Re: Automatic linking to a mhonarc archive

[Earl Hood <earl@earlhood.com>]

On August 14, 2002 at 00:26, Alan Burlison wrote:

> I've done this, but I don't get a callback to mhonarc::CBMessageHeadRead if
> a message is moved between one index page and another, due to the insertion
> of an earlier message.

mhonarc::CBMessageHeadRead is only called for new messages added to
an archive.

> How do I figure out that a message has been moved
> onto a different index page?  Would it be better to iterate through all the
> messages in say %mhonarc::IndexNum at the end of processing?

It is a little tricky if you use OTHERINDEXES since if you use
mhonarc::replace_li_var() will then be based upon the last index type
you specified.

If not, then yes, you can iterate over %mhonarc::IndexNum, or
@mhonarc::MListOrder to avoid have to call keys.  @mhonarc::MListOrder
is the array of message indexes as they will appear on the main index.

> BTW, what is the control character stuck on the end of lots of the hash keys
> in .monharc.db?  

The control character is used within each key index for a message.
The actual character used is represented by the cryptix $mhonarc::X
variable.  Here is the comments of it in mhinit.pl:

$X = "\034";    # Value separator (should equal $;)
                # NOTE: Older versions used this variable as
                #       the list value separator.  Its use should
                #       now only be for extracting time from
                #       indexes of messages or for processing
                #       old version data.

It is a relic of Perl 4 days and each message key index actually
containes information used by MHonArc: the unix date of the message.
It is not very efficient, but avoided the need for yet another hash to
store the Unix time for a message and not as expensive as parsing the
original raw date string each time.  I'd definitely do things differently
if I were start over again.

Since Perl 5 added support for references, I finally was able to move
away from the Perl 4-way of having "arrays" as hash values (see the
CHANGES file when that finally happend), but still keep compatibility
with older version of MHonArc.  See mhopt.pl for the auto-upgrade code
MHonArc executes if it encounters an older archive.

Unfortunately, the use of main multiple hashes to store data is
still around.  Changing would basically be a rewrite and a compatibility
challange.  Ideally, a archive would be represented as a blessed hash
(an "object" in OO-terminology) with all associated data contained
within the hash.

--ewh

---------------------------------------------------------------------
To sign-off this list, send email to majordomo@mhonarc.org with the
message text UNSUBSCRIBE MHONARC-USERS