Datamapper DataMapper

Comments and changes to this ticket

• 

  Dan Kubb (dkubb) January 8th, 2009 @ 02:19 PM

  • State changed from “new” to “unconfirmed”
  • Assigned user cleared.
• 

  Michael Klishin (antares) January 9th, 2009 @ 06:05 PM

  • Assigned user set to “Michael Klishin (antares)”
  • Title changed from “Basic usage failure...” to “Valid parent entity is not saved though child entity has FK set (to nil obviously)”
• 

  Dirkjan Bussink January 10th, 2009 @ 02:19 AM

  • State changed from “unconfirmed” to “confirmed”

  I have experienced this issue in my own project just yesterday. I know work around it by explicitly setting the FK again just before saving, but that's pretty ugly.

• 

  Dan Kubb (dkubb) January 10th, 2009 @ 03:47 AM

  • Assigned user changed from “Michael Klishin (antares)” to “Dan Kubb (dkubb)”

  I can confirm this is still an issue in dkubb/dm-core, and it isn't necessarily related to validations, so I am going to assign this to myself.

• 

  Michael Klishin (antares) January 10th, 2009 @ 03:51 AM

  • Tag set to “bug, dm-core”

  I only picked this one because I have a friend who was bit by this issue and wanted to contribute a fix this weekend.

• 

  Dan Kubb (dkubb) January 10th, 2009 @ 03:54 AM

  As a side note, the reason this is a bug is because when assigning the Account to the User#account= mutator, it look at the current Account object's key and sets the corresponding FK inside User.

  The problem with this is that when the Account is actually saved it doesn't go out and update the FK originally set. The Account has no knowledge of what objects depend on it, and I'm not entirely sure we want to go down that route.. with observers and other things.

  What I'm thinking we should do instead is not set the FK at all on assignment. We should cache the Account object inside @account in the User object. Anytime User#account_id is called, it should return whatever @account.id does. The moment the @account object is updated, and we save the User object, @account.id will return the correct value, and things will be persisted as expected.

  We should also make it so that User#account_id= temporarily saves the value in @account_id, but when User#account is called, we initialize the Account instance and remove the @account_id. That way User#account and User#account_id will work as above, and we won't have any cases where @account.id != @account_id.

• 

  Dan Kubb (dkubb) January 11th, 2009 @ 02:14 AM

  • Assigned user changed from “Dan Kubb (dkubb)” to “Michael Klishin (antares)”

  @Michael: ahh ok, I'll assign it back to you. Please be aware though that in dkubb/dm-core relationships have been completely rewritten, so any fixes you make to sam/dm-core probably won't be carried forward into future releases.

  I tend to encourage people to make fixes to dkubb/dm-core, and backport anything to sam/dm-core that is critical, rather than directly in sam/dm-core. I am trying to keep sam/dm-core as close to frozen as possible, and only making critical fixes to it.

• 

  Dan Kubb (dkubb) January 11th, 2009 @ 02:15 AM

  @Michael: BTW: if you decide to make the fix in sam/dm-core, because you think this is critical, please re-assign this to me because I will have to make sure dkubb/dm-core is fixed as well.

• 

  Martin Gamsjaeger (snusnu) July 9th, 2009 @ 07:10 PM

  • State changed from “confirmed” to “resolved”

  As can be seen below, this is fixed in current datamapper/dm-core/next.

  require "rubygems"
  require "dm-core"
  require "dm-validations"
  require "spec"
  
  DataMapper::Logger.new(STDOUT, :debug)
  DataMapper.setup(:default, "sqlite3::memory:")
  
  class Account
    include DataMapper::Resource
    property :id, Serial
    property :name, String
  end
  
  class User
    include DataMapper::Resource
    property :id, Serial
    belongs_to :account
  end
  
  describe "saving m:1 relationships" do
    before(:all) do
      DataMapper.auto_migrate!
    end
    it "should save a valid parent" do
      a = Account.new(:name => "foo")
      u = User.new
      u.account = a
      a.save.should be_true
      u.save.should be_true
      
      Account.all.size.should == 1
      Account.get(a.id).name.should == 'foo'
      User.all.size.should == 1
    end
  end
  
  # mungo:snippets snusnu$ spec -cfs 619.rb 
  # 
  # saving m:1 relationships
  #  ~ (0.000143) SELECT sqlite_version(*)
  #  ~ (0.000110) DROP TABLE IF EXISTS "accounts"
  #  ~ (0.000026) DROP TABLE IF EXISTS "users"
  #  ~ (0.000023) PRAGMA table_info("accounts")
  #  ~ (0.000360) CREATE TABLE "accounts" ("id" INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT, "name" VARCHAR(50))
  #  ~ (0.000017) PRAGMA table_info("users")
  #  ~ (0.000113) CREATE TABLE "users" ("id" INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT, "account_id" INTEGER NOT NULL)
  #  ~ (0.000167) CREATE INDEX "index_users_account" ON "users" ("account_id")
  #  ~ (0.000037) INSERT INTO "accounts" ("name") VALUES ('foo')
  #  ~ (0.000057) INSERT INTO "users" ("account_id") VALUES (1)
  #  ~ (0.000037) SELECT "id", "name" FROM "accounts" ORDER BY "id"
  #  ~ (0.000024) SELECT "id", "name" FROM "accounts" WHERE "id" = 1
  #  ~ (0.000022) SELECT "id", "account_id" FROM "users" ORDER BY "id"
  # - should save a valid parent
  # 
  # Finished in 0.010888 seconds
  # 
  # 1 example, 0 failures