View Single Post
Old 04-16-2011, 05:19 AM   #74
kiwidude
Calibre Plugins Developer
kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.kiwidude ought to be getting tired of karma fortunes by now.
 
Posts: 4,637
Karma: 2162064
Join Date: Oct 2010
Location: Australia
Device: Kindle Oasis
Quote:
Originally Posted by chaley View Post
- Did "find dups". It found the expected 4 groups. I switched to "show non_dups", selected them all, and said to remove exemptions. It did. I then pressed "find dups", expecting to see the now 6 groups. I saw the original 4. I suggest that pushing the find dups button after removing exemptions should redo the alg.
That's a good point. Should it present the whole UI search dialog again, or just re-run the duplicate search in the background (I'm thinking the latter).
Quote:
- Same as above, but pressed manually selected find dups. Saw all six, but the sort order is wrong. The two new groups are sorted at the bottom. This is probably caused by the multisort thinking it has already sorted the screen. I suggest that you force a multisort (don't set the ignore flag) after every find_dups.
Will fix this.
Quote:
- The 'are you sure' messages should offer the checkbox to not show the message again. In particular, the mark exempt dialog should have this checkbox. The confirm method in gui2.dialogs.confirm_delete.py should be able to do the job.
Thanks, will take a look into that.
Quote:
- A hygiene-factor thing. When the user selects 'mark group exempt', it might be a good idea to check if any books not in the group are selected. If they are, then the user is probably confused and thinks that the selection will be used in lieu of the group. (Guess who almost did that ... )
Funnily enough I had started coding exactly that and then I ripped it out to see what people thought. That you have come up with it tells me it should be back in. The reason I removed it was I couldn't decide what to do if the selections don't match. For instance, should I require that only selected rows be on the current matching group? What happens if they intersect with another group? And do I just simply tell the user they are not on the right group and wait for them to select the right one, or do I warn them with a question dialog and then let them go ahead if they say yes?
Quote:
- Idea: if you connect to gui.search.cleared, you will be notified if the user clicks the clear button on the search bar. That would permit clearing the search to do a "clear duplicate results". Should it?
Sounds a good idea to me, I'll add that too, thanks.

I've been thinking more about the title/author independent algorithm thing, the more I think about it the more I am tempted. I would like to see two sliders with labels on the tickmarks (which sadly Qt cannot do out of the box). The slider would have a range values something like:
"Identical", "Similar", "Vaguely Similar", "Ignore"

One slider for each of title and author. If you set both title and author to "Ignore", then it does an ISBN match. A descriptive text box would summarise the combination you had selected a little bit like it does now.

A first time user would get it set to "Identical Title", "Identical Author". The "Vaguely Similar" (or "Fuzzy" or some better name!) author and title selections would do the fuzzier algorithms I suggested above.

Any thoughts? I'm just concerned about the permutations - break them apart and the problem goes away.
kiwidude is offline   Reply With Quote