[boost] [gsoc2008] YAML archives for Boost.serialization

From: Esteve Fernandez (esteve_at_[hidden])
Date: 2008-03-31 15:08:28

Hi all, it seems that the period for student applications has been extended to 
April the 7th. Although I briefly discussed the idea for YAML archives 
following a message by Sohail, I would like to explain it in more detail, 
discuss some concerns and see if there's interest.

We already talked about YAML not being a superset of JSON, however given the 
recent great news of the relicensing of TinyJSON, it still can be used as a 
basis to build YAML parser.

My main concern is that YAML is sensitive to indentation, actually it uses 
indentation as a substitute for { block }. One can write YAML to look like 
JSON, use TinyJSON and everything will be fine, but it will not be able to 
parse documents that use spaces/tabs to delimit blocks, only those that use 
curly brackets. I guess it will have to be implemented at the scanner level 
in Boost.Spirit, but it becomes more complicated. Anyway, with more 
complications, comes more fun :-) I might be mistaken about where to 
implement this stuff so feel free to correct me, but I think it's the way to 
go, as I took a look at the source code for the Python interpreter and for 
LibYAML, and they implemented the indentation stuff in their scanner, not in 
the parser.

The good thing about doing YAML instead of JSON is that all the concerns about 
object tracking and class versioning disappear, as they're already 
implemented in the official spec. JSPON might or not be added to the official 
JSON spec, although there are good chances that it will be, but it's still 
not officially accepted and may change in the future. Also, a parser needs to 
implement JSONPath (an XPath-like language for querying JSON documents), 
since a reference to the root object can be written as "$ref" : "$[0].child"

A side advantage of implementing the YAML archives is that a C++ parser for 
YAML would be available for those who don't want to use the official LibYAML 
C parser.

Anyway, what do you think? Is there interest for this project?

Cheers.