sequel 4.26.0 → 5.37.0

data/doc/release_notes/4.18.0.txt

@@ -1,36 +0,0 @@
-= New Features
-
-* An :auto_increment key has been added to the schema information for
-  primary key columns on JDBC, PostgreSQL, MySQL, MSSQL, DB2, and
-  SQLite.  This fixes issues in the schema_dumper extension where
-  non-auto-incrementing integer primary keys are no longer dumped as
-  auto-incrementing.
-
-  For adapters that don't have specific support for detecting
-  auto incrementing primary keys, Sequel now assumes a primary key
-  is auto incrementing only if it is not a composite primary key
-  and the type contains int (e.g. int, integer, bigint).
-
-= Other Improvements
-
-* Dataset#empty? now ignores any order on the dataset.  Previously,
-  calling empty? on a dataset ordered by an alias in the SELECT list
-  could raise an exception.
-
-* Schema qualified tables are now handled correctly in
-  many_through_many associations.
-
-* Using a hash as the value for the :eager association option now
-  works correctly.
-
-* All PG::ConnectionBad exceptions are now treated as disconnect
-  errors in the postgres adapter.  This should be more robust than
-  the previous method of trying to recognize disconnect errors by
-  trying to parse the exception message.
-
-* Sequel now skips a hash allocation when issuing queries through
-  datasets if sharding is not used.
-
-* Sequel no longer uses the JDBC schema parsing in the jdbc/sqlserver
-  adapter.  Instead, it uses the MSSQL schema parsing, which should
-  be more accurate than the generic JDBC schema parsing.

data/doc/release_notes/4.19.0.txt

@@ -1,45 +0,0 @@
-= New Features
-
-* Model#get_column_value and #set_column_value have been added for
-  getting/setting column values.  Historically, to get column
-  values, you would just send the column name, and to set column
-  values you would send the column name suffixed by =.  However,
-  this doesn't work when such methods are already defined by
-  ruby or Sequel itself (e.g. class, model, object_id).
-
-  Both #get_column_value and #set_column_value are just aliases to
-  #send, but you can safely override the methods to handle column
-  names that conflict with existing method names.  Both the core
-  model code and all of the plugins that ship with Sequel have
-  been updated to use these new methods.  External plugins are
-  strongly encouraged to switch to these new methods.
-
-* A column_conflicts plugin has been added to automatically handle
-  columns that conflict with existing method names.  So if you
-  have a column named "model" in your table, you can just load the
-  column_conflicts plugin and Sequel will handle things correctly.
-
-* A accessed_columns plugin has been added, which records which
-  columns have been accessed for a model instance.  This is useful
-  in development when you are planning on restricted the columns
-  selected by the dataset that retrieved the instance.  SELECTing
-  only the columns you need can result in significant performance
-  increases, and the accessed_columns plugin makes that easier.
-
-* Model#cancel_action has been added for canceling actions in
-  before hooks, instead of having the before hook methods return
-  false (which is still supported).  In addition to being easier to
-  use, this also makes it possible to use custom exception messages
-  for hook failures, if you are using the default behavior of raising
-  exceptions on save failures.
-
-= Other Improvements
-
-* Dataset#union, #intersect, and #except now automatically handle
-  datasets with raw SQL, by wrapping such datasets in subqueries.
-
-* The integer migrator now stores the correct migration number
-  when migrating with allow_missing_migration_files set.
-
-* A :timeout=>nil Database option on MySQL no longer sets a
-  wait_timeout.

data/doc/release_notes/4.2.0.txt

@@ -1,129 +0,0 @@
-= New Features
-
-* LATERAL subqueries are now supported on PostgreSQL 9.3+, HSQLDB,
-  and DB2 via Dataset#lateral:
-
-    DB.from(:a, DB[:b].where(:c=>:a__d).lateral)
-    # SELECT * FROM a,
-    # LATERAL (SELECT * FROM b WHERE (c = a.d)) AS t1
-
-  You can use a similar syntax when joining tables:
-
-    DB[:a].cross_join(DB[:b].where(:c=>:a__d).lateral)
-    # SELECT * FROM a
-    # CROSS JOIN LATERAL (SELECT * FROM b WHERE (c = a.d)) AS t1
-
-  If you are using Microsoft SQL Server, you can use the new
-  mssql_emulate_lateral_with_apply extension to emulate LATERAL
-  subqueries via CROSS/OUTER APPLY.
-
-* The static_cache plugin now supports a :frozen=>false option.  When
-  this option is used, instead of returning the frozen cached values,
-  the model now returns new, unfrozen objects that can be modified.
-  Note that if you make any database modifications, you are
-  responsible for updating the cache manually.
-
-* A pg_static_cache_updater extension has been added.  This extension
-  can automatically update the caches used by the static_cache plugin,
-  whenever the underlying database table is updated, using PostgreSQL's
-  notification channels.
-
-  This works by defining triggers on the underlying model tables that
-  use NOTIFY, and spinning up a thread in your application processes
-  that uses LISTEN, and refreshes the cache for the related model
-  whenever it receives a notification that the underlying table has
-  been modified.
-
-  This extension should make it possible to use the static_cache plugin
-  with the :frozen=>false option for any table that is small and not
-  frequently updated.
-
-* A from_block extension has been added that makes Database#from operate
-  like Dataset#from in regards to a passed block, allowing you to write
-  code like:
-
-    DB.from{table_returning_function(arg1, arg2)}
-
-* Database#supports_partial_indexes? has been added for checking for
-  partial index support.  Partial indexes are now supported on SQLite
-  3.8.0+.
-
-* A pg_loose_count extension has been added for fast approximate counts
-  of PostgreSQL tables.  This uses the system tables and should be
-  fairly accurate if the table statistics are up to date:
-
-    DB.loose_count(:table)
-
-* The Dataset#use_cursor method in the postgres adapter now supports
-  a :cursor_name option.  You can set this option if you want to
-  use nested cursors.
-
-* The mysql2 adapter now supports a :flags Database option allowing to
-  set custom mysql2 flags (e.g. ::Mysql2::Client::MULTI_STATEMENTS).
-
-= Other Improvements
-
-* Dataset#freeze has been implemented.  Previously, it was not
-  implemented, so Object#freeze was used, which resulted in a dataset
-  that wasn't cloneable.  Dataset#freeze now works as expected,
-  resulting in a cloneable dataset, but it doesn't allow methods to
-  be called that mutate the receiver.
-
-* Dataset#dup has been implemented.  Previously, it was not
-  implemented, so Object#dup was used, which resulted in a dataset
-  that shared an options hash with the receiver, so modifying the
-  dup's opts could also change the original dataset. Now dup works
-  similarly to clone, except that the returned object will not be
-  frozen.
-
-* Model#dup has been implemented.  Previously, it was not implemented,
-  so Object#dup was used, which resulted in a model instance that
-  shared the values hash with the receiver, so modifying the dup's
-  values also changed the original's values.  Now, dup does a shallow
-  copy of some of the internal data structures as well, so the copy
-  is more independent.
-
-  Note that you still need to be careful if you mutate objects:
-
-    m = Model.new(:a=>'a')
-    m2 = m.dup
-    m.a.gsub!('a', 'b') # also changes m2
-
-* Model#clone has been implemented.  Previously, it had the same
-  issues as dup.  Now, it calls the new Model#dup, but also
-  freezes the returned object if the receiver is frozen.
-
-* Placeholder literal strings with an empty parameter hash are now
-  handled correctly.
-
-= Backwards Compatibility
-
-* The static_cache plugin now disallows saving/destroying instances
-  unless the :frozen=>false option is used.  As the cached objects
-  returned by the model were frozen anyway, this affects creating
-  new instances or saving/destroying instances returned from the
-  underlying dataset.
-
-* Model#set_values has been removed (it was deprecated starting in
-  Sequel 4.0).
-
-* The following Model class methods are no longer defined:
-  insert_multiple, set, to_csv, paginate, query, set_overrides,
-  set_defaults.  By default, these methods used call the
-  dataset method of the same name, but as those methods are no
-  longer defined on datasets by default, they also resulted in a
-  NoMethodError.
-
-* Dataset#query!, #set_defaults!, and #set_overrides! are no longer
-  defined on all datasets.  They are now only defined on datasets
-  that use the query or set_overrides extensions.
-
-* Partial indexes are no longer returned by Database#indexes on MSSQL,
-  for consistency with PostgreSQL.  Note that the same change was
-  desired for SQLite, but SQLite currently does not offer reflection
-  support for determining which indexes are partial.
-
-* Database#foreign_key_list on MSSQL now will return a
-  SQL::QualifiedIdentifier instead of a symbol for the :table entry if
-  the schema of the referenced table does not match the schema of the
-  referencing table.

data/doc/release_notes/4.20.0.txt

@@ -1,79 +0,0 @@
-= New Features
-
-* A :before_retry option has been added to Database#transaction, which
-  specifies a proc to call when retrying if the :retry_on option
-  is used.  This can be used to implement additional logging, sleeping
-  between retries, or other things.
-
-* The to_json method :root option in the json_serializer plugin can now
-  be a string value to specify the name for the object key, instead of
-  using the underscored model name.
-
-* Dataset#paged_each now returns an enumerator if not passed a block.
-
-* You can now set the :instance_specific association option to false.
-  Previously, it was automatically set to true in some cases.  If you
-  know the association does not depend on anything instance-specific
-  other than the foreign/primary key, setting this option can allow
-  Sequel to perform some additional optimizations.
-
-= Other Improvements
-
-* Eager loading queries are now skipped if there are no matching keys.
-  There was code to check this previously, but it was accidently
-  removed in an earlier refactoring.
-
-* Eager loading an association with a limit and an eager block and
-  cascaded associations now works correctly when the window_function
-  limit strategy is used (the default on databases that support
-  window functions).
-
-* Eager loading an association with a limit with an eager block now
-  works correctly on databases do not support window functions but do
-  support correlated subqueries.
-
-* The parent association is now set on associated objects when loading
-  descendants in the rcte_tree plugin.  This allows the parent method
-  on any of the descendants to work without issuing a database query.
-
-* The prepared_statements_associations plugin now only uses prepared
-  statements if association metadata is being cached.  Previously, it
-  would use prepared statements even if association metadata was not
-  cached, which could leak the prepared statements.
-
-* Model#dup now duplicates the associations hash for the object.
-
-* Model#freeze no longer validates an object if the the errors for the
-  object are already frozen.  The static_cache plugin now freezes the
-  errors before freezing the object, so that it doesn't validate the
-  object.  This can skip many database queries when the
-  auto_validations plugin is used and there is a unique constraint or
-  index on the related table.
-
-* AUTOINCREMENT is now used again on SQLite by default for primary
-  keys.  It was removed when :auto_increment was added to the schema
-  hashes, but the removal changed SQLite's behavior.  This restores
-  the previous behavior.
-
-* Microsoft SQL Server's bit type is now recognized as a boolean type
-  by the schema dumper.
-
-* The pg_enum extension's create_enum method can now be used in
-  reversible migrations.
-
-* set_column_type with the :auto_increment=>true option once again
-  works on MySQL.  It had been broken since Sequel started adding
-  :auto_increment to the schema hashes.
-
-* The mysql2 adapter now recognizes the :charset option as a synonym
-  for :encoding.
-
-* The swift adapter now respects database and application timezone
-  settings.
-
-= Backwards Compatibility
-
-* AssociationReflection#apply_ruby_eager_limit_strategy no longer
-  checks that the strategy is :ruby, callers are now expected to
-  check the value themselves.  This should only matter if you are
-  using custom association types.

data/doc/release_notes/4.21.0.txt

@@ -1,94 +0,0 @@
-= New Features
-
-* SQL::GenericExpression#=~ has been added as an alternative method
-  of specifying equality/inclusion/identity.  Previously, you had to
-  use a hash.  This led to some slightly weird looking syntax when
-  used inside virtual rows:
-
-    DB[:items].where{{function(:column)=>0}}
-    # SELECT FROM items WHERE function(column) = 0
-
-  You can now use =~ as an equivalent:
-
-    DB[:items].where{function(:column) =~ 0}
-    # SELECT FROM items WHERE function(column) = 0
-
-  Like when using a hash, this works also for inclusion:
-
-    DB[:items].where{function(:column) =~ [1,2,3]}
-    # SELECT FROM items WHERE function(column) IN (1, 2, 3)
-
-  for identity:
-
-    DB[:items].where{function(:column) =~ nil}
-    # SELECT FROM items WHERE function(column) IS NULL
-
-  and for matching (on MySQL/PostgreSQL):
-
-    DB[:items].where{function(:column) =~ /foo/i}
-    # SELECT FROM items WHERE function(column) ~* 'foo'
-
-  This new syntax makes more complex conditions simpler to express:
-
-    DB[:items].where{(function(:column) =~ 0) | (column =~ 1)}
-    # SELECT FROM items WHERE function(column) = 0 OR column = 1
-
-  compared to previous versions of Sequel:
-
-    DB[:items].where{Sequel.|({function(:column) => 0}, {:column => 1})}
-
-  On ruby 1.9+, you can also use SQL::GenericExpression#!~ to invert
-  the condition:
-    
-    DB[:items].where{function(:column) !~ 0}
-    # SELECT FROM items WHERE function(column) != 0
-
-    DB[:items].where{function(:column) !~ [1,2,3]}
-    # SELECT FROM items WHERE function(column) NOT IN (1, 2, 3)
-
-    DB[:items].where{function(:column) !~ nil}
-    # SELECT FROM items WHERE function(column) IS NOT NULL
-
-    DB[:items].where{function(:column) !~ /foo/i}
-    # SELECT FROM items WHERE function(column) !~* 'foo'
-
-  This makes it simpler to write inverted conditions.  Ruby 1.8
-  doesn't support overriding the !~ method, but you can still use the
-  unary ~ method to invert:
-
-    DB[:items].where{~(function(:column) =~ 0)}
-
-* Database#add_named_conversion_proc has been added on PostgreSQL to
-  make it easier to add conversion procs by name instead of by OID:
-
-    DB.add_named_conversion_proc(:citext){|s| s}
-
-* Database#full_text_search on PostgreSQL now supports :tsquery and
-  :tsvector options for using existing tsquery and/or tsvector
-  arguments, instead of assuming the arguments are query terms or
-  the text to be search.
-
-= Other Improvements
-
-* Database#transaction now works inside after_commit and
-  after_rollback hooks.  Previously, it didn't work correctly as it
-  thought it was already inside the previously committed/rolled back
-  transaction.
-
-* Sequel.pg_jsonb now returns JSONBOp instances instead of JSONOp
-  instances when passed other than Array or Hash.
-
-* The tinytds adapter no longer tries to cancel a query on a closed
-  connection, which was causing an exception to be raised.
-
-= Backwards Compatibility
-
-* The default root name used in the JSON serializer is now demodulized
-  before being underscored.  This changes the behavior when the model
-  is namespaced.  For example, if the model class name is Mod::Model,
-  the previous default root name would be "mod/model", the new default
-  root name is "model".
-
-* If you were calling =~ or !~ on SQL::GenericExpression objects and
-  expecting the default ruby behavior of returning nil for =~ and
-  true for !~, you'll have to update your code.

data/doc/release_notes/4.22.0.txt

@@ -1,72 +0,0 @@
-= New Features
-
-* A csv_serializer plugin has been added, for serializing model
-  objects and datasets to CSV, or parsing CSV into a model
-  object or array of model objects.  Behavior and API is similar to
-  the existing xml_serializer and json_serializer plugins.
-
-* Sequel::MassAssignmentRestriction is now raised for mass assignment
-  errors in strict mode (the default).  Previously the generic
-  Sequel::Error was used.
-
-= Other Improvements
-
-* On Ruby 1.9+, Sequel now uses condition variables instead of busy
-  waiting for connections in the threaded conection pools.  This
-  can significantly decrease overhead when waiting for connections,
-  which can improve performance on machines that are compute
-  bottlenecked.  This also makes the connection pool checkouts more
-  fair, reducing the chance that a request for a connection will
-  fail with a PoolTimeout when under heavy resource contention.
- 
-* Sequel now attempts to avoid hash allocations and rehashing in
-  performance sensitive code.  This can speed up Dataset#clone,
-  Model#clone, and #Model#dup by about 60%, and speed up method
-  chains such as:
-
-    ds.select(:a).where(:a=>1).order(1)
-
-  by almost 20%.
-
-* Symbol#to_proc is used instead of explicit blocks across the
-  the library, which should improve performance slightly on Ruby 1.9+.
-
-* When Model#cancel_action is used in association before hooks,
-  Sequel will now return false if raise_on_save_failure = false,
-  instead of raising an exception.  This mirrors the behavior
-  when Model#cancel_action is used inside model save hooks when
-  raise_on_save_failure = false.
-
-* Dataset#to_hash and #to_hash_groups now work correctly on model
-  datasets when given a single array argument.
-
-* The auto_validations plugin now works correctly on columns that
-  have a default value, but where the default value is not parseable
-  into a ruby object by the adapter.
-
-* The tree plugin now correctly sets the reciprocal association
-  in the children association it creates.
-
-* In the pg_array extension, if the :default value when creating a
-  column is set to a ruby array, Sequel will now convert it to a
-  PostgreSQL array.
-
-* Sequel no longer adds a :max_length entry to the schema for
-  varchar(max) columns on Microsoft SQL Server.
-
-* Adapters now are specified to set the :default schema entry for
-  columns to nil if the adapter can determine the :default is nil.
-  Adapters that ship with Sequel already did this, but previously
-  it was unspecified behavior.
-
-* Sequel no longer silently ignores the :jdbc_properties Database
-  option in the jdbc adapter.  Previously, it only used the
-  :jdbc_properties option if it was not able to connect without it.
-
-* Bit types are now converted to boolean values in the ODBC adapter.
-
-= Backwards Compatibility
-
-* The db2, dbi, fdbsql, firebird, jdbc/fdbsql, informix, and openbase
-  adapters are now deprecated and will be removed in a future version
-  of Sequel.