Added a visitor version of unpack API,

[redboltz]

The current unpacking APIs are constructed on the visitor mechanism.
(Fixed #418)

[redboltz]

There is no significant difference in performance.

[jpetso]

Since this test is also listed on the wiki documentation page, I'll point out two potential issues:

• end_map_value() doesn't append a comma to separate map items like end_array_item() does.
• visit_str() doesn't process strings as Unicode, therefore non-ASCII codepoints will result in invalid JSON (which encodes them as e.g. "\u00f8").

Neither of these is required by the test case so that's okay for this particular piece of code, however you might want to point it out in the documentation.

[redboltz]

Thank you for the comments.

end_map_value() doesn't append a comma to separate map items like end_array_item() does.

I will fix it.

visit_str() doesn't process strings as Unicode, therefore non-ASCII codepoints will result in invalid JSON (which encodes them as e.g. "\u00f8").

True. The point of the test and the document is not accurate JSON, I just demonstrate a visitor usage. However, I should use a precise term. I will fix the test and the document. I will use the term 'json like structure ( string is not espaced)'.

[jpetso]

One more thing to consider: When I was having a look at https://github.com/msgpack/msgpack-c/wiki/v2_0_cpp_unpack_visit, I realized that this code isn't actually unpacking the msgpack buffer, at least not in the traditional sense of unpack as "parse buffer into msgpack::object".

Since the function name (unpack_visit()) is already not the same as the traditional unpack(), I wonder whether it makes sense to give it a name that's less tied to the traditional meaning of "unpack". Maybe parse()? I'm interested to hear about your thoughts on this idea :)

[redboltz]

Good point. unpack_visit() doesn't unpack anything. I think that parse() is good too. Another candidate is traverse(). In this context, parse() is better. I will rename to parse().

[redboltz]

I've fixed all problems that you pointed out. In addition, I found end_map_value() calling point was wrong, and fixed it. https://github.com/redboltz/msgpack-c/blob/add_unpack_visitor_api/include/msgpack/v2/unpack.hpp#L434

[redboltz]

@jpetso , could you review my fix? If my fix would be ok, I will merge it.

[redboltz]

I am going to merge the PR this weekend if no comments are post.

[jpetso]

Hi, sorry, and well done! I had a pretty tight few weeks and didn't manage to find the time for the review. Since the first version of the patch was already very good, I'm assuming it's quite alright :)

Anyway, I'm glad it's in, and thank you for your hard work!

[redboltz]

@jpetso , thank you!