Boost logo

Boost :

From: Dave Harris (brangdon_at_[hidden])
Date: 2002-12-16 17:27:37


In-Reply-To: <01C2A35A.28116D10_at_[hidden]>
On Sat, 14 Dec 2002 10:18:32 -0800 Robert Ramey (ramey_at_[hidden]) wrote:
> The fundamental problem that Serialization addresses is the reversable
> deconstruction of an arbitrary set of C++ objects to a stream of bytes.

I agree that will be the main reason to add Serialisation support to a
User-Defined Class.

> If you do have a requirement to generate a CSV file compatible for some
> particucular application (e.g. loading into an excel spreadsheet or
> into a database table) I can't imagine how a serialization library
> would offer benefit over the usage of iostreams directly.

Once I've added Serialisation support to my UDT, I want to get as much
value out of that code as possible. Adding a single CSV archive will
sometimes be less work than adding iostreams code to many UDTs.

Actually, I am more interested in XML than CSV. XML offers me enormous
benefit:

o Human readable output. Eg for debugging.
o Human editable input. Eg for preparing test cases, or
   patching data files.
o Platform neutral format.
o Email-safe format. Eg for discussing data in this boost
   list.
o Standard format. Eg 3rd party products understand it.

> This is a hugely ambitious undertaking that is only tenously related to
> serialization.

I don't think boost should take on every possible format. We should be
using this discussion time to establish which formats we should take on.
(I wish someone other than me would say something concrete about this.)
The remainder can be done as and when people need them.

Although a full validating XML parser is a big project, I don't think such
is necessary to deliver the benefits listed above. The scope is within
reason.

-- Dave Harris


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk