From ajuckel at gmail.com  Sat Oct  2 01:49:13 2004
From: ajuckel at gmail.com (Anthony Juckel)
Date: Fri, 1 Oct 2004 20:49:13 -0500
Subject: [cl-prevalence-devel] Patch to serialization.lisp for sbcl
Message-ID: <a45b92ac041001184956b9c23e@mail.gmail.com>

I was having some problems getting objects to properly serialize when
using cl-prevalence, until I found that there was no implementation of
serializable-slots for sbcl.  This patch contains such an
implementation.  Also, I've modified it somewhat from the formula
followed by the other lisps.  Rather than simply returning all slots
for the object, I only return those slots which are currently bound
for the object.  I leave it up to you to decide whether this is a
useful addition or not.

Anthony W. Juckel
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cl-prevalence.patch
Type: application/octet-stream
Size: 1064 bytes
Desc: not available
URL: <https://mailman.common-lisp.net/pipermail/cl-prevalence-devel/attachments/20041001/7798b4fc/attachment.obj>

From randall at randallsquared.com  Sat Oct  2 08:14:19 2004
From: randall at randallsquared.com (Randall Randall)
Date: Sat, 2 Oct 2004 04:14:19 -0400
Subject: [cl-prevalence-devel] Patch to serialization.lisp for sbcl
In-Reply-To: <a45b92ac041001184956b9c23e@mail.gmail.com>
References: <a45b92ac041001184956b9c23e@mail.gmail.com>
Message-ID: <0F08E248-144B-11D9-A072-000A95A0F1E8@randallsquared.com>

On Oct 1, 2004, at 9:49 PM, Anthony Juckel wrote:
> implementation.  Also, I've modified it somewhat from the formula
> followed by the other lisps.  Rather than simply returning all slots
> for the object, I only return those slots which are currently bound
> for the object.  I leave it up to you to decide whether this is a
> useful addition or not.

I was planning to add (if Sven doesn't object) some
sort of unbound slot handling (since I would expect
unbound slots to remain unbound after unserialization,
and currently the serialization code just dies if a
slot is unbound.

So I agree that something needs to be done about
unbound slots, but I don't agree that ignoring them
is the best solution.

--
Randall Randall <randall at randallsquared.com>
"And no practical definition of freedom would be complete
  without the freedom to take the consequences. Indeed, it
  is the freedom upon which all the others are based."
  - Terry Pratchett, _Going Postal_




From scaekenberghe at common-lisp.net  Mon Oct  4 15:16:27 2004
From: scaekenberghe at common-lisp.net (Sven Van Caekenberghe)
Date: Mon, 4 Oct 2004 17:16:27 +0200
Subject: [cl-prevalence-devel] Patches to serialization
Message-ID: <5CE204E2-1618-11D9-8598-000A95B3E0D2@common-lisp.net>

Hi all,

The following patches were applied to serialization in CVS (not yet the 
release):

1

bugfix: serialization of an empty string is <string></string> but
deserialization of that XML construct was NIL which is not a string
we now return an empty string in that case
thanks to randall randall

2

added serializable-slots implementation for sbcl
thanks to Anthony W. Juckel

3

added code to deal with unbound slots when serializing (noted by 
anthony juckel)
and code to deal with missing slots when deserializing

4

added reset-class-slots

5

now testing xml and sexp serialization seperately, added empty string 
test
thanks to randall randall

Sven




From scaekenberghe at common-lisp.net  Tue Oct  5 11:54:34 2004
From: scaekenberghe at common-lisp.net (Sven Van Caekenberghe)
Date: Tue, 5 Oct 2004 13:54:34 +0200
Subject: [cl-prevalence-devel] Indexes
Message-ID: <5313453A-16C5-11D9-8598-000A95B3E0D2@common-lisp.net>

Hi All,

Randall Randall contributed some code to add indexes to 
managed-prevalence,
today I merge his code in CVS head (not yet in the release tarball)

Exported functions:

index-on
drop-index-on
find-object-with-slot

Comments:

The ideas and most code comes from Randall,
I added code to drop indexes,
added code to add existing objects to newly created indexes,
added a fallback for find-object-with-slot in case there is no index,
changed the code that checked for unchanged slots so that it conses 
less.

Todo:

Since indexing was added to improve performance,
I think current the generation of root and index names done each time
we access a root or index is too slow (especially conses too much).

I have a new scheme of root naming that I am thinking about that might 
solve this problem.
It might also allow for iteration over all current indexes of a class 
which could be usefull when deleting objects,
and might bring back the index->object association instead of the 
current index->id association
which is currently adding another indirection.

Sven