pod4 0.10.6 → 1.0.0

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: pod4
 version: !ruby/object:Gem::Version
-  version: 0.10.6
+  version: 1.0.0
 platform: ruby
 authors:
 - Andy Jones
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2018-09-06 00:00:00.000000000 Z
+date: 2019-05-31 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: devnull
@@ -47,12 +47,11 @@ email:
 executables: []
 extensions: []
 extra_rdoc_files:
-- md/fixme.md
-- md/roadmap.md
-- md/typecasting.md
+- md/breaking_changes.md
 files:
 - ".bugs/bugs"
 - ".bugs/details/3979ce1679bc4f8c4aef8436e344e10e3773480d.txt"
+- ".bugs/details/b5368c7ef19065fc597b5692314da71772660963.txt"
 - ".hgignore"
 - ".hgtags"
 - ".rspec"
@@ -63,6 +62,8 @@ files:
 - lib/pod4.rb
 - lib/pod4/alert.rb
 - lib/pod4/basic_model.rb
+- lib/pod4/connection.rb
+- lib/pod4/connection_pool.rb
 - lib/pod4/encrypting.rb
 - lib/pod4/errors.rb
 - lib/pod4/interface.rb
@@ -75,17 +76,20 @@ files:
 - lib/pod4/sequel_interface.rb
 - lib/pod4/sql_helper.rb
 - lib/pod4/tds_interface.rb
+- lib/pod4/tweaking.rb
 - lib/pod4/typecasting.rb
 - lib/pod4/version.rb
-- md/fixme.md
-- md/roadmap.md
-- md/typecasting.md
+- md/breaking_changes.md
 - pod4.gemspec
 - spec/README.md
 - spec/common/alert_spec.rb
 - spec/common/basic_model_spec.rb
-- spec/common/model_new_validate_spec.rb
+- spec/common/connection_pool_parallelism_spec.rb
+- spec/common/connection_pool_spec.rb
+- spec/common/connection_spec.rb
+- spec/common/model_ai_missing_id_spec.rb
 - spec/common/model_plus_encrypting_spec.rb
+- spec/common/model_plus_tweaking_spec.rb
 - spec/common/model_plus_typecasting_spec.rb
 - spec/common/model_spec.rb
 - spec/common/nebulous_interface_spec.rb
@@ -129,7 +133,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
       version: '0'
 requirements: []
 rubyforge_project: 
-rubygems_version: 2.5.1
+rubygems_version: 2.7.6
 signing_key: 
 specification_version: 4
 summary: Totally not an ORM
@@ -137,8 +141,12 @@ test_files:
 - spec/README.md
 - spec/common/alert_spec.rb
 - spec/common/basic_model_spec.rb
-- spec/common/model_new_validate_spec.rb
+- spec/common/connection_pool_parallelism_spec.rb
+- spec/common/connection_pool_spec.rb
+- spec/common/connection_spec.rb
+- spec/common/model_ai_missing_id_spec.rb
 - spec/common/model_plus_encrypting_spec.rb
+- spec/common/model_plus_tweaking_spec.rb
 - spec/common/model_plus_typecasting_spec.rb
 - spec/common/model_spec.rb
 - spec/common/nebulous_interface_spec.rb

data/md/fixme.md

@@ -1,3 +0,0 @@
-Things I Wish I Could Do
-========================
-

data/md/roadmap.md

@@ -1,125 +0,0 @@
-Transactions
-============
-
-I'd like to support basic transactions because without it we can't really claim to do optimistic
-locking properly. And it would be nice to claim that.
-
-Thought I had a solid idea of how to do it without any faffing around, but, it had some holes. Will
-have to think again.
-
-But, this is top of my wish list.
-
-For the record, my current thinking: 
-
-    customer.new(4).read.or_die
-    customer.transaction do |c|
-      c.update(foo: 'bar')
-      c.orders.update(foo: 'bar')
-    end.or_die
-
-* a method supports_transactions() will control whether the interface does that. sql_helper will
-  define it to return true.
-
-* sql_helper will define a sql_transaction method which wraps SQL as a transaction.
-
-* interface methods create() delete() and update now accept an extra parameter, a boolean; if true,
-  they will return sql (and values), rather than doing anything. This is defined in Pod4::Interface.
-
-* BasicModel defines @in_transaction = false; @tx_sql = ""; @tx_vals = [].
-
-* When a Model is @in_transaction, the create, delete and update methods pass the extra parameter
-  to the corresponding interface methods. The results are accumulated in @tx_sql and @tx_vals.
-
-* the method BasicModel.transaction will:
-
-    * set interface.in_transaction = true, or raise an error if the interface doesn't support them
-    * yield a block passing the model instance so that the caller can run methods inside it
-    * set in_transaction back to false.
-    * call interface.executep( interface._sql_transaction( @tx_sql ), @tx_vals )
-
-Notes:
-
-* We will either have to standardize the execute method or check for it each time?
-* Ditto with executep. Ditto with whether an interface supports parameterisation.
-* trying to do a transaction across databases will fall over, but, really, no expectation there.
-* You can't have a transaction that uses the result of the first half to do the last half.
-* You can no longer call select() in a create, as we currently do for some interfaces? This is the
-  real problem I am wrestling with -- how to allow a transaction that returns a value from create().
-
-
-Migrations
-==========
-
-This will almost certainly be something crude -- since we don't really control the database
-connection in the same way as, say, ActiveRecord -- but I honestly think it's a worthwhile feature.
-Just having something that you can version control and run to update a data model is enough,
-really.
-
-I'm not yet sure of the least useless way to implement it.  Again, I favour SQL as the DSL.
-
-We will clearly need transactions first, though.
-
-My Current thoughts:
-
-* a migration against a database is on a par with a model but very different. You subclass
-  migration and give it an interface, pointing to the table that stores the current migration
-  state.  
-
-* The methods in a module are exectute() up() and down() -- the last two call the first one.
-
-* Each instance of a model is stored in a file and contains up and down SQL somehow. Each instance
-  has a version number.
-
-* You run a migration by running up or down on your migration class, passing a version?
-
-
-Connection Object
-=================
-
-PgInterface and TdsInterface both take a connection Hash, which is all very well, but it means that
-we are running one database connection per model.  Presumably this is a bad idea. :-)
-
-This actually hasn't come up in my own use of Pod4 -- for complex reasons I'm either using
-SequelInterface or running transient jobs which start up a couple of models, do some work, and then
-stop entirely.
-
-Connection is baked into those interfaces, and interface dependant. So I'm thinking in terms of a
-memoising object that stores the connection hash and then gets passed to the interface. When the
-interface wants a connection, then it asks the connection object. If the connection object doesn't
-have one, then the interface connects, and gives the connection to the connection object.
-
-It's looking as if we don't need this right now. One connection per model might not be as daft as
-it seems.
-
-
-JDBC-SQL interface
-==================
-
-For the jdbc-msssqlserver gem.  Doable ... I *think*.
-
-    driver = Java::com.microsoft.sqlserver.jdbc.SQLServerDriver.new
-    props = java.util.Properties.new
-    props.setProperty("user", "username")
-    props.setProperty("password", "password")
-    url = 'jdbc:sqlserver://servername;instanceName=instance;databaseName=DbName;'
-
-    conn = driver.connect(url, props)
-    #or maybe conn = driver.get_connection(url, "username", "password")
-
-    stmt = conn.create_statement
-    sql = %Q|blah;|
-
-    rsS = stmt.execute_query(sql)
-
-    while (rsS.next) do
-      veg = Hash.new
-      veg["vegName"] = rsS.getObject("name")
-      # etc
-    end
-
-    stmt.close
-    conn.close
-
-    see https://github.com/jruby/jruby/wiki/JDBC
-
-

data/md/typecasting.md

@@ -1,80 +0,0 @@
-TypeCasting
-===========
-
-Example
--------
-
-```
-require 'pod4'
-require 'pod4/someinterface'
-require 'pod4/typecasting'
-
-class Foo < Pod4::Model
-  include Pod4::TypeCasting
-
-  class Interface < Pod4::SomeInterface
-    # blah blah blah
-  end
-
-  set_interface Interface.new($stuff)
-
-  attr_columns :name, :issue, :created, :due, :last_update, :completed, :thing
-
-  # Now the meat
-  typecast :issue,         as: Integer
-  typecast :created, :due, as: Date
-  typecast :last_update,   as: Time
-  typecast :completed,     as: BigDecimal, ot_as: Float
-  typecast :thing,         use: mymethod
-end
-```
-
-What You Get
-------------
-
-### Every attribute named in a typecast gets:
-
-* An accessor. (Probably it already has one, if it is named in attr_columns, but it doesn't have to
-  be. Note, though, that we don't add the attribute to the column list and it does not get output
-  in to_ot by default.)
-
-* An attempt to force the value to that data type on set().  If the value cannot be coerced, it is
-  *untouched*. 
-
-* A second attempt to cast on to_interface(). This time, if the value cannot be coerced, it is set
-  to nil.
-
-* if the optional `ot_as` type is set, then we cast a third time in the `to_ot()` method;
-  additionally we guard the OT with the base type using Octothorpe.guard. Note that this only
-  effects to_ot().
-
-### Additionally the user can call these methods:
-
-* `typecast?(:columnname, value)` returns true if the value can be cast; value defaults to the
-  column value if not given.
-
-* `typecast(type, value, strict)` returns a typecast value, or either the original value, or nil if
-  strict is `:strict'.  
-
-* `guard(octothorpe)` will set guard conditions for nil values on the given octothorpe, based on
-  the attributes typecast knows about.
-
-### The following types will be supported:
-
-* Integer
-* BigDecimal
-* Float
-* Date
-* Time
-* :boolean
-
-Also: custom typecasting (`use: mymethod`, above). This must accept two parameters: the value, and
-an option hash.
-
-
-What You Don't Get
-------------------
-
-Validation. It's entirely up to you to decide how to validate and we won't second guess that.  But
-we do provide the `typecast?` method to help.
-

data/spec/common/model_new_validate_spec.rb

@@ -1,204 +0,0 @@
-require 'octothorpe'
-
-require 'pod4/model'
-require 'pod4/null_interface'
-
-
-##
-# This is purely here to test that model works when you have a validate that accepts the new
-# vmode parameter
-#
-describe 'Customer Model with new validate' do
-
-  let(:customer_model_class) do
-    Class.new Pod4::Model do
-      attr_columns :id, :name, :groups
-      attr_columns :price  # specifically testing multiple calls to attr_columns
-      set_interface NullInterface.new(:id, :name, :price, :groups, [])
-
-      def map_to_model(ot)
-        super
-        @groups = @groups ? @groups.split(',') : []
-        self
-      end
-
-      def map_to_interface
-        x = super
-        g = (x.>>.groups || []).join(',')
-        x.merge(groups: g)
-      end
-
-      def fake_an_alert(*args)
-        add_alert(*args) #private method
-      end
-
-      def validate(vmode)
-        add_alert(:error, "falling over for mode #{vmode}") if name == "fall over"
-      end
-
-      def reset_alerts; @alerts = []; end
-    end
-  end
-
-  let(:records) do
-    [ {id: 10, name: 'Gomez',     price: 1.23, groups: 'trains'       },
-      {id: 20, name: 'Morticia',  price: 2.34, groups: 'spanish'      },
-      {id: 30, name: 'Wednesday', price: 3.45, groups: 'school'       },
-      {id: 40, name: 'Pugsley',   price: 4.56, groups: 'trains,school'} ]
-
-  end
-
-  let(:records_as_ot)  { records.map{|r| Octothorpe.new(r) } }
-
-  # model is just a plain newly created object that you can call read on.
-  # model2 and model3 are in an identical state - they have been filled with a
-  # read(). We have two so that we can RSpec 'allow' on one and not the other.
-
-  let(:model) { customer_model_class.new(20) }
-
-  let(:model2) do
-    m = customer_model_class.new(30)
-
-    allow( m.interface ).to receive(:read).and_return( Octothorpe.new(records[2]) )
-    m.read.or_die
-  end
-
-  let(:model3) do
-    m = customer_model_class.new(40)
-
-    allow( m.interface ).to receive(:read).and_return( Octothorpe.new(records[3]) )
-    m.read.or_die
-  end
-
-  ##
-
-
-  describe '#create' do
-
-    let (:new_model) { customer_model_class.new }
-
-    it 'calls validate and passes the parameter' do
-      # validation tests arity of the validate method; rspec freaks out. So we can't 
-      # `expect( new_model ).to receive(:validate)`
-
-      m = customer_model_class.new
-      m.name = "fall over"
-      m.create
-      expect( m.model_status ).to eq :error
-      expect( m.alerts.map(&:message) ).to include( include "create" )
-    end
-
-    it 'calls create on the interface if the record is good' do
-      expect( customer_model_class.interface ).to receive(:create)
-      customer_model_class.new.create
-
-      new_model.fake_an_alert(:warning, :name, 'foo')
-      expect( new_model.interface ).to receive(:create)
-      new_model.create
-    end
-
-    it 'doesnt call create on the interface if the record is bad' do
-      new_model.fake_an_alert(:error, :name, 'foo')
-      expect( new_model.interface ).not_to receive(:create)
-      new_model.create
-    end
-
-  end
-  ##
-
-
-  describe '#read' do
-
-    it 'calls validate and passes the parameter' do
-      # again, because rspec is a bit stupid, we can't just `expect(model).to receive(:validate)`
-
-      allow( model.interface ).
-        to receive(:read).
-        and_return( records_as_ot.first.merge(name: "fall over") )
-
-      model.read
-      expect( model.model_status ).to eq :error
-      expect( model.alerts.map(&:message) ).to include( include "mode read" )
-    end
-
-  end
-  ##
-
-
-  describe '#update' do
-
-    before do
-      allow( model2.interface ).
-        to receive(:update).
-        and_return( model2.interface )
-
-    end
-
-    it 'calls validate and passes the parameter' do
-      # again, we can't `expect(model2).to receive(:validate)` because we're testing arity there
-      model2.name = "fall over"
-      model2.update
-      expect( model2.model_status ).to eq :error
-      expect( model2.alerts.map(&:message) ).to include( include "mode update" )
-    end
-
-    it 'calls update on the interface if the validation passes' do
-      expect( model3.interface ).
-        to receive(:update).
-        and_return( model3.interface )
-
-      model3.update
-    end
-
-    it 'doesn\'t call update on the interface if the validation fails' do
-      expect( model3.interface ).not_to receive(:update)
-
-      model3.name = "fall over"  # triggers validation
-      model3.update
-    end
-
-  end
-  ##
-
-
-  describe '#delete' do
-
-    before do
-      allow( model2.interface ).
-        to receive(:delete).
-        and_return( model2.interface )
-
-    end
-
-    it 'calls validate and passes the parameter' do
-      # again, because rspec can't cope with us testing arity in Pod4::Model, we can't say
-      # `expect(model2).to receive(:validate)`. But for delete we are only running validation as a
-      # courtesy -- a validation fail does not stop the delete, it just sets alerts. So the model
-      # status should be :deleted and not :error
-      model2.name = "fall over"
-      model2.delete
-      expect( model2.alerts.map(&:message) ).to include(include "mode delete")
-    end
-
-    it 'calls delete on the interface if the model status is good' do
-      expect( model3.interface ).
-        to receive(:delete).
-        and_return( model3.interface )
-
-      model3.delete 
-    end
-
-    it 'calls delete on the interface if the model status is bad' do
-      expect( model3.interface ).
-        to receive(:delete).
-        and_return( model3.interface )
-
-      model3.fake_an_alert(:error, :price, 'qar')
-      model3.delete 
-    end
-
-  end
-  ##
-
-end
-