[Imaplib2-devel] recovering message numbering for EXISTS and EXPUNGE

[Ross B. <ro...@bi...>]

I think that it is not possible to reconstruct exactly which messages
were added and removed in some circumstances.  Is my understanding
correct?

If this sequence came back from the server
* 4 EXPUNGE
* 5 EXISTS
then the new message is sequence 5.

If this sequence came back
* 5 EXISTS
* 4 EXPUNGE
then the new message is sequence 4.

Since imaplib2 only gives the relative time sequence within a single
class of untagged response (i.e., EXPUNGE or EXISTS) there is no way to
tell which of these scenarios happened assuming one recovers the 2
separate responses.

For my purposes that level of detail is probably not necessary; I can
just err on the side of caution.

Perhaps imaplib2 should have a callback that would enable tracking the
exact sequence of messages.

Context: iterating slowly through the messages in a large folder, so
that messages might appear or vanish during the iteration.  It looks to
me as if doing this by message sequence number rather than UID is most
natural.(*) The commands are like
SEARCH 1:50 <some additional filters>
it would be natural to do
SEARCH 51:100 <some additional filters>
for the next round, but this could miss messages if some were deleted in
the  interim.

Hmm, there is still a problem with a race even if I check the message
sequences immediately before the SEARCH, since a message could be
deleted in between checking the status and issuing the SEARCH
anyway.....

(*) Most natural because EXISTS and EXPUNGE use sequence numbers, and
because UIDs may in principle have large gaps.  The latter implies there
could be many searches by UID range that get nothing.  Perhaps a more
robust approach would be to collect the UIDs of messages at sequence 1,
50, 100, 150, .... at the start, and then fetch by the implied UID
range.

Ross Boylan