUCache<String, String[]> getLobFileListCache() {
return null;
}
public int getLastReconnect() {
return lastReconnect;
}
@Override
public TempFileDeleter getTempFileDeleter() {
if (tempFileDeleter == null) {
tempFileDeleter = TempFileDeleter.getInstance();
}
return tempFileDeleter;
}
@Override
public boolean isReconnectNeeded(boolean write) {
return false;
}
@Override
public SessionInterface reconnect(boolean write) {
return this;
}
@Override
public void afterWriting() {
// nothing to do
}
@Override
public LobStorageInterface getLobStorage() {
if (lobStorage == null) {
lobStorage = new LobStorageFrontend(this);
}
return lobStorage;
}
@Override
public synchronized int readLob(long lobId, byte[] hmac, long offset,
byte[] buff, int off, int length) {
checkClosed();
for (int i = 0, count = 0; i < transferList.size(); i++) {
Transfer transfer = transferList.get(i);
try {
traceOperation("LOB_READ", (int) lobId);
transfer.writeInt(SessionRemote.LOB_READ);
transfer.writeLong(lobId);
if (clientVersion >= Constants.TCP_PROTOCOL_VERSION_12) {
transfer.writeBytes(hmac);
}
transfer.writeLong(offset);
transfer.writeInt(length);
done(transfer);
length = transfer.readInt();
if (length <= 0) {
return length;
}
transfer.readBytes(buff, off, length);
return length;
} catch (IOException e) {
removeServer(e, i--, ++count);
}
}
return 1;
}
@Override
public JavaObjectSerializer getJavaObjectSerializer() {
initJavaObjectSerializer();
return javaObjectSerializer;
}
private void initJavaObjectSerializer() {
if (javaObjectSerializerInitialized) {
return;
}
synchronized (this) {
if (javaObjectSerializerInitialized) {
return;
}
String serializerFQN = readSerializationSettings();
if (serializerFQN != null) {
serializerFQN = serializerFQN.trim();
if (!serializerFQN.isEmpty() && !serializerFQN.equals("null")) {
try {
javaObjectSerializer = (JavaObjectSerializer) JdbcUtils
.loadUserClass(serializerFQN).newInstance();
} catch (Exception e) {
throw DbException.convert(e);
}
}
}
javaObjectSerializerInitialized = true;
}
}
/**
* Read the serializer name from the persistent database settings.
*
* @return the serializer
*/
private String readSerializationSettings() {
String javaObjectSerializerFQN = null;
CommandInterface ci = prepareCommand(
"SELECT VALUE FROM INFORMATION_SCHEMA.SETTINGS "+
" WHERE NAME='JAVA_OBJECT_SERIALIZER'", Integer.MAX_VALUE);
try {
ResultInterface result = ci.executeQuery(0, false);
if (result.next()) {
Value[] row = result.currentRow();
javaObjectSerializerFQN = row[0].getString();
}
} finally {
ci.close();
}
return javaObjectSerializerFQN;
}
@Override
public void addTemporaryLob(Value v) {
// do nothing
}
}
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论