85 4 true 2008-06-10T18:54:40+01:00 22887 2009-01-31T00:00:00+00:00 383 activerecord-should-use-savepoints-for-nested-transactions 21 8994 committed activerecord nested savepoint transaction 2009-01-16T21:07:58+00:00 17202 Greg Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 From http://dev.rubyonrails.org/ticket/5457 in the old rails trac. Jonathan Viney has created a patch that enables doesn't break existing transaction semantics but allows for transactional behaviour to be tested in transactional fixtures. http://svn.viney.net.nz/things/rails/plugins/savepoints/ From http://dev.rubyonrails.org/ticket/5457 in the old rails trac. Jonathan Viney has created a patch that enables doesn't break existing transaction semantics but allows for transactional behaviour to be tested in transactional fixtures. http://svn.viney.net.nz/things/rails/plugins/savepoints/ <div><p> From <a href="http://dev.rubyonrails.org/ticket/5457">http://dev.rubyonrails.org/ticke...</a> in the old rails trac. </p><p> Jonathan Viney has created a patch that enables doesn't break existing transaction semantics but allows for transactional behaviour to be tested in transactional fixtures. </p><p> <a href="http://svn.viney.net.nz/things/rails/plugins/savepoints/">http://svn.viney.net.nz/things/r...</a> </p></div> 0 From http://dev.rubyonrails.org/ticket/5457 in the old rails trac. Jonathan Viney has created a patch that enables doesn't break existing transaction semantics but allows for transactional behaviour to be tested in transactional fixtures. http://svn.viney.net.nz/things/rails/plugins/savepoints/ <div><p> From <a href="http://dev.rubyonrails.org/ticket/5457">http://dev.rubyonrails.org/ticke...</a> in the old rails trac. </p><p> Jonathan Viney has created a patch that enables doesn't break existing transaction semantics but allows for transactional behaviour to be tested in transactional fixtures. </p><p> <a href="http://svn.viney.net.nz/things/rails/plugins/savepoints/">http://svn.viney.net.nz/things/r...</a> </p></div> false 2008-06-10T18:54:40+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 new 2008-06-10T18:54:40+01:00 22887 Sean Kirby Sean Kirby http://rails.lighthouseapp.com/projects/8994/tickets/383 85 0 I'd love to support savepoints. We've had the discussion regarding the default behavior of nested transactions, though, and it's not the way to go. Nested transactions should be explicit, otherwise we're going to flood the database with junk transactions that really only mean "include me in a transaction" not "start a new transaction". Let's extend the existing semantics instead of stepping on them. And add better test coverage ;) <div><p> I'd love to support savepoints. We've had the discussion regarding the default behavior of nested transactions, though, and it's not the way to go. Nested transactions should be explicit, otherwise we're going to flood the database with junk transactions that really only mean "include me in a transaction" not "start a new transaction". Let's extend the existing semantics instead of stepping on them. And add better test coverage ;) </p></div> false 2008-06-10T22:35:35+01:00 22887 --- :assigned_user: 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 new 2008-06-10T22:35:35+01:00 85 Jeremy Kemper Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 0 false 2008-06-10T22:35:49+01:00 22887 --- :milestone: 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 new 2008-06-10T22:35:49+01:00 85 Jeremy Kemper Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 0 Jeremy, the plugin by Jonathan Viney actually does make transaction{} only create a nested transaction if an explicit ":force => true" option is passed to it. The only exception being transactional tests where one additional level of non-explicit transactions can be created (because the test class itself uses the first level). So these changes should not change rails behavior outside tests. This plugin however does not include the ability to configure rails behavior for implicit transactions (those automatically created by save/destroy). Which means that you will not be able to make some activerecord model always create an implicit transaction for saving itself, even if a higher level transaction already exists. But this is a capability not strictly neccessary for the nested transactions feature to be useful, as you can always create an explicit "transaction :force => true" block around a save if you need it. <div><p> Jeremy, the plugin by Jonathan Viney actually does make transaction{} only create a nested transaction if an explicit ":force => true" option is passed to it. The only exception being transactional tests where one additional level of non-explicit transactions can be created (because the test class itself uses the first level). </p><p> So these changes should not change rails behavior outside tests. </p><p> This plugin however does not include the ability to configure rails behavior for implicit transactions (those automatically created by save/destroy). Which means that you will not be able to make some activerecord model always create an implicit transaction for saving itself, even if a higher level transaction already exists. But this is a capability not strictly neccessary for the nested transactions feature to be useful, as you can always create an explicit "transaction :force => true" block around a save if you need it. </p></div> false 2008-06-16T20:02:06+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 new 2008-06-16T20:02:06+01:00 14890 Tarmo Tänav Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 0 I'd like to see #533 included first (moving the transaction counter from Thread.current to the connection object), and then I'll add an updated patch for the savepoints. <div><p> I'd like to see <a href="/projects/8994/tickets/533" title="Ticket #533">#533</a> included first (moving the transaction counter from Thread.current to the connection object), and then I'll add an updated patch for the savepoints. </p></div> false 2008-07-03T13:12:57+01:00 22887 --- :tag: 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 new activerecord nested savepoint transaction 2008-07-03T13:12:57+01:00 17814 jonathan Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 0 #533 is in -- interested in porting your work over to Rails master? <div><p><a href="/projects/8994/tickets/533" title="Ticket #533">#533</a> is in -- interested in porting your work over to Rails master?</p></div> false 2008-08-30T05:10:22+01:00 22887 --- :state: new 383 activerecord-should-use-savepoints-for-nested-transactions-2 0 8994 open activerecord nested savepoint transaction 2008-08-30T05:10:22+01:00 85 Jeremy Kemper Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 0 I've attached a patch for master. The handling of increment/decrement_open_transactions has been moved to the connection object rather than handling it in AR::Base.transaction, seemed better to me. I've added a transactional_fixtures accessor to the connection which is set by the fixtures. Not the nicest, but I wasn't sure what else to do. There are some failures for databases that don't support savepoints (eg. sqlite). These are a bit hard to avoid unless we add a supports_savepoints? to the connection. Anyway, this should get the ball rolling. <div><p>I've attached a patch for master.</p> <p>The handling of increment/decrement_open_transactions has been moved to the connection object rather than handling it in AR::Base.transaction, seemed better to me.</p> <p>I've added a transactional_fixtures accessor to the connection which is set by the fixtures. Not the nicest, but I wasn't sure what else to do.</p> <p>There are some failures for databases that don't support savepoints (eg. sqlite). These are a bit hard to avoid unless we add a supports_savepoints? to the connection.</p> <p>Anyway, this should get the ball rolling.</p></div> false 2008-08-31T11:38:50+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions-3 0 8994 open activerecord nested savepoint transaction 2008-08-31T11:38:50+01:00 17814 jonathan Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 1 I think ':force => true' is a bit confusing. How about ':nested => true'? <div><p>I think ':force =&gt; true' is a bit confusing. How about ':nested =&gt; true'?</p></div> false 2008-10-01T15:19:19+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-10-01T15:19:22+01:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 1 savepoints-3.diff doesn't apply cleanly om master. Here's an updated patch, with conflicts resolved. It doesn't pass the unit tests however. More investigation will be required. <div><p>savepoints-3.diff doesn't apply cleanly om master. Here's an updated patch, with conflicts resolved.</p> <p>It doesn't pass the unit tests however. More investigation will be required.</p></div> false 2008-10-09T12:22:35+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-10-09T12:22:38+01:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 2 Jonathan, is there a reason why you changed def transaction(start_db_transaction = false) to def transaction(start_db_transaction = false) ? <div><p>Jonathan, is there a reason why you changed</p> <pre> <code>def transaction(start_db_transaction = false) </code> </pre> <p>to</p> <pre> <code>def transaction(start_db_transaction = false) </code> </pre> <p>?</p></div> false 2008-10-09T12:25:37+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-10-09T12:25:42+01:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 2 Never mind my last message, I already figured it out. :) Here's an updated patch which passes all unit tests, for MySQL, PostgreSQL and SQLite3. Changes are: * Much improved documentation, including documentation about potential caveats on MySQL. * :force option renamed to :nest, which is probably clearer. * I had to revert commit 045713ee240fff815edb5962b25d668512649478 by Jeremy Kemper, "PostgreSQL: introduce transaction_active? rather than tracking activity ourselves". It conflicts with the savepoint code. <div><p>Never mind my last message, I already figured it out. :)</p> <p>Here's an updated patch which passes all unit tests, for MySQL, PostgreSQL and SQLite3. Changes are:</p> <ul> <li>Much improved documentation, including documentation about potential caveats on MySQL.</li> <li>:force option renamed to :nest, which is probably clearer.</li> <li>I had to revert commit 045713ee240fff815edb5962b25d668512649478 by Jeremy Kemper, "PostgreSQL: introduce transaction_active? rather than tracking activity ourselves". It conflicts with the savepoint code.</li> </ul></div> false 2008-10-09T15:27:43+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-10-09T15:27:48+01:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 3 After speaking with Koz I've reimplemented Jeremy's transaction state introspection feature. The old implementation was incompatible with the old 'postgres' driver, which I fixed. I also added unit tests for this feature. <div><p>After speaking with Koz I've reimplemented Jeremy's transaction state introspection feature. The old implementation was incompatible with the old 'postgres' driver, which I fixed. I also added unit tests for this feature.</p></div> false 2008-10-09T17:00:49+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-10-09T17:00:56+01:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 Koz said that it's probably not feasible to include this feature in 2.2, seeing that we're in a feature freeze. If anybody wants to help, I'm maintaining this savepoint stuff in the following git repository: git://git.phusion.nl/rails.git branch 'savepoints' <div><p>Koz said that it's probably not feasible to include this feature in 2.2, seeing that we're in a feature freeze. If anybody wants to help, I'm maintaining this savepoint stuff in the following git repository:</p> <p>git://git.phusion.nl/rails.git branch 'savepoints'</p></div> false 2008-10-09T17:18:56+01:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-10-09T17:19:01+01:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 false 2008-10-17T17:14:35+01:00 22887 --- :milestone: 12671 9903 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-10-17T17:14:41+01:00 1366 Pratik Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 2.x 85 4 false 2008-12-03T19:08:41+00:00 22887 --- :milestone: 9903 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-12-03T19:08:45+00:00 141 Michael Koziarski Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 Is git://git.phusion.nl/rails.git still available? <div><p>Is git://git.phusion.nl/rails.git still available?</p></div> false 2008-12-09T00:00:53+00:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-12-09T00:00:57+00:00 85 Jeremy Kemper Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 Yes it is. I've just merged the patches against latest edge, and all unit tests pass. <div><p>Yes it is. I've just merged the patches against latest edge, and all unit tests pass.</p></div> false 2008-12-09T00:45:45+00:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-12-09T00:45:49+00:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 @@@ $ git remote add phusion git://git.phusion.nl/rails.git $ git fetch phusion fatal: The remote end hung up unexpectedly @@@ <div><pre><code> $ git remote add phusion git://git.phusion.nl/rails.git $ git fetch phusion fatal: The remote end hung up unexpectedly </code></pre></div> false 2008-12-09T02:40:54+00:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-12-09T02:40:56+00:00 85 Jeremy Kemper Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 Oops, a little git-daemon configuration problem, sorry. It should be up now. <div><p>Oops, a little git-daemon configuration problem, sorry. It should be up now.</p></div> false 2008-12-09T11:24:21+00:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2008-12-09T11:24:23+00:00 10679 Hongli Lai Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 @@@ ruby def transaction(start_db_transaction = false) start_db_transaction ||= open_transactions == 0 || (open_transactions == 1 && transactional_fixtures) @@@ This is an unexpected change of behavior: your transactions are nested in tests by default, but not in development or production without :nest => true. There should be a clearer distinction between wanting to participate in a transaction and requiring a new transaction boundary. To keep Base#transaction compatible with existing apps, we should :nest => true by default but disable for save/destroy callback transactions. <div><pre><code class="ruby"> def transaction(start_db_transaction = false) start_db_transaction ||= open_transactions == 0 || (open_transactions == 1 &amp;&amp; transactional_fixtures) </code></pre> <p>This is an unexpected change of behavior: your transactions are nested in tests by default, but not in development or production without :nest =&gt; true.</p> <p>There should be a clearer distinction between wanting to participate in a transaction and requiring a new transaction boundary.</p> <p>To keep Base#transaction compatible with existing apps, we should :nest =&gt; true by default but disable for save/destroy callback transactions.</p></div> false 2009-01-10T20:50:59+00:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 open activerecord nested savepoint transaction 2009-01-10T20:51:02+00:00 85 Jeremy Kemper Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 http://github.com/rails/rails/commit/ab0ce052ba23a4cce7a84ecade0d00d9cc518ebd <div><p><a href="http://github.com/rails/rails/commit/ab0ce052ba23a4cce7a84ecade0d00d9cc518ebd"> http://github.com/rails/rails/co...</a></p></div> true 2009-01-10T23:13:55+00:00 22887 --- :state: open 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 committed activerecord nested savepoint transaction 2009-01-10T23:13:59+00:00 85 Jeremy Kemper Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 (from [ab0ce052ba23a4cce7a84ecade0d00d9cc518ebd]) Introduce transaction_joinable flag to mark that the fixtures transaction can't joined, a new savepoint is required even if :requires_new is not set. Use :requires_new option instead of :nest. Update changelog. [#383 state:committed] http://github.com/rails/rails/commit/ab0ce052ba23a4cce7a84ecade0d00d9cc518ebd <div><p>(from [ab0ce052ba23a4cce7a84ecade0d00d9cc518ebd]) Introduce transaction_joinable flag to mark that the fixtures transaction can't joined, a new savepoint is required even if :requires_new is not set. Use :requires_new option instead of :nest. Update changelog.</p> <p>[<a href="/projects/8994/tickets/383" title="Ticket #383">#383</a> state:committed] <a href="http://github.com/rails/rails/commit/ab0ce052ba23a4cce7a84ecade0d00d9cc518ebd"> http://github.com/rails/rails/co...</a></p></div> true 2009-01-10T23:30:45+00:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 committed activerecord nested savepoint transaction 2009-01-10T23:30:46+00:00 17393 Repository Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 85 4 Can we ensure there's a way to easily debug issues one has when using the feature. I'm thinking of setting out the development database logs with the SQL trace in a fashion where it's easy to see what: (a) transaction it's part of (b) what save point section it's part of (if that makes sense) <div><p>Can we ensure there's a way to easily debug issues one has when using the feature. I'm thinking of setting out the development database logs with the SQL trace in a fashion where it's easy to see what: (a) transaction it's part of (b) what save point section it's part of (if that makes sense)</p></div> true 2009-01-16T21:07:53+00:00 22887 --- {} 383 activerecord-should-use-savepoints-for-nested-transactions 0 8994 committed activerecord nested savepoint transaction 2009-01-16T21:07:58+00:00 17202 Greg Sean Kirby Jeremy Kemper http://rails.lighthouseapp.com/projects/8994/tickets/383 c4a2d136fdcaa16b98fdec8ede0e52d21be74f5d text/plain 2008-08-31T11:38:50+01:00 savepoints-3.diff 42811 14985 17814 http://rails.lighthouseapp.com/attachments/42811/savepoints-3.diff b8c27ff0dcc30f77e4005e625b54915f1d22a605 text/plain 2008-10-09T12:22:36+01:00 savepoints-4.diff 52735 14940 10679 http://rails.lighthouseapp.com/attachments/52735/savepoints-4.diff 5b7410738809c89635bd4e0ebeccdad07913c4bb text/plain 2008-10-09T15:27:43+01:00 savepoints-5.diff 52774 39922 10679 http://rails.lighthouseapp.com/attachments/52774/savepoints-5.diff 8b78d68d0069d2e783e5972c0a39a46c40cf5f92 text/plain 2008-10-09T17:00:49+01:00 savepoints-6.diff 52798 45558 10679 http://rails.lighthouseapp.com/attachments/52798/savepoints-6.diff