sequel 5.28.0 → 5.33.0

data/doc/release_notes/4.46.0.txt

@@ -1,404 +0,0 @@
-= Deprecated Features
-
-* Symbol splitting is now deprecated by default.  Sequel has
-  split symbols since the very first version, but it has caused
-  many problems over the years and while terse, it isn't
-  intuitive to new Sequel users and causes significant
-  problems when using databases that use double/triple
-  underscores in identifiers.
-
-  If you are using symbols with embedded double/triple
-  underscores, such as:
-
-    :table__column
-    :column___alias
-    :table__column___alias
-
-  you either need to turn symbol splitting on by doing:
-
-    Sequel.split_symbols = true
-
-  or you need to convert the symbols to Sequel objects:
-
-    Sequel[:table][:column]
-    Sequel[:column].as(:alias)
-    Sequel[:table][:column].as(:alias)
-
-  Sequel ships with multiple extensions that make creation
-  of those Sequel objects less verbose, so consider using
-  the symbol_aref, symbol_aref_refinement, symbol_as, 
-  symbol_as_refinement, and/or s extensions.
-
-  To automatically convert symbols with double/triple
-  underscores to their Sequel object equivalents, you can
-  use the sequel-unsplit tool available at
-  https://github.com/jeremyevans/sequel-unsplit.
-
-  This deprecation also affects virtual row block methods that use
-  double underscores.  For example:
-
-    DB[:table].where{table__column > 3}
-
-  should be changed to:
-
-    DB[:table].where{table[:column] > 3}
-
-* Automatically treating plain strings passed to filtering/update
-  methods as literal strings has been deprecated, with support moved
-  to the auto_literal_strings extension.  The automatic conversion
-  of plain strings to literal SQL is the most common cause of
-  SQL injections in applications using Sequel, since many methods
-  pass their arguments down to the filtering methods, without
-  considering whether the argument might be a string derived from
-  user input.  By requiring explicit marking of literal SQL strings,
-  SQL injections are less likely and easier to audit for.
-  
-  This change means that unless you want to use the
-  auto_literal_strings extension, code such as:
-
-    DB[:table].where("a = 1")
-    DB[:table].where("a > ?", 1)
-
-  should to be converted to:
-
-    DB[:table].where(Sequel.lit("a = 1"))
-    DB[:table].where(Sequel.lit("a > ?", 1))
-
-  or even better, avoid literal SQL completely by converting it to
-  use equivalent Sequel expressions:
-
-    DB[:table].where(:a => 1)
-    DB[:table].where{a > 1}
-
-  This change also affects passing Dataset#update a string:
-
-    # Before
-    DB[:table].update("a = a + 1")
-
-    # Change to
-    DB[:table].update(Sequel.lit("a = a + 1"))
-    DB[:table].update(:a => Sequel[:a] + 1)
-
-  Note that this deprecation does not affect cases where literal
-  SQL is used for the entire query, such as when using any of the
-  following:
-
-    DB["SELECT * FROM foo"]
-    DB.fetch("SELECT * FROM foo WHERE a = ?", 1)
-    DB.dataset.with_sql("SELECT * FROM foo WHERE a = ?", 1)
-
-* Passing blocks to virtual row methods has been deprecated,
-  with support moved to the virtual_row_method_block
-  extension.  Historically, passing blocks to virtual row methods
-  changed how the methods were handled, but in recent years
-  alternative methods have been added to get the same results.
-  If you don't want to use the virtual_row_method_block extension,
-  conversion is fairly simple:
-
-    # WHERE a()
-    # Before
-    where{a{}}
-    # Change to
-    where{a.function}
-
-    # SELECT count(*)
-    # Before
-    select{count(:*){}}
-    # Change to
-    select{count.function.*}
-    
-    # SELECT count(DISTINCT c)
-    # Before
-    select{count(:distinct, :c){}}
-    # Change to
-    select{count(:c).distinct}
-    
-    # SELECT sum(c) OVER (PARTITION BY a)
-    # Before
-    select{count(:over, :args=>c, :partition=>:a){}}
-    # Change to
-    select{count(:c).over(:partition=>:a)}
-
-* Model.set_allowed_columns and Model#{set,update}_{all,only}
-  have been deprecated, with support moved to the
-  whitelist_security plugin.  These were the historical mass
-  assignment methods supported by Sequel, but set_fields and
-  update_fields have been recommended instead for many years.
-
-* Model.finder and .prepared_finder have been deprecated by default,
-  with support moved to the finder plugin.  Model.finder was
-  originally added to make it easy to create optimized finder
-  methods, but few Sequel users actually use it, so it makes more
-  sense to move it to a plugin.
-
-* Model.def_dataset_method and Model.subset have been deprecated
-  by default, with support moved to the def_dataset_method plugin.
-  It's been recommended for many years to use Model.dataset_module
-  to define dataset methods, instead of calling def_dataset_method
-  and subset on the model class.
-
-* Using ` in virtual rows to create literal SQL is now deprecated,
-  switch to using Sequel.lit instead:
-
-    # Before
-    DB[:table].where{`a = 1`}
-    # Change to
-    DB[:table].where(Sequel.lit('a = 1'))
-
-* Corner cases in argument handling in the filtering methods are now
-  deprecated, including:
-
-  * Ignoring a filtering method called without an argument or block.
-    In Sequel 5, this will raise an exception.
-
-  * Ignoring empty string arguments or other objects that respond to
-    empty? and return true.  In Sequel 5, only an empty array or hash
-    will be ignored.
-
-  * Ignoring an explicit nil argument when a block is passed.  In
-    Sequel 5, this will use a NULL filter.
-
-  * Ignoring an explicit nil argument when there is no existing
-    filter on the dataset. In Sequel 5, this will use a NULL
-    filter.
-
-* Using a joined dataset as a Sequel::Model dataset is now
-  deprecated.  Such datasets should now be wrapped in a subquery.
-  In Sequel 5, such datasets will automatically be wrapped in
-  a subquery aliased to the first table.
-
-    # Before
-    Model.dataset = DB[:a].join(:b, :id=>:b_id)
-    # Change to
-    Model.dataset = DB[:a].join(:b, :id=>:b_id).from_self(:alias=>:a)
-
-* Model.first_where has been deprecated, Model.first should be used
-  instead.
-
-* Database#log_yield is now deprecated.  This does not affect any of
-  the adapters that ship with Sequel, but external adapters that
-  have not yet been updated to support #log_connection_yield will need
-  to be updated.
-
-* The set_overrides extension is now deprecated.  Anyone using it
-  should consider supporting it as an external extension.
-
-* Many internal Database and Dataset regexp and string constants
-  that were previously used internally have been deprecated.
-  Additionally, some historical aliases for existing constants
-  have also been deprecated, such as Sequel::Schema::Generator.
-  Ruby 2.3+ is required to receive deprecation warnings related to
-  these constants.
-
-* Passing model classes as the first argument to Dataset#join_table
-  and Dataset#graph is now deprecated.  Pass the model's table name
-  or the model's datasets instead.
-
-* Passing model instances to Dataset#insert and #insert_sql is now
-  deprecated. Call values on the model instance to get the values
-  hash, and pass that as the argument instead.
-
-* Calling Dataset#set_graph_aliases before Dataset#graph is now
-  deprecated.  Dataset#set_graph_aliases should now be called
-  after Dataset#graph, not before.
-
-* The sequel/no_core_ext file is deprecated.  Sequel hasn't loaded
-  the core extensions by default since Sequel 3.  You can use the
-  following if you want to support both Sequel 3 and Sequel 5:
-
-    begin
-      require 'sequel/no_core_ext'
-    rescue LoadError
-      require 'sequel'
-    end
-
-* Database#pragma_get and #pragma_set on SQLite are now deprecated,
-  along with any method that calls them, such as auto_vacuum,
-  temp_store, foreign_keys, case_sensitive_like, synchronous, and
-  their setter methods.  To set these pragrams for all SQLite
-  database connections, the appropriate options should be passed
-  when creating the Database instance.
-
-* Automatically looking up the dataset class for a Database
-  instance by looking for a DatasetClass constant in the Database's
-  class is now deprecated.  All adapters that ship with Sequel have
-  been converted, but external adapters should now define the
-  Database#dataset_class_default private method appropriately
-  to return the correct dataset class.
-
-* Calling Model.db= on a model with a dataset is now deprecated.
-  If a model already has a dataset, you must now use set_dataset
-  or dataset= to change the dataset, not db=.
-
-* Sequel::SQL::Expression#sql_literal and #lit are now deprecated.
-  These aren't used internally and aren't expected to be used
-  externally.
-
-* {Integer,Timestamp}Migrator::DEFAULT_SCHEMA_{COLUMN,TABLE}
-  are now deprecated.  They have been replaced by
-  default_schema_column and default_schema_table instance methods.
-
-* Passing a Schema::CreateTableGenerator instance as the
-  second argument to Database#create_table is now
-  deprecated.  Database#create_table still supports passing the
-  generator via the :generator option.
-  
-* Passing a second argument to Database#alter_table is now
-  deprecated.
-
-* Sequel::BasicObject.remove_methods! is now deprecated. It has
-  always been a no-op on ruby 1.9+.
-
-* Referencing the PG_NAMED_TYPES constant in your code is now
-  deprecated.  Previously, adding entries to the PG_NAMED_TYPES
-  was deprecated, but no deprecation message would be issued by
-  referencing the constant.
-
-* The conversion of - to _ in adapter schemes is now deprecated.
-  This does not affect any internal adapters, but it may affect
-  external ones.
-
-* The Database#jdbc_* methods in the jdbc/db2 adapter (e.g.
-  jdbc_tables) are now deprecated.  Call the regular versions
-  instead (e.g. tables).
-
-* Dataset#_filter and #_filter_or_exclude private methods have
-  been deprecated.  If you have an extension that was calling these
-  methods, switch to the new #add_filter private method.
-
-= New Features
-
-* The class_table_inheritance plugin now supports an :alias option.
-  If provided, this wraps subclass datasets in subqueries, avoiding
-  problems with ambiguous columns and cases where the wrong table
-  name is used.  Due to the deprecation of joined datasets for
-  models, use of the class_table_inheritance plugin without this
-  :alias option will result in deprecation warnings.  In Sequel 5,
-  class_table_inheritance will default to using an :alias option
-  with the same as the name of the parent table.
-
-* The Dataset#sqltime_precision private method has been added.
-  Adapters can use override this if the precision for time values
-  is different from the precision for timestamp values. Sequel
-  uses this support on Microsoft SQL Server, so that time values
-  now support microsecond precision, instead of millisecond
-  precision.
-
-= Other Improvements
-
-* Sequel::Model classes that use a SQL::Identifier or
-  SQL::QualifiedIdentifier FROM table value will now use optimized
-  lookups and deletes, just as is done for those that use a
-  Symbol or String.
-
-* Dataset#simple_select_all? now handles aliased subqueries
-  correctly, returning false instead of true.
-
-* If Sequel.application_timezone is set to :utc,
-  Sequel::SQLTime.create will create instances using utc time
-  instead of local time.
-
-* If there is an exception while rolling back a transaction when
-  using the :rollback=>:always option, the exception is now
-  raised instead of being ignored.
-
-* If a migration file does not contain a migration or contains
-  multiple migrations, the exception raised will now include
-  the file name in the exception message.
-
-* In the jdbc/sqlserver adapter, time values with fractional
-  seconds and datetimeoffset values are now handled better when
-  using some versions of the underlying JDBC driver.
-
-* An additional disconnect error is recognized when using the
-  mysql and mysql2 adapters.
-
-* Dataset#full_text_search on Microsoft SQL Server now works
-  correctly if the no_auto_literal_strings extension is used.
-
-* Calling Database#disconnect when using the single connection
-  pool without an active connection works correctly again. It was
-  broken starting in 4.43.0 during changes to implement
-  Database#freeze.
-
-* Model class methods are no longer added for private methods
-  defined in a dataset_module block.  Previously, a public model
-  class method was defined, but it would raise an error when
-  called.
-
-* Fixnum is no longer referenced in the sqlanywhere shared
-  adapter, fixing deprecation warnings on ruby 2.4.
-
-* Sequel no longer uses constants for building SQL queries,
-  relying on frozen string literal support for better
-  performance on ruby 2.3+.  However, this decreases SQL
-  query building performance on ruby <2.3.  For the fastest
-  SQL query building, update to a recent version of ruby.
-
-* Sequel no longer ignores an empty object argument to a
-  filtering method if a block is provided.  Previously,
-  this could raise an exception or produce invalid SQL.
-
-* Many small modifications were made to reduce array
-  allocations, providing minor speedups.
-
-* Internal use of Array#at has been replaced with Array#[],
-  providing minor speedups on recent ruby versions.
-
-* The jdbc/db2 adapter no longer adds jdbc_* methods to
-  JDBC::Database.
-
-* Sequel no longer issues deprecation warnings on ruby 1.8.7.
-  Sequel 5 will drop support for ruby 1.8.7, and it doesn't make
-  sense to issue a deprecation warning if you couldn't upgrade
-  anyway.
-
-= Backwards Compatibility
-
-* When specifying the :fields option to a nested_attributes
-  setter, set_fields is now used internally instead of set_only.
-  set_fields has been recommended over set_fields since it's
-  introduction in Sequel 3.12, but nested_attributes was added
-  in Sequel 3.4, before set_fields was available.  The result
-  of this change is that if additional fields are provided that
-  do not match the fields in the :fields option, they will be
-  ignored instead of an exception being raised.
-
-* When specifying a function name using a Sequel::SQL::Identifier
-  instance, the function name is no longer quoted unless
-  Sequel::SQL::Function#quoted is used to create a quoted
-  function.  The reason for this is to make converting virtual
-  row method block code easier.
-
-    # Before
-    Sequel.function(Sequel[:a]) # "a"()
-
-    # Now
-    Sequel.function(Sequel[:a]) # a()
-    Sequel.function(Sequel[:a]).quoted # "a"()
-
-* When passing an SQL::PlaceholderLiteralString instance to a
-  dataset filtering method, the placeholder string is now always
-  wrapped in parentheses:
-
-    ds.where(Sequel.lit('? OR ?', :a, :b)).where(:c)
-    # Before: WHERE a OR b AND c
-    # Now: WHERE (a OR b) AND c
-
-  This is more of a bugfix than a backwards compatibility issue,
-  but is listed in the backwards compatibility section as there
-  may be applications that could break due to this change.
-
-* Model.subset now calls Model.dataset_module.subset, instead of
-  the other way around.  If your code depends on this, you will
-  need to make modifications.
-
-* The private Database#column_definition_order method no longer
-  uses const_get(:COLUMN_DEFINITION_ORDER).  External adapters
-  that defined COLUMN_DEFINITION_ORDER but did not override
-  this method must now override this method.
-
-* The private Database#native_function_name method no longer
-  uses const_get(:EMULATED_FUNCTION_MAP). External adapters
-  that defined EMULATED_FUNCTION_MAP but did not override
-  this method must now override this method.

data/doc/release_notes/4.47.0.txt

@@ -1,56 +0,0 @@
-= Deprecated Features
-
-* Setting an invalid dataset for a model is now deprecated.
-  Historically, Sequel has swallowed exceptions for this to keep
-  backwards compatibility, but it generally just results in code
-  breaking later.  To allow invalid datasets to be used:
-
-    Sequel::Model.require_valid_table = false
-
-* The association_autoreloading and many_to_one_pk_lookup plugins
-  are now deprecated.  They were moved from plugins to standard
-  model behavior in Sequel 4.0, and have been no-ops since.
-
-* The pg_typecast_on_load plugin is now deprecated.  It is only useful
-  on the already deprecated do and swift adapters.
-
-= New Features
-
-* Database#with_server in the server_block extension now accepts an
-  optional second argument for the read only server to use.  This
-  allows for overriding the default server while providing a separate
-  default for read only queries:
-
-    DB.with_server(:server1, :server1ro) do
-      DB[:a].all       # Uses server1ro
-      DB[:b].insert(1) # Uses server1
-    end
-
-* Model.default_association_type_options has been added, allowing the
-  ability to set default options per association type.  This can be
-  used to make some association types read_only by default:
-
-    opts = Sequel::Model.default_association_type_options
-    opts[:one_to_many] = opts[:many_to_many] = {:read_only=>true}
-  
-* Database#views on PostgreSQL now accepts a :materialized option to
-  return materialized views instead of regular views.
-
-= Other Improvements
-
-* Setting Sequel::Model.require_valid_table = true no longer raises
-  an exception when using a valid dataset that selects from a subquery
-  or table returning function or uses a join.
-
-* The defaults_setter plugin now inherits any custom default
-  values when subclassing.
-
-* The schema_dumper extension now handles Oracle 11g XE behavior of
-  appending not null to the database type.
-
-= Backwards Compatibility
-
-* External callers of Database#check_non_connection_error (private
-  method) should update their code to call it with a true or false
-  argument specifying whether to raise an error for exceptions that
-  are not connection errors.