Quote:
Originally Posted by kiwidude
Allowing the user to choose sounds interesting but not without some technical challenges. The way this plugin works is it scans the text of book pages using a regular expression to find numbers that look like an ISBN. So right now all it could do is show the user a list of numbers - not particularly useful given there are no rules for publishers about what order they should be defined and for what purpose. They may even be from an advertisement for another book by that publisher...
So you would have to try to give at least a snippet of text preceding each (which still wouldn't help in the case of an advert). Even that might not necessarily be trivial given how this plugin strips stuff out prior to searching for ISBNs, would have to check.
Then you have all the UI and workflow aspects to consider - right now this runs as a background job in calibre so someone can queue up 50 books and forget about it. You would have to force a UI interaction on them when the job is completed (at least in the circumstances where more than one match is found). Perhaps displaying all the isbns across all the books in a grid and force the user to perhaps tick a box as to which they want from each book group. As one book at a time through some sort of wizard would be too painful to use.
Hence IMO it is a non-trivial change to make that falls very squarely into the rainy day category. It is an interesting suggestion though, and has the potential to reduce some of the noise that has happened in this thread over the years when people come here to exclaim that the plugin should apply rule xyz just because they have a book or two that would work for...
I'm sure if we hooked this into a cloud AI with some learning patterns it would make better choices - but no that isn't going to happen  |
Yea, it's definitely an interesting problem to solve.
To try and get an idea of the various ways I've seen publishers display the eISBN, below is a collage of screenshots from 15 different eBooks in my Calibre library.
https://i.imgur.com/SLgu71w.png
I feel like if the plugin could possibly use some sort of regex that matches all those variations, it might catch the eISBN 99% of the time. Thoughts?