>>I thought one argument against automatically using nofilter is that it can be slower as recorda hev to be physically written everytime. I usually check _TALLY to find how many records have been selected. Do you think that is always accurate.

>If I say no then some lurkers would request reproducable code:) I know that it doesn't work right 100% but do not have reproducable code just at the moment so I prefer to say yes (also from version to version reliability matters). I think you'd never see its problems, using collating machine and English only.

The place to suspect _tally is the command window - in case there was an intellisense script which may have run something which would change it anywhere between your record-related command and the time you checked _tally.

That would, of course, mean that we need to preserve _tally in any such scripts we write, but then if the default set of scripts doesn't, then I don't see why should we.

So if _tally is suspect in command window, we can always select the set of commands, rightclick and use eXecute Selection from the menu. That should call no Intellisense between the select/replace/whatever command and the ?_tally command.