Hmm, so there are... I suppose if you still have the source data, the thing to do would be to check the output of sort before it gets fed to uniq -c, and see if things appear out of place there. If not, try getting a diff of the lines out of uniq -c before they go into the next sort (I presume you used a sort on the initial numeric field at the end?). Getting the difference between lines might be fiddly though. Maybe "od" or similar might be of more use? Personally I wouldn't want to resort to somebody else's program though, I'd just have to spend time learning the thing and likely find it isn't what I'm after.
Oh, I still have the source data, of course, but the obsessive spell broke and I wandered away from the problem. When I get around to doing the same thing with the full logs (a few years worth), I'll want to make sure it all sorts correctly then.
I'm wondering whether a 3-dimensional graph can be arranged in a way that'll show me interesting things about the relative poularity of different pages over time, rather than just looking pretty. (Maybe if I limit the graph to the few most popular pages, instead of trying to crowd all of them in there...)
Re: Backups
Maybe "od" or similar might be of more use? Personally I wouldn't want to resort to somebody else's program though, I'd just have to spend time learning the thing and likely find it isn't what I'm after.
Re: Backups
I'm wondering whether a 3-dimensional graph can be arranged in a way that'll show me interesting things about the relative poularity of different pages over time, rather than just looking pretty. (Maybe if I limit the graph to the few most popular pages, instead of trying to crowd all of them in there...)