ee7abd8acc621807e5128269db269b5a32218fbc Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/10f69a2bd22e39ba03f4687af196cd476e88d162 10f69a2bd22e39ba03f4687af196cd476e88d162

2009-05-04T09:44:29-07:00

Bump version to 3.0.0 and add release notes ca615016f597f14b3224800ab09378887fd15f28 Jeremy Evans code@jeremyevans.net 82483c0b3c13fe4843ce7b94afb2e1e33273a21f Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/ee7abd8acc621807e5128269db269b5a32218fbc ee7abd8acc621807e5128269db269b5a32218fbc

2009-05-01T14:48:01-07:00

Remove dead threads from connection pool if the pool is full and a connection is requested Thanks to a report on IRC by Crash2burn, I was able to track down this bug. The bug is related to Sequel's connection pool's lack of understanding of ruby's threading implementation. Sequel's connection pool was apparently designed with the idea that using ensure will ensure that the code is run. However, that isn't true in all cases. For example, if Thread#exit! is called, the thread terminates immediately without running any ensure blocks. One would assume that nobody would use Thread#exit! or a similar method without a very good reason (looks like both jruby and ruby 1.9 don't support it at all), but apparently if ruby is used with many threads, it will call that or something similar. I was able to recreate the issue using the following code: ds = DB[:table] 1000.times{Thread.new{sleep 1; rand > 0.5 ? ds.insert : ds.first}} Basically, what happens is that the connection pool's @allocated hash is left with dead threads. These should be released by the ensure block, but apparently that isn't always run. The fix is fairly simple. If a connection request comes in and the pool is full, check the @allocated hash for dead threads and release any that are dead. This approach is lazy in that it doesn't release dead thread connections right away. However, it will only keep the dead thread around if you aren't fully utilizing the pool. As soon as you start to fully utilize the pool, it will clean up the garbage. I expect this has a modest performance impact, but I don't think it can be done faster and still be correct. Send me a patch if I'm wrong. There's no change to the single threaded pool, as this situation will only happen in multithreaded code. While here, stop keeping track of the connection counts separately. I'm not sure of the performance impact of that change, but it simplifies the code and makes future bugs less likely. This also moves the acquire call inside the begin/ensure block in #hold, which should fix a problem if a thread died directly after requiring a connection but before the begin block was entered. This also changes #release so that it doesn't take a connection object, and it now assumes the mutex is already in use. I could make a similar change to #remove, but there isn't any reason to at this point. 374abb23c90fbce5fc2337d45b42841bdf6192c4 Jeremy Evans code@jeremyevans.net 75921495c06011ed7eb3c14912c0c4698e7a5bba Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/82483c0b3c13fe4843ce7b94afb2e1e33273a21f 82483c0b3c13fe4843ce7b94afb2e1e33273a21f

2009-05-01T14:29:02-07:00

Add a failing connection pool spec to test Thread#exit! inside ConnectionPool#hold cdb0000e060d4afee120b8996cae6f0c9a2304c6 Jeremy Evans code@jeremyevans.net d5b2b5d53b92094e8f267f9c268ccd5bb3cd6792 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/75921495c06011ed7eb3c14912c0c4698e7a5bba 75921495c06011ed7eb3c14912c0c4698e7a5bba

2009-04-30T08:15:08-07:00

Add autoincrementing primary key support in the Oracle adapter, using a sequence and trigger This support is based on the Firebird adapter, so thank you wishdev for that. The Oracle adapter still needs quite a bit more work to be fully supported, though hopefully that'll come in the next few releases. 63d765e7a0504dbfedfec5e9b5e9790bc91c9c89 Jeremy Evans code@jeremyevans.net 048c7e66d142d34c475b6d557f13d3dac3d0288e Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/d5b2b5d53b92094e8f267f9c268ccd5bb3cd6792 d5b2b5d53b92094e8f267f9c268ccd5bb3cd6792

2009-04-29T10:05:37-07:00

Fix spec issue in the STI specs due to the Model#save change 7d47bb86e03154e29006db279b82e6b18fad4d97 Jeremy Evans code@jeremyevans.net 51be000c7c529b6ccf91238ead1d3efdcc15cc48 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/048c7e66d142d34c475b6d557f13d3dac3d0288e 048c7e66d142d34c475b6d557f13d3dac3d0288e

2009-04-29T10:05:05-07:00

Make the schema sort the indexes in name order when dumping them, for consistency e40cd9e2a3db708ef56566c6569b5599d9a574bb Jeremy Evans code@jeremyevans.net 30af82cebec03657ce750c5f5188a627b56c990d Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/51be000c7c529b6ccf91238ead1d3efdcc15cc48 51be000c7c529b6ccf91238ead1d3efdcc15cc48

2009-04-29T09:43:45-07:00

Make Model#save use the same server it uses for saving as for retrieving the saved record This change should only affect those who are using Sequel's master/slave database functionality with Models. Because Model#save runs in a transaction by default, and because Model#refresh does a select and Model#save does an insert for new records, before, Model#save would insert the record on the :default server and then refresh. Since refresh does a read, it operates on the :read_only server by default. Because it was inside a transaction, the changes from the master weren't available on the slave yet. Anyway, the fix is to use the :default server for refreshes that are called inside of save. 8dcfa20069e708a24648863989b1219c20fa3970 Jeremy Evans code@jeremyevans.net f9224af0582759bc40db90853662cc56c3a04194 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/30af82cebec03657ce750c5f5188a627b56c990d 30af82cebec03657ce750c5f5188a627b56c990d

2009-04-27T16:40:42-07:00

Add Database#database_type method, for identifying which type of database the object is connecting to The RDoc explains the reasoning: The database type for this database object, the same as the adapter scheme by default. Should be overridden in adapters (especially shared adapters) to be the correct type, so that even if two separate Database objects are using different adapters you can tell that they are using the same database type. Even better, you can tell that two Database objects that are using the same adapter are connecting to different database types (think JDBC or DataObjects). 83c72fa5caafc29a03d5873171b402fd27010c43 Jeremy Evans code@jeremyevans.net 7ba72fc1b0f7b07ce326e6acdbd8746830215cfa Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/f9224af0582759bc40db90853662cc56c3a04194 f9224af0582759bc40db90853662cc56c3a04194

2009-04-27T16:02:42-07:00

Make private select_*_sql methods only take a single argument Now that select_sql doesn't take any options, we can streamline the code so that instead of passing the @opts instance variable to every method as an argument, we just refer to it inside the methods. This is probably a little faster based on a few spec runs that I did, but the main reason I'm doing this is that the code is cleaner. a7d4e304fa6b63a21e50a68a48f6fd7ec621e187 Jeremy Evans code@jeremyevans.net fd0400ee75278630c6cab34ad688b503ba811d88 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/7ba72fc1b0f7b07ce326e6acdbd8746830215cfa 7ba72fc1b0f7b07ce326e6acdbd8746830215cfa

2009-04-27T14:50:18-07:00

Add ability to reset primary key sequences in the PostgreSQL adapter This adds the reset_primary_key_sequence Database method to the PostgreSQL adapter. It's based on the code in ActiveRecord. This commit also fixes a bug where sequences were quoted twice instead of once if they weren't simple lower case names. This changes the code to not use extra quotes, since PostgreSQL adds them anyway. 7eb7a557124bc27895575e3673f1fb4d80a68360 Jeremy Evans code@jeremyevans.net 7183329a93cc1dbb9af6ef676877279af21c3e52 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/fd0400ee75278630c6cab34ad688b503ba811d88 fd0400ee75278630c6cab34ad688b503ba811d88

2009-04-27T14:15:28-07:00

Allow schema_dumper extension to dump indexes separate from columns You can now use the dump_indexes_migration method to dump just indexes in a separate migration. This is helpful when cloning databases, as it is faster to transfer all data first and then add indexes. This includes some small refactoring in dump_table_schema. It also modified Generator#dump_indexes to accept :add_index and :drop_index methods. This also documents the :indexes option to dump_schema_migration. 291a95082f152222a3dcfea4decec11ffd24888e Jeremy Evans code@jeremyevans.net 81f828a05e1bb26b59dec17e98fc95ec7f7f66e9 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/7183329a93cc1dbb9af6ef676877279af21c3e52 7183329a93cc1dbb9af6ef676877279af21c3e52

2009-04-27T13:38:55-07:00

Allow schema_dumper extension to use an :indexes=>false option not to include indexes e508c5b176d6fb59e3e5900e2fdc46fb18ed4743 Jeremy Evans code@jeremyevans.net ccb0f62f70386cbf929d11b9a4b04f4d4b4726fa Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/81f828a05e1bb26b59dec17e98fc95ec7f7f66e9 81f828a05e1bb26b59dec17e98fc95ec7f7f66e9

2009-04-27T13:07:22-07:00

Fix issues with the index parsing The JDBC adapter was completely broken in the last commit, fix that. The MySQL adapter's indexes method was broken, fix that as well. The SQLite adapter's indexes method was broken on JDBC, handle that. Add tests to make sure primary key indexes are not included, and modify the SQLite, JDBC, and H2 adapters to make sure of that. beebba592bdffb9e34703978af57b09649e4fcb9 Jeremy Evans code@jeremyevans.net 54fbbd5a564c619af33191536d8344b96c8aa67f Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/ccb0f62f70386cbf929d11b9a4b04f4d4b4726fa ccb0f62f70386cbf929d11b9a4b04f4d4b4726fa

2009-04-27T12:26:59-07:00

Add index parsing to PostgreSQL, MySQL, SQLite, and JDBC adapters, support dumping indexes This commit adds a Database#indexes method to the above adapters, which share the following API (from the RDoc): Return a hash containing index information. Hash keys are index name symbols. Values are subhashes with two keys, :columns and :unique. The value of :columns is an array of symbols of column names. The value of :unique is true or false depending on if the index is unique. Here's an example of use: DB.indexes(:songs) => {:songs_name_index=>{:unique=>true, :columns=>[:name]}, :songs_lyricid_index=>{:unique=>false, :columns=>[:lyricid]}} This information is used by the schema_dumper extension to dump indexes when dumping tables. There will be an option later to skip the dumping of indexes. This also changes the schema_dumper extension to dump composite primary key constraints, or really any primary key constraints that isn't an autoincrementing primary key. While here, I'm refactoring the tables and schema_parse_table methods to make them more DRY. To that end I'm adding a metadata_dataset private method that uses the default identifier input and output methods, as well as input_identifier_meth and output_identifier_meth, both of which return method objects. The point of all this is so that things don't break when a user uses an input or output identifier method different from the default. 8abd376c78b6c3d9a9c8bf7d75bf53795b919025 Jeremy Evans code@jeremyevans.net bb5b3b19212fb93d72954e89314ba5f93cbf8d5e Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/54fbbd5a564c619af33191536d8344b96c8aa67f 54fbbd5a564c619af33191536d8344b96c8aa67f

2009-04-26T20:30:05-07:00

Correctly quote SQL Array references, and handle qualified identifiers with them (e.g. :table__column.sql_subscript(1)) I didn't notice that this didn't work as I would expect until I had to use it to implement the PostgreSQL index parsing code (which will be coming in the near future). It does break backwards compatibility for users who were creating SQL::Subscripts directly via: Sequel::SQL::Subscript.new("column", [1]) Before, the SQL would be column[1], now it is 'column'[1]. You should either use a Symbol or a LiteralString instead now. 3740998e83ac18732c69b5e087c039b75ec0d9da Jeremy Evans code@jeremyevans.net d9295db936fa0dc56d1061c8e081999141344aa4 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/bb5b3b19212fb93d72954e89314ba5f93cbf8d5e bb5b3b19212fb93d72954e89314ba5f93cbf8d5e

2009-04-26T20:25:05-07:00

Allow dropping an index with a name different than the default name I'm not sure why this wasn't supported before. Sequel let's you choose a specific name for an index, but before you couldn't use that name when dropping the index. Anyway, use the :name option to drop_index to drop an index with a specific name. 9d7134fe7e55fed3cc2a14ce94e7ba3f33ecdfd5 Jeremy Evans code@jeremyevans.net bf0ff5840bc3b6377db03030656742313b21190c Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/d9295db936fa0dc56d1061c8e081999141344aa4 d9295db936fa0dc56d1061c8e081999141344aa4

2009-04-26T18:09:08-07:00

Allow Dataset#from to remove existing FROM tables when called without an argument, instead of raising an error later This change is backwards incompatible, but previously there was no way to remove all FROM tables. I chose to backwards compatibility and make from consistent with select, in that still works if you call it without an argument. I thought about adding a method similar to select_all, but couldn't think of a good name (from_none was my original choice, but I don't really like that). 54dbaa2bc42fd906330cfa8a2b08357159c76d3b Jeremy Evans code@jeremyevans.net 7aad75df8bc4e692290733afc878e8f705d96f79 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/bf0ff5840bc3b6377db03030656742313b21190c bf0ff5840bc3b6377db03030656742313b21190c

2009-04-23T15:37:42-07:00

Fix string quoting on Oracle so it doesn't double backslashes ba613d6eb73e9f76c41018499d93c955513d1f0f Jeremy Evans code@jeremyevans.net f9c4ee8871d422f07bc161bf24f0f8accfd94c3b Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/7aad75df8bc4e692290733afc878e8f705d96f79 7aad75df8bc4e692290733afc878e8f705d96f79

2009-04-22T09:14:26-07:00

2009-04-22T09:03:28-07:00

Fix spec breakage caused by alias use in Dataset#count 4cc1346c1c9f3d81cab307345e56559cc4919a35 Jeremy Evans code@jeremyevans.net cf2fa139d32bb64ffa90afca34b10319985070d3 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/f9c4ee8871d422f07bc161bf24f0f8accfd94c3b f9c4ee8871d422f07bc161bf24f0f8accfd94c3b

2009-04-21T21:23:32-07:00

Alias the count function call in Dataset#count, fixes use on MSSQL b2b57b16a877855cb3b8c354485bf23f9f19207d Jeremy Evans code@jeremyevans.net e010b007b508306090679e2a940fbec1e05ee98c Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/cf2fa139d32bb64ffa90afca34b10319985070d3 cf2fa139d32bb64ffa90afca34b10319985070d3

2009-04-21T21:17:40-07:00

Allow QualifiedIdentifiers to be qualified, to allow :column.qualify(:table).qualify(:schema) 500ee34c087bbff28b1719a41d106978e0d080f2 Jeremy Evans code@jeremyevans.net 214c278ba03aecb4c81ea937c5f75eb383b5f380 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/e010b007b508306090679e2a940fbec1e05ee98c e010b007b508306090679e2a940fbec1e05ee98c

2009-04-21T17:52:28-07:00

Update CHANGELOG for akitaonrails's DBI change a5874383956ea8e02874ae5920d2051f7f5536dd Jeremy Evans code@jeremyevans.net 2186724561eeb6f2476997e9555121099ace9aec Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/214c278ba03aecb4c81ea937c5f75eb383b5f380 214c278ba03aecb4c81ea937c5f75eb383b5f380

2009-04-21T17:51:01-07:00

Revert MSSQL quoted_identifier change 3f8ca12f8b89a7bb57dac9d210a97737491cd59e Jeremy Evans code@jeremyevans.net 0ca44c253a81928282b6df8092e21b0d74faac89 109071f8d88f091a467a2d01d307062688336f3f Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/2186724561eeb6f2476997e9555121099ace9aec 2186724561eeb6f2476997e9555121099ace9aec

2009-04-21T17:48:41-07:00

Merge commit 'akitaonrails/master' 43e1dbc584f246c413d9438e9c989578036a48d4 Jeremy Evans code@jeremyevans.net 63d10b7436eab8c134077ac5509cf7cf456e366c AkitaOnRails fabioakita@gmail.com http://github.com/jeremyevans/sequel/commit/109071f8d88f091a467a2d01d307062688336f3f 109071f8d88f091a467a2d01d307062688336f3f

2009-04-21T16:44:00-07:00

Taking care of correctly quoting table names such as Person.Address into [Person].[Address] in MSSQL Also making sure the :db_type => 'mssql' properly include the adapters/shared/mssql for DBI af31d327b78b7b399a85c5130b535818db8139a9 AkitaOnRails fabioakita@gmail.com d6a973b5e62ba61824f6e92d3e469df6d192280b Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/0ca44c253a81928282b6df8092e21b0d74faac89 0ca44c253a81928282b6df8092e21b0d74faac89

2009-04-21T16:24:24-07:00

Add schema_dumper extension, for dumping schema of tables This adds three functions to Schema::Generator: dump_columns, dump_constraints, and dump_indexes. The expectation is that if you instance eval each inside of another instance, that instance will be a copy of current instance. This adds a Database#dump_table_schema method, which returns a string representing a create_table method call that will recreate the table. It isn't perfect, but it recognizes primary keys, defaults, and NULL/NOT NULL determinations. It also is able to parse database types back into ruby types, and use those. It converts types by default, so if you wanted to copy the table schema from a PostgreSQL database to an SQLite database, you could use it. If you are copying to the same database type, you can use the :same_db option to not convert database types. The conversion from database types to ruby types is lossy. Most of the time you will get the same or similar type, but if you have a single precision float, it'll probably give you a double precision float. And if you have a smallint, you'll end up with a regular integer. These cases could be dealt with, but I'd like to keep the code as simple as possible. A Database#dump_schema_migration method was also added, that can be used to dump the entire database as a single migration. Currently this only supports dumping columns. It will be expanded to dump indexes as well soon. 4e5e77871dc9a11ea353555acd564a7a96865c52 Jeremy Evans code@jeremyevans.net ad2ef833bce8c18630cb8ca2e019218dd9943ec0 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/d6a973b5e62ba61824f6e92d3e469df6d192280b d6a973b5e62ba61824f6e92d3e469df6d192280b

2009-04-21T15:56:55-07:00

Add spec for invalid class used as Database column 2fcb23d73ffaae1e69177190db2af98688e02eed Jeremy Evans code@jeremyevans.net d0d48614f1576d13fea63730fd7a6a7a93e2e132 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/ad2ef833bce8c18630cb8ca2e019218dd9943ec0 ad2ef833bce8c18630cb8ca2e019218dd9943ec0

2009-04-21T15:47:56-07:00

2009-04-21T15:47:41-07:00

Add Database#schema_autoincrementing_primary_key? for doing a guess at whether the column schema could be an autoincrementing primary key f4df0faea6e4cf05e779672fcf8e6076d2be5c35 Jeremy Evans code@jeremyevans.net c59e955a862a71848c1f5ee77b05b3da2cca0257 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/d0d48614f1576d13fea63730fd7a6a7a93e2e132 d0d48614f1576d13fea63730fd7a6a7a93e2e132

2009-04-21T15:46:02-07:00

Make sure the eval in the schema generator contains file and line info 7e8f43e6893f76411d74ba74a0b3db2767beb3c6 Jeremy Evans code@jeremyevans.net 4711520cd89ae08cedec723c8fa0cba775b47569 Jeremy Evans code@jeremyevans.net http://github.com/jeremyevans/sequel/commit/c59e955a862a71848c1f5ee77b05b3da2cca0257 c59e955a862a71848c1f5ee77b05b3da2cca0257

2009-04-21T15:44:10-07:00

Get 100% spec coverage for plugins and extensions This commits removes the update_values definition in the caching plugin, since it was removed from Sequel::Model. 8226bf6238e6c7b4841a2bef3afb80f7542c358a Jeremy Evans code@jeremyevans.net