It looks to me as though the Wizard only remembers the matches which are done manually, not those where the match is against the list chosen at the top of the matching window. . For example if I select a specified list  and match against this and then have some unmatched species left, which I then match against oher lists, then only these extra matches are saved. Next time I match against the selected list it rematches the ones it did  automatically and uses also the matches I have done manually. This  matching  does not seem to happen if I choose a differnt list.  The table which holds the matches has the key for a taxon list in it  and its seems that this is  somehow this involved in the matching.  I don't  understand what is happening here, but it does seem to work as long as I take a consistent approach to matching and do not use different lists.  The main problem is that none of this is explained anywhere in the detail required.

Dave

The only matches which are saved are those where the users finds a match in a Dictionary. Those made automatically  are not saved.  When the  manually matches are saved the matched_checklist_key  is taken from the checklist currently in use at the time of the save. Where 'All preferred lists' has been chosen then the matched_check_list_key is null.  When search usd to get the saved taxa, it first looks at saved matches for the currently selcted dictionary list, then for any automatic matches against this lists. The way the matches are saved against different dictionaries can cause problems at this stage as the matches may not have been saved against the current list and will therefore not be accessed.

The approach which appears to works best is to do the search against 'All Preferred Lists'. As far as I can see this does what it says and finds the matches on preferred list, not just on the species to be matched, but on any synonmyms, identified in Index_taxon_name via NameServer.  Anything which is not automatically matched this way or any which are not considered right,   can then be matched manually against the appropriate list, but it is important not to change to the checklist (All Preferred).

If this process is followed every time then matching will pick up everything matched before and those new ones which can be matched automatically. Only those new entries which can't be matched automatically will need to be matched manually.

During the testing of the process I have noticed that delete/replace is not working for 'All Preferred Lists'. What is happening instead is that a new entry is being created. This means that subsequent matching of those species is arbitary.   This has been raised as a high priority bug, because there seems to be no easy way to get around this.

Can I just say that I am not defending what is happening, just attempting to explain it. I rarely use the Wizard as most of the work I do is importing data where the Wizard can't cope or would take forever to set up, so I was not that familar with exactly how it worked until I looked into it yesterday. I suspect that making the matching work on a combination of taxon name and chosen dictionary is intended to make the matching more flexible. Some users don't like matching against the preferred list because the taxa allocated does not have the same name as  that originally recorded.  What they can do under the current system is to match a list which uses the older names against say the R3.3. list and attached the manual matches to that. Then for a list which uses more modern names thay can use a preferred list and attach perhaps a different set of manual matches to this.  The testing I have done suggests that apart from the one thing I picked up it is working, but I agree that the way it works can be confusing and that it could do with looking at in more detail. It would also help if the import wizard tables could be updated by a Batch Update as this would at least give users more control.

I am still experimenting with this. See my post below following John's comments.

From John's explanation the way this should work with a  list of taxa which cover several groups is that you process each dctionary list involved  in turn. Use search to pick up the automatic matches, then do any manual matches for the list and commit these before moving onto the next list. This approach should  solve the homonym issue and mean that you will always get the matches you need as long as you follow the same procedure. The main thing is not to commit matches with the wrong dictionary list selected and to always to commit manually. If you leave the system to commit when the import is run then it will put all the matches against the last list selected.

Hi Rob
In effect what you are suggesting is that the search against the preferred list becomes the default behaviour and that the ability to browse other lists during matching is made an "advanced" option so most users won't need to bother with it. Also that if there are duplicate matches, rather than no match being made, list the potentials along with their taxon group and let the user select?