JSON Parser + Boost Fusion + adapted structs
Hi everybody, I am writing a prototype for the JSON parser for the GSoC with the support of Bjorn Reese. When discussing the proposal, Michael Marcin suggested an interface that allows the user to define a struct reflecting the JSON data structure. I have been looking at Boost.Fusion to enable the parser to "fill in" the values in the struct. It looks like I'm after a solution described in the following post: http://stackoverflow.com/questions/13830792/simulate-compile-time-reflection... Could anyone tell me if this is a good starting point, and/or if there are other ways of achieving this. Thank you, Stephan.
Stephan Bourgeois
I have been looking at Boost.Fusion to enable the parser to
"fill in" the values in the struct.
It looks like I'm after a solution described in the following post: http://stackoverflow.com/questions/13830792/simulate-compile-time- reflection-in-c
I used BOOST_FUSION_ADAPT_STRUCT in conjuction with boost.serialization to automate serialization of messages between clients and server in my application. It just reflects simple C++ structs with data to XML/TEXT for pass them throughout network. After few experiments with RTTI, some kind of type erasure (with "variant" and "any" types), true object-oriented approach and boost.fusion I have found that the last one gives more benefits and fit in static type system of C++ more naturally.
Could anyone tell me if this is a good starting point, and/or if there are other ways of achieving this. Thank you, Stephan.
As alternate you look at implementation of serialization in Different ORM libraries, but they use external tools Message passing and RMI libraries like Apache.Thief or libcppa, but they use RTTI AFAIK. I think boost.fusion is the only one way to implement true compile-time reflection. With certain restrictions, of course.
This may be useful for your library. Suggestion: http://bloglitb.blogspot.com/2010/07/access-to-private-members-thats-easy.ht... Code sample: https://gist.github.com/1528856 Motivating discussion: http://thread.gmane.org/gmane.comp.lib.boost.devel/226883/focus=226889
Hi, A few things to note: The ISO C++ committee is preparing compile time reflection for C++. It could be worth to have a look at what they have in mind. Sadly the associated forum looks empty ( https://groups.google.com/a/isocpp.org/forum/?fromgroups#!forum/reflection) There's probably something that can be done with boost.spirit to write efficient parser/generators. I don't know how well they interface with Fusion. I personally like this implementation better than fusion (it would be a mistake to bring that in boost but it has one advantage: the macros are slightly less verbose as they don't require the types of the members, just their names) https://github.com/bytemaster/boost_reflect Then, there's boost.mirror that hasn't given much news recently, but that had one thing right: you need something to generate the metadata to avoid the pitfalls of writing the macro by hand. I had in mind a tool based on clang to do that but never had time to care about it since my very naive perl code was enough for my small needs. Finally, if you do that kind of effort, the applications will be far greater than JSON, it's definitely worth it ! Suggestion:
http://bloglitb.blogspot.com/**2010/07/access-to-private-** members-thats-easy.htmlhttp://bloglitb.blogspot.com/2010/07/access-to-private-members-thats-easy.ht...
I think that should not be a starting point at all. The types being parsed or serialized should be POD-like, ideally. It's much simpler to write code that deals with variables than with functions. If not, they should be meant to be serializable in their design. Then, if given no choice, with unmodifiable 3rd party code, we hack the privacy away. I could do without that last step. Regards, Julien
[Please do not mail me a copy of your followup]
boost@lists.boost.org spake the secret code
There's probably something that can be done with boost.spirit to write efficient parser/generators. I don't know how well they interface with Fusion.
They interface great; it's common practice to adapt a structure with BOOST_FUSION_ADAPT_STRUCT and have that tied into a Spirit grammar. -- "The Direct3D Graphics Pipeline" free book http://tinyurl.com/d3d-pipeline The Computer Graphics Museum http://computergraphicsmuseum.org The Terminals Wiki http://terminals.classiccmp.org Legalize Adulthood! (my blog) http://legalizeadulthood.wordpress.com
participants (5)
-
Julien Nitard -
legalize+jeeves＠mail.xmission.com -
Max Skvortsov -
Michael Marcin -
Stephan Bourgeois