mthiede / rgen Goto Github PK

Hi,
I had implemented for my application a version of eql? which takes in account attributes and relations. Would you be interested in integrating in rgen? In case I would send you a patch.

Hi,
I have a non containment relation:

class McWithNonContRef < RGen::MetamodelBuilder::MMBase
    has_one 'mcwta',McWithTwoAttrs
end

class McWithNonContRefMany < RGen::MetamodelBuilder::MMBase
    has_many 'mcwtas',McWithTwoAttrs
end

I add twice the same objects:

c1.addMcwtas(mcwta1)    
c1.addMcwtas(mcwta1)
assert_equal 2,c1.mcwtas.count

The assertion fails. Is this the supposed behavior? It seems quite surprising to me. While I understand that I could contain just once an object I don't see why I can not refer the same object twice.

Hi,
using RGen in a small script I found out that it is influenced by $VERBOSE.
I think this is a bad idea because that name could be potentially be used for something else (as it was in my case). I would suggest to rename it to $RGEN_VERBOSE.

In my case I got tons of lines like these:

[code]
/home/federico/.rvm/gems/jruby-1.7.4/gems/rgen-0.6.5/lib/rgen/metamodel_builder/builder_extensions.rb:265 warning: instance variable @abstract not initialized
(eval):8 warning: instance variable @name not initialized
(eval):7 warning: instance variable @abstract not initialized
(eval):7 warning: instance variable @interface not initialized
(eval):4 warning: instance variable @eStructuralFeatures not initialized
(eval):3 warning: instance variable @eStructuralFeatures not initialized
(eval):19 warning: instance variable @eContainingClass not initialized
(eval):19 warning: instance variable @eContainingClass not initialized
(eval):19 warning: instance variable @eContainingClass not initialized
(eval):19 warning: instance variable @eContainingClass not initialized
(eval):19 warning: instance variable @eContainingClass not initialized
(eval):19 warning: instance variable @eContainingClass not initialized
(eval):7 warning: instance variable @ePackage not initialized
(eval):4 warning: instance variable @eSuperTypes not initialized
(eval):3 warning: instance variable @eSuperTypes not initialized
(eval):7 warning: instance variable @instanceClassName not initialized
(eval):4 warning: instance variable @eAnnotations not initialized
(eval):7 warning: instance variable @changeable not initialized
(eval):7 warning: instance variable @defaultValueLiteral not initialized
(eval):7 warning: instance variable @derived not initialized
(eval):7 warning: instance variable @lowerbound not initialized
(eval):8 warning: instance variable @name not initialized
(eval):7 warning: instance variable @ordered not initialized
(eval):7 warning: instance variable @transient not initialized

[/code]

If it is not an error is issued at rgen/metamodel_builder/builder_runtime.rb (line 85)

targetId += "(" + target.name + ")" if target.respond_to?(:name) and target.name

what if you sum target.name.to_s instead of target.name?

The method RGen::MetamodelBuilder::BuilderRuntime#is_a? is currently where the CPU spends most of its time in some of our performance tests. I understand the need for overriding this very basic method to see the ClassModule type inheritance but perhaps there is less intrusive approach than to actually change the is_a? contract?

An alternative that I've experimented with is to implement a specific eIsA? method and then call that method instead when the ClassModule inheritance is relevant. In my experiment I changed the BuilderExtensions.type_check_code slightly so that it tests for eIsA? when the value is an MMBase but use is_a? when it's not. All tests are still green after changing 'test_inheritance' unit test to also use eIsA?. Our own performance tests show a significant improvement.

There are probably more places that needs changing in the Rgen code base (and more tests to write) and I'd be happy to submit a PR for this but I'd like to discuss it first.

Similar to #34 I see 0.10.0 and 0.10.1 are missing on Rubygems. I'm mostly interested in 9b7f42d.

As a related note: the homepage there also links to an expired domain. Please consider merging #25 and release 0.10.2.

The same kind of problem as fixed in #2 for :to_hash occurs for the method :to_str which is used internally buy other ruby methods.

As an example, when calling join on an array, containing a nested array, and that array is empty causes a failure.

x = []
['hello', x].join(':')

The problem with this is that it happens in 3d party logic. Just as for :to_hash, the expectancy is that Array should raise an error (NoSuchMethod).

This can be fixed, by treating :to_str the same way as :to_hash - in array_extensions.rb e.g. like this:

    return super unless (size == 0 &&
      !(m == :to_hash || m == :to_str)) ||
      compact.any?{|e| e.is_a? RGen::MetamodelBuilder::MMBase}

The following is more a question than an issue. I have the following situation: I have two meta-models, let's say "car.rb" and "wheel.rb". Now I would like to make within the "Car" ModelBuilder a reference to a Wheel created by the Wheel model builder. Is there any standard way? Note, in the Car metamodel I manually inserted the following:

require_relative 'wheel'
...
CarMetaModel::Car.contains_many_uni 'wheels', WheelMetaModel::Wheel

When instantiating a model from an xml file that includes a tag named File using DefaultXMLInstantiator, the program crashes. I believe this is due to line 56 in default_xml_instantiator.rb: mod.const_get(class_name).new. The method call: const_get will, by default, also searchn up the ancestral chain. When I set the inherit flag to false: mod.const_get(class_name, false).new, the model was instantiated correctly.

When serializing a model built with metamodel_builder, the result contains eSubTypes extension and the result is not XMI 2.0 compliant.

I could not figure out how to get rid of this since (I think) the transformer that is involved drags in ecore_ext which defines the super/sub types to be bidirectional.

To get past this problem I simply commented out the line while performing serialization.

Not sure if there is a way to load a model and serialize it to XMI without triggering the code path in question. If not, this is problem that is hard to work around.

The XMI20 outputs < as literal characters in the XML stream. This does not work well - they have to encoded as < or there will be problems reading it back in again.

(I worked around this problem by manually editing the XML output).

Hi,
on an instance of class RGen::MetamodelBuilder::MMBase (so an 'RGen eObject' I would say) there are methods to navigate children (direct children and indirect children)?
To get children of a given type?

Hi,
I had a problem with my code because a Metaclass was defining a reference named "value" while inheriting another reference with the same name.

I would suggest to issue a warning in such cases, the behavior is a bit confusing in my opinion.

When using attributes of Enum type, the created ECore model is wrong in that it always uses a type_name of "anonymous". This means that the model generated by metamodel builder must be patched after being created before the model can be used in EcoreToRuby transformation to again create the runtime classes.

It does not help to do the Enum creation outside of the modeled classes (i.e. assigning them to constants), patching is still required.

The package name is correct, but the type name is not.

Hi,
I noticed that there is the git tag v0.9.0 , but the latest release on rubygems is 0.8.2. Could you publish 0.9.0 on rubygems as well?

Trying to build on Fedora 21 with ruby 2.1.5, tests fail with:

Warning: you should require 'minitest/autorun' instead.
Warning: or add 'gem "minitest"' before 'require "minitest/autorun"'
From:
  /usr/share/ruby/test/unit.rb:1:in `<top (required)>'
  rgen_test.rb:3:in `<main>'
MiniTest::Unit::TestCase is now Minitest::Test. From /usr/share/ruby/test/unit/testcase.rb:8:in `<module:Unit>'
/usr/share/ruby/test/unit.rb:676:in `<class:Runner>': undefined method `_run_suite' for class `Test::Unit::Runner' (NameError)
        from /usr/share/ruby/test/unit.rb:261:in `<module:Unit>'
        from /usr/share/ruby/test/unit.rb:15:in `<module:Test>'
        from /usr/share/ruby/test/unit.rb:7:in `<top (required)>'
        from /usr/share/rubygems/rubygems/core_ext/kernel_require.rb:55:in `require'
        from /usr/share/rubygems/rubygems/core_ext/kernel_require.rb:55:in `require'
        from rgen_test.rb:3:in `<main>'

With ruby 2..2.1 fails with:

+ ruby rgen_test.rb
/usr/share/rubygems/rubygems/core_ext/kernel_require.rb:54:in `require': cannot load such file -- test/unit (LoadError)
        from /usr/share/rubygems/rubygems/core_ext/kernel_require.rb:54:in `require'
        from test/rgen_test.rb:3:in `<main>'

The instantiator can not handle references to external elements (i.e. elements which are not within the same document). In particular it can not resolve references to standard ECore datatypes like http://www.eclipse.org/emf/2002/Ecore#//EInt.

We see these test failures with Ruby 3.0

Finished in 0.617611s, 422.5960 runs/s, 2833.4978 assertions/s.

  1) Failure:
MetamodelFromEcoreTest#test_has_attr [/<<PKGBUILDDIR>>/test/metamodel_builder_test.rb:229]:
Expected /In (\w+::)+SimpleClass : Can not use a Symbol\(:someSymbol\) where a \[true,false\] is expected/ to match "In MetamodelFromEcoreTest::MetamodelFromEcore::SimpleClass : Can not use a Symbol(someSymbol)(:someSymbol) where a [true,false] is expected".

  2) Failure:
MetamodelFromEcoreTest#test_many_to_many [/<<PKGBUILDDIR>>/test/metamodel_builder_test.rb:619]:
Expected /In (\w+::)+BClassMM : Can not use a Symbol\(:notaaclass\) where a (\w+::)+AClassMM is expected/ to match "In MetamodelFromEcoreTest::MetamodelFromEcore::BClassMM : Can not use a Symbol(notaaclass)(:notaaclass) where a MetamodelFromEcoreTest::MetamodelFromEcore::AClassMM is expected".

  3) Failure:
MetamodelBuilderTest#test_many_to_many [/<<PKGBUILDDIR>>/test/metamodel_builder_test.rb:619]:
Expected /In (\w+::)+BClassMM : Can not use a Symbol\(:notaaclass\) where a (\w+::)+AClassMM is expected/ to match "In MetamodelBuilderTest::TestMetamodel::BClassMM : Can not use a Symbol(notaaclass)(:notaaclass) where a MetamodelBuilderTest::TestMetamodel::AClassMM is expected".

  4) Failure:
MetamodelBuilderTest#test_has_attr [/<<PKGBUILDDIR>>/test/metamodel_builder_test.rb:229]:
Expected /In (\w+::)+SimpleClass : Can not use a Symbol\(:someSymbol\) where a \[true,false\] is expected/ to match "In MetamodelBuilderTest::TestMetamodel::SimpleClass : Can not use a Symbol(someSymbol)(:someSymbol) where a [true,false] is expected".

261 runs, 1750 assertions, 4 failures, 0 errors, 0 skips
ERROR: Test "ruby3.0" failed. Exiting.

The tests won't run on 1.8.7 without this fix.