15142 - Need to do secondary sort on thread pane

Status: VERIFIED FIXED

(MailNews Core :: Backend, defect, P2)

Description

[Phil Peterson]

When I sort the thread pane by sender, the messages come out in random order,
within each sender. Secondary sort by date would be more desirable.

4.x sort of does this, since it presents messages in ID order, which is usually
date order (but sometimes not).

In some perfect world, it might even be nice to specify multiple sort keys, but
I'd be happy with secondary sort on date.

Comment 1

[Phil Peterson]

M15. rjc, do we need to extend RDF searching for this?

Comment 2

[Robert John Churchill]

I'll check in some changes when the tree opens today [Nov 10] to enable this.
You'll just need add a [optional] "rdf:resource2" attribute [to accompany the
"rdf:resource" attribute]  to relevant 'treecol' tag(s).

I'll add another comment on this bug when the check-in is complete.

Comment 3

[Robert John Churchill]

Support for secondary sort keys is now checked in. Feel free to play with it.

Comment 4

[scottputterman]

great! thanks.

Comment 5

[Phil Peterson]

*** Bug 18846 has been marked as a duplicate of this bug. ***

Comment 6

[Michael La Guardia]

I'm still seeing the random list when sorting on sender with the 11/17/99 build.

Comment 7

[scottputterman]

that's because I haven't done anything to incoporate rjc's changes into mailnews
yet.

Comment 8

[scottputterman]

I checked in the fix.  When you sort by anything besides date, the secondary
sort is by date.

Comment 9

[fenella]

Linux Redhat 6.0 (2000-01-20-08 M13)
Win_nt 4.0 (2000-01-20-09 M13)
Mac (2000-01-20-08 M13)
Secondary sort by date is working in both POP and IMAP accounts on all
platforms.

Comment 10

[sol]

Reopening - for large folders, secondary sort works in chunks (incorrectly). For 
example, if you sort on sender, you see messages from that sender organized as 
follows:
     12/17/99
     1/1/00
     1/4/00
     7/31/99
     8/15/99
     1/10/00
     2/1/00
     6/15/99
     10/31/99

Within a "chunk", the messages are sorted correctly by date, however, the sort 
should occur across all of these chunks.

Comment 11

[scottputterman]

reassigning to rjc.  I'm pretty sure I'm setting the secondary sort key 
correctly.  

Comment 12

[Robert John Churchill]

Sol, can you retest this?  I'm not seeing this problem.  Do you have have
threading on or off?  Can you determine a threshhold size for which this is
exhibited, but below it isn't?

Comment 13

[sol]

Rich: I'm still seeing this problem using the following builds on WinNT4:
   2000-02-04-10
   2000-02-06-08

I have threading turned *off*.

I sort by sender. If I have a lot of mail from the sender in the folder, then
the first 130 or so messages from that sender are correctly sorted by date.
After this first block of correctly sorted messages, another block of messages
from the same sender (again sorted by date within the 2nd block) is presented.

Let me know if you want to see this behavior on my machine.

Comment 14

[Robert John Churchill]

As soon as I figure out who Rich is, I'll send him over.  <grin>

Comment 15

[sol]

Not sure how this bug got stuck to me, but please, get it off me!

Reassigning to putterman - not sure who the right owner is at this point.

Comment 16

[scottputterman]

I'm not sure why this is getting reassigned to me!

I figure Sol mixed up rhp with rjc.  Reassigning back to rjc.

Comment 17

[Phil Peterson]

This bug is driving me crazy. It makes sorting in the thread pane pretty much
useless. Can we pretty please have a fix for this soon?

Comment 18

[Robert John Churchill]

Scott, I'm looking at "mozilla/mailnews/base/resources/content/threadPane.xul"
and I don't see any "rdf:resource2" (note the 2 for secondary!) attributes on
the <treecol> tags in this file.  

[Are you sure that your change from 1/12/2000 went in?  Or, am I looking at the
wrong file? Or, ... ?]

Comment 19

[scottputterman]

I'm pretty sure I set it programatically in commandglue.js in SortColumn.

Comment 20

[Robert John Churchill]

Yeah, I see the JavaScript. (Any reason why, instead of just setting it in
XUL?)  Continuing to look at it.

Comment 21

[scottputterman]

Yeah, I was trying to make it so that if we ever had a ui to change the 
secondary sort that we were set up for it.

Comment 22

[Robert John Churchill]

I'm noticing that if I comment out TEST_SORT  (line #96 in 
mozilla/rdf/content/src/nsXULSortService.cpp)  which reverts to older sorting 
code, the problem seems a lot better.

Phil/Scott, if you comment that out and rebuild RDF, do you see this also?

Comment 23

[scottputterman]

that would be amusing.  If that's the case, then I can take a look at it, though 
if you see anything obviously wrong, I'd appreciate the help.

Comment 24

[scottputterman]

ok.  It was my fault.  I accidentally typed node1 instead of node2 someplace and
that caused all of the problems.  Changing it to node2 makes it work for me.

Comment 25

[Robert John Churchill]

Great!   :^)

Comment 26

[scottputterman]

fix was checked in yesterday.

Comment 27

[fenella]

Linux (2000-04-11-09-M15)
Win32 (2000-04-11-09-M15)
Mac (2000-04-11-08-M15)
I have a folder that contains 2315 messages. The problem still exists in certain
block of senders. This occurs on all 3 platforms. Re-open the bug.

Comment 28

[lchiang]

When you reopen this bug, the target milestone should be set to a more
appropriate one.

Comment 29

[fenella]

Set it to M17 per Scott Putterman

Comment 30

[selmer (gone)]

Mail Review recommends not in this release.  Marking M20.

Comment 31

[scottputterman]

so, this is just a bug.  for some reason it works fine in debug builds but not 
in release builds.  I'm not sure we should remove this from the product just 
yet.  We should spend some time on it.  On the other hand, how much time do we 
want to spend on this if it turns out that there's some sorting bug that only 
manifests itself in the release build and might be hard to track down?

Comment 32

[Phil Peterson]

I'd invest a lot in it since it makes sorting by anything other than date look
wrong. Bummer that it's release-only. Are you going to move the target milestone?

Comment 33

[scottputterman]

I think we should nominate this.

Comment 34

[Jim Roskind]

This sure would be nice... but there are too many critical beta2 bugs.  Putting 
on beta2-, but nominating for beta3.  As stated, this is a bug in the non-debug 
build, which is a hassle, and often a time sink.

Comment 35

[lchiang]

+ per mail triage

Comment 36

[lchiang]

P2 per mail triage

Comment 37

[scottputterman]

cc'ing nhotta.

I have a theory that this is a collation key problem.  I have output for a 
folder of mine where it's telling me that comparing Scott MacGregor to Scott 
MacGregor is returning a sort order of 1 rather than a sort order of 0.  The 
collation keys appear to be different.  The first one gets ...................a0
and the second gets ......................a.  I only see the bug in release 
mode.  Any ideas on what is happening?

Comment 38

[scottputterman]

Don't know much about collation keys, but debugging, I'm seeing one Scott 
MacGregor get "???????????h???aa" and another get "???????????h???aa ????a"  Is 
it possible some buffer's being reused and garbage is in it?

Comment 39

[nhottanscp]

Scott, could you give me a simple data and I will investigate if this is a 
collation key problem.

Comment 40

[scottputterman]

nhotta gave me a patch that makes this work on my release build.

Comment 41

[scottputterman]

I checked in the most recent patch Naoki sent to me.

Comment 42

[scottputterman]

Reopening this seems to be happening on Linux and Mac still.

Comment 43

[fenella]

Linux (2000-08-21-12 M18)
Win32 (2000-08-21-08 M18)
Mac (2000-08-21-08 M18)
This bug is still on Mac POP and IMAP
Linux in IMAP only, Not POP.
Win32 is OK.
Show it to Scott. M. He will re-open it.

Comment 44

[fenella]

Oh. I meant Scott P, not Scott M.

Comment 45

[scottputterman]

reassigning to nhotta. I'm pretty sure this is a collation key bug based on the 
fact that Windows got fixed with this and if I remember correctly, this code is 
different per platform. I'm pretty sure the fact that it's not working for 
certain folder types is coincidental.

Comment 46

[nhottanscp]

Checked in a fix. I tested on my Macintosh release build and the sort problem 
was fixed with this change.

Comment 47

[fenella]

Mac (2000-09-08-20 M18)
Linux (2000-09-11-08 M18)
This problem is fixed.