Add support for null values to AnyMap

[pelliott-mathworks]

rapidjson supports null values but AnyMap did not. This change adds support for null values to the AnyMap class. Signed-off-by: The MathWorks, Inc. pelliott@mathworks.com

[codecov-commenter]

Codecov Report

Merging #694 (ca793ee) into development (c4a8d62) will increase coverage by 0.08%.
The diff coverage is 25.00%.

@@               Coverage Diff               @@
##           development     #694      +/-   ##
===============================================
+ Coverage        90.32%   90.41%   +0.08%     
===============================================
  Files              249      248       -1     
  Lines            13629    13636       +7     
===============================================
+ Hits             12311    12329      +18     
+ Misses            1318     1307      -11     

Impacted Files	Coverage Δ
framework/include/cppmicroservices/AnyMap.h	52.38% <0.00%> (-20.96%)	⬇️
framework/src/bundle/BundleManifest.cpp	86.71% <100.00%> (+0.21%)	⬆️
...clarativeServices/src/manager/ReferenceManager.hpp	87.50% <0.00%> (-12.50%)	⬇️
...nfigurationAdmin/src/ConfigurationAdminPrivate.hpp	80.00% <0.00%> (-11.67%)	⬇️
framework/include/cppmicroservices/detail/Log.h	97.14% <0.00%> (-2.86%)	⬇️
.../cppmicroservices/detail/ServiceTrackerPrivate.tpp	81.69% <0.00%> (-1.17%)	⬇️
framework/src/bundle/BundleResource.cpp	95.10% <0.00%> (-0.42%)	⬇️
framework/src/util/Properties.h	100.00% <0.00%> (ø)
framework/include/cppmicroservices/ServiceEvent.h	100.00% <0.00%> (ø)
...ramework/include/cppmicroservices/BundleResource.h	100.00% <0.00%> (ø)
... and 11 more

[jeffdiclemente]

should we have a function in the Any class to test is an Any is NullValue?

Currently, I don't think Any::Empty() and Any::ToJSON() are in alignment with this new concept of NullValue. There might also be other Any member functions which are not consistent with NullValue.

  /**
   * returns true if the Any is empty
   */
  bool Empty() const { return !_content; }


  /**
   * Returns a JSON representation for the content.
   *
   * Custom types should specialize the any_value_to_json template function for meaningful output.
   * The values of increment and indent are passed around to be able to be used for nicer
   * formatting. The code that makes use of this is in the any_value_to_json specializations for the
   * various containers.
   *
   * To get pretty output, simply pass a value greater than zero in as the first argument of ToJSON
   * and the rest of the code will take care of things.
   *
   * @param increment The amount of extra indentation to add for each level of JSON. An increment of
   *                  zero indicates no special formatting
   * @param indent    The current amount of indent to apply to the current line. 
   */
  std::string ToJSON(const uint8_t increment, const int32_t indent) const
  {
    return Empty() ? "null" : _content->ToJSON(increment, indent);
  }

[pelliott-mathworks]

Do you think Empty should return true if the value is null?

[jeffdiclemente]

given that we have a user-defined type, NullValue, representing a JSON null, I think the answer is no.

I also think ToJSON returning "null" may be confusing with a ToJSON representation of JSON null

[jeffdiclemente]

I would like to see test coverage for NullValue within the core framework's unit tests, i.e. within framework\test\gtest\AnyTest.cpp. Take a look at the existing tests in here for guidance on what to test. off the top of my head, testing Any::Type() for NullValue should be done.