The index files read are created by ezmlm-idx(1) on a per-list basis and by ezmlm-send(1) on a per-message archive for a indexed list.

The output files created are:

dir/archive/threads/yyyymm

The thread index. It contains one line per subject, starting with the number of the first message with that subject within the set investigated, ``:'', a 20 character subject hash, blank, ``[n]'' where ``n'' is the number of messages in the thread, blank, and the subject. The file ``yyyymm'' contains entries for all threads that have messages in the month ``yyyymm'' or that have messages both before and after that month. The subject hash is a key to the subject files; the message number is a key to the index file. The lines are in ascending order by message number when the index is created de novo on an existing archive. When the messages are added one-by-one as in normal archive operation, ``n'' is the number of message in the thread for the particular month and the order is in reverse of latest message, i.e. the last extended thread is shown last. The message number accompanying a thread is always a message within the thread. It is the first in archives created on existing lists, and the last message in incrementally created archives. Use the corresponding subject index file to get a list of all messages in the thread in ascending order.

dir/archive/subjects/xx/yyyyyyyyyyyyyyyyyy

A subject file. The first line is the subject hash, a space, and the subject. This is followed by one line per message with this subject, in the format message number, ``:'', date (yyyymm), ``:'', author hash, blank, author from line. The lines are sorted by message number. The author hash is a key to the author files; the message number is a key to the index file. The file in the example would be for the subject hash ``xxyyyyyyyyyyyyyyyyyy''.

dir/archive/authors/xx/yyyyyyyyyyyyyyyyyy

An author file. The first line is the author hash, a space, and the author from line. This is followed by one line per message with this author, in the format message number, ``:'', date (yyyymm), ``:'', subject hash, blank, subject. The lines are sorted by message number. The subject hash is a key to the subject files; the message number is a key to the index file. The file in the example would be for the author hash ``xxyyyyyyyyyyyyyyyyyy''.

dir/archnum keeps track of the last message processed. Normally, ezmlm-archive will process entries for messages from one above the contents of this file up to an including the message number in dir/num.

-c

Create a new index. This overrides dir/archnum causing ezmlm-archive to start with the first message in the archive. Synonym for -f0. NOTE: ezmlm-archive does not remove files in the index. While it will overwrite/update old files it will not remove files that are obsolete for other reasons.

-C

(Default.) Process entries starting with the message after the message listed in dir/archnum.

-f msg1

Process messages from the archive section (set of 100 messages) containing message msg1. This is useful if you have removed part of the archive, as it will shorten processing time and decrease memory use. NOTE: ezmlm-archive does not remove files in the index. While it will overwrite/update old files it will not remove files that are obsolete for other reasons. The number of messages per thread will be incorrect when using of the -f and -t switches leads to partial re-indexing of already indexed messages.

-F

(Default.) Do not change the starting message from the default (see -C).

-s

Always sync files.

-S

(Default.) Sync files, except when on of the message range modifying options is used.

-t msg2

Process messages to message msg2 instead of the last message in the archive. Again, files written are corrected, but other files are not explicitly removed.

-T

(Default.) Process entries for messages up to the last message in the archive.

-v

Display ezmlm-archive version info.

-V

Display ezmlm-archive version info.

In normal list use, it processes only at most a few messages at a time, but for initial processing of a large archive, considerable amounts of memory may be used. Assuming 40 bytes for subject/from line, 5 messages per thread, 100,000 messages, and 1000 authors, this is 2.5 MB. For 1,000,000 messages this is about 20 MB.

Thus, for large archives, it may be useful to use the -t switch to process the archive in multiple subsets, starting with e.g. the first 100,000, then the next, and so on.