Revisions for imp/lib/Mailbox.php

Start moving what I can from my local branch to HEAD so that the branches
don't get too divergent and I pull my hair out later trying to merge the
boatload of changes coming due to the IMAP/MIME changes.

IMAP_Cache has been merged into IMP_IMAP_Cache::.

Cleanups on some missed items from the removal of IMP_IMAP_Search.

Remove IMP_IMAP_Search:: and IMP_IMAP_Search_Query::.

MFH:
1.12      +24 -8     imp/lib/IMAP/MessageCache.php
1.268     +2 -2      imp/lib/Mailbox.php
2.869     +3 -4      imp/mailbox.php
1.1171    +1 -0      imp/docs/CHANGES

Tags: IMP_4_3, IMP_4_3_RC2

Cache results of the IMP_UI_Mailbox::getFrom() call, since calls to the
IMP::parseAddressList are expensive.

MFH: 1.267

Tags: IMP_4_2_1_RC1

Fix parentheses.

MFH 1.266

Make sure we are receiving a copy of the cache data, so we can't
unknowingly modify the data ourselves.

Bug: 6449
MFH 1.265

Tags: IMP_4_2, IMP_4_2_RC4

Bug: 6449
Fix name of cache entry.

Bug: 6507, 6508
MFH 1.264

Bug: 6507, 6508
Right idea, wrong implementation.  We can't keep track of the hide deleted
status in the cache key since the key needs to be an unadultered mailbox name
so that other code can reliably access the entry.  Instead, keep track of
hide deleted status within the data that is affected by the setting - namely
arrival time and the message list.

Bug: 6449
MFH 1.263

Bug: 6449
Missed several conversions to using _cachedMailbox().

MFH: CHANGES - 1.1104; Mailbox.php - 1.262

Fix mailbox list caching when hiding deleted messages.

IMP_MailboxCache -> IMP_MessageCache.

Tags: IMP_4_2_RC3

This has bothered me for the longest time - we are storing messages in this
cache, not mailboxes.  Change the library name now and corresponding config
options so we don't need to change them after 4.2 is released.

MFH 1.260

Allow header information to be returned in getMailboxArray().

Bug: 5234
MFH: Add hook for adding formatting information to a message list row.

Tags: IMP_4_2_RC2

Happy New Year

Happy New Year

Bug: 5234
Merge after: 3 days
Add hook for adding formatting information to a message list row.

MFH 1.257

Bug: 6023
Don't expire cache when traversing through virtual INBOX via the message
screen (we don't want flagging a message to destroy the cached value of the
mailbox).

Merge from HEAD.

Revert direct link to thread view.

Tags: BEFORE_MERGE_3_2, IMP_4_2_RC1

Add direct link to the thread view from the mailbox (Request #5902).

Reduce session storage space (especially with large mailboxes) by not storing
mailbox names for each uid if we are in a non-search mailbox.

Sanity checking

Use our own set of IMAP flags to not unnecessarily reopen mailboxes.

Tags: IMP_4_2_ALPHA

getSortedList() should return an array with message numbers starting at 1,
not 0.

Only rebuild mailbox on message count determination when we are not dealing
with a virtual folder.

Makes more sense to return message number starting at 1 rather than an
array-like 0.

Our internal range indices are 1 less than the given ranges.

Rename buildMailboxArray() to getMailboxArray() to more clearly indicate that
we are not necessarily generating a "page" of information.  Furthermore,
change parameters to allow a defined set of message numbers we want information
about.

Make sure mailbox is updated before we try to get message count.

Make sure we always resort the sorted list because we are unsure of what
an underlying function might do with the data.
Shortcircuit some cache clearing code if we explicitly rebuild the mailbox.

The cached imap_status() object must be cleared on flag change.

Need to give public access to imap cache status string.

true == 2

Add way to override pref and always return preview information.

unset doesn't throw notices, so this should protect us (isset does if
a key inside the isset expression isn't set, since that's evaluated
before the isset).

Bug: 4830
Silence undefined warning.

Happy New Year

Tags: IMP_4_1_6, IMP_4_1_5, IMP_4_1_4, IMP_4_1_4_RC1

Happy New Year

Assign copyright to The Horde Project for those copyright holders that already agreed to it.

Make sure arrival cache data is updated when we delete messages.

Fix *extremely* difficult to track down bug - we were doing an update instead
of a store when switching sort directions, which was causing the cache to
not be correctly updated with the updated cache key.  Was causing 1 particular
installation to have 1 sort column on 1 mail server not updated correctly when
toggling - and I can't event begin to explain this behavior.

Add _getSortedList().

coding standards: don't use is_null().

Add parameter to determine whether to include preview information.

Yuck.  The way we were storing data in the cache could cause a message list in
a virtual folder to have missing messages - since UIDs are not unique across
mailboxes.  Fix this by changing the storage structure in the cache.

Remove debug line.

Tags: IMP_4_1_3

Bug: 3171
Remove current mailbox information from session - allows multiple
simultaneous open IMP windows.

Fix undefined index, use logic from HEAD.

Remove 'stream' as a session variable.  it is nothing more than a disguised
global since it doesn't persist between page accesses.
Don't require us to pass the imap stream to several overriden functions since
we can do the necessary stream passing in the overriden functions themselves.

Refactor code so Auth_imp:: only deals with authenticating to the IMAP server.
Auth_imp no longer automatically opens a connection to the current mailbox. We
must now manually call IMP_IMAP::changeMbox() anytime we are working with the
IMAP stream and require a connection to a mailbox.  Much cleaner, and should
eliminate any issues with clearing the 'recent' flag incorrectly.

learn to spell michael

eliminate repetitive function calls in a loop.

Bug: 3877
I *think* this fixes the issues in Bug 3877.  Namely, we may not have the
current up-to-date mailbox list when we are removing the messages - since we
may be somewhere, like a search screen, where we only use the cached mailbox
and make no attempt to update the mailbox from the current state. Thus, if
messages have come in since the last time we directly accessed the mailbox,
they won't be in the sorted list and won't appear when we rewrite the cache
entry with the current imap_status information.  Solution: simply update the
current cache entry.  If this caused the mailbox to change (imap_status() info
message changes), this will already be caught the next time we need to view
the full mailbox.

More tweaking (cleanups as I go through this code line-by-line).

Rather than comparing objects directly (I don't trust PHP), this is probably
a better check to make sure we don't do multiple updates of a single object.

Correctly fix sortlimit check.

Add sort preferences to the IMAP cache information. Needed because the sort
preference may change in a mailbox without any user request - i.e. a mailbox
exceeded the sort limit.

Only call getSort() when needed.

Much improved sort preference handling - instead of handling in 3 different
functions in 2 different libraries, a single IMP::getSort() will now return
the correct sort preference for the current mailbox.

1 less imap_search() call, but only if we are not hiding deleted messages.
imap_status() has truly become our friend.

Move some display logic from IMP_Mailbox:: to mailbox.php.
If we don't need to hide deleted messages, just use imap_status() information
in newMessageCount() instead of doing a search.

Restore PHP 4 BC.

Fix message arrival number generation from cached data.
Remove unneeded var.

A more extensible way of allowing us to (eventually) store more information
about message indices without increasing the number of class variables and
drastically increasing sessions sizes.
Clean up _setSorted() to make it easier to understand what it is being asked
to do at any given time.

Cache the arrival order in the mailbox list cache - saves us an IMAP access
on every mailbox page load when the cache is still valid.

only need imap_cache here in this method.

re-add - we do need this.

Compromise for mailbox caching - having mailbox_start set to either first
unseen or last unseen messages was killing the whole concept of caching - since
we would have to do a search every page to determine location.  Instead, only
honor preference the first time we load mailbox; subsequent visits to mailbox
will start up where the user last left the mailbox.  Honestly, this seems
to be better behavior anyway (and more like other mail clients) IMHO.

clean up some unneeded/duplicative variables/functions.

Bug: 3913
Fix seen flag when stripping part.
Save a call to flag() because it seems imap_append() accepts a list of flags to
set (not documented very well).
Make 2nd parameter to IMP_Mailbox::updateMailbox() optional.
Make sure if we invalidate the current mailbox object that we ensure
_buildMessage() will correctly re-build the next time we run it.

getMailboxArrival() makes more sense in IMAP_Cache instead of mailbox caching.
Simplify cache expiration by having a single function that can do 3 different
types of cache expiration.
We need to make sure we expire cached results of imap_search() arrival searches
since mailbox actions may alter the results of these searches.

Simplify code

Need to expire any cached imap_status() information if we do anything to alter
the mailbox during the page access.

Bug: 3877
Yet more improvements to mailbox list cache updates.
Biggest fix: when in search mailboxes, make sure to not only update list of
indices in the base mailbox of the messages, but also update the mailbox list
in the search mailbox entry.  Therefore, we need to pass in mailbox information
with indices to updateMailbox() so that the search mailbox object can correctly
determine the array index from the UID.

Rework IMP_Message:: so that the trigger to update the mailbox list cache
is via either expunging or flagging.
Do some fancy, smart indices manipulation when we are given an exact range
of messages to expunge.  This allows us to selectively keep other messages
in the mailbox marked as deleted even after an expunge.  This is great when
there are messages marked as deleted in a mailbox and we are using a trash
folder, we can now move messages when not using trash and the old message
is deleted instead of simply marked as deleted, etc. This costs us an
extra imap_search for deleted flags, but this shouldn't be too expensive of
a call and once we improve caching a bit more will save us in the long run.

Don't need a 'move' update action (since a move is really either a
'flag' or 'delete' action).

Move incrementing current indices to message.php.  We only need to increment
these indices if we are in the message view and, from a design standpoint, this
is a UI issue anyway.
Vastly simplify IMP_Message::undelete().
No need to directly update mailbox cache on either copy or undelete.

Correctly update mailbox list cache if expunging messages.

More improvements to mailbox list cache updating.

Can return success information and list of updated indices using a single
parameter. Now just need to use the list of updated indices to update the
mailbox status in updateMailbox().

no longer need this function now that setIndex() is public.

Bug: 3876
Make the set index function public.

Bug: 3867
Fix undefined function error.

correct object

Make sure _buildMessage() is only processed once per instantiation.
If $index is passed to singleton(), make sure it gets propogated to object.

minor cleanups

These changes seems to make caching work much, much better.
Search results seem to be working again.
We now cache thread objects.
We should only rebuild the mailbox list now in the following situations:
+ first access to mailbox
+ if mailbox has changed (per imap_status() call)
+ in message view, only when we reach the last message in list.
+ for search mailboxes, every mailbox access
What are we gaining?  In cases 2 and 3 above, we used to always generate
the mailbox list - now we only regenerate if needed.
In case 1 above, we used to regenerate every time we switched mailboxes.  now
we only regenerate when switching mailboxes if the mailbox has changed.

Actually, it seems as if there are a few bugs left.  Therefore, in order to
keep CVS working as expected, just rebuild the mailbox on every page load for
now.  Inefficient, but it works.

combine some duplicative functions.
many less calls to _buildMailbox().
Still not working entirely properly with search mailboxes, but doesn't seem
to be throwing errors anymore with these mailboxes and seems to be working just
fine for regular mailboxes.

Improve mailbox cache updating.

This got deleted accidentally.