Author(s) vs Author sort

[TarquiniusX]

Hi all,
I apologise in advance if this issue has already been raised; I have searched but failed to find and answers.

Since the 0.7 release of Calibre, the content server has started listing books by the value in the 'Author(s)' column rather than the value in 'Author sort'. Prior to this I would have, for example, 'Douglas Adams' as the Author and 'Adams, Douglas' as the Author sort value. All was well and good. When looking for books on the contnet server Douglas Adams would appear right before Robert Adams. Now however Douglas Adams appears in the 'D' list and Robert Adams appears in the 'R' list (along with all the many other Roberts out there).

I started editing the meta data for the books (again) to make both fields 'Adams, Douglas' but that defeats the purpose of having both fields. Having changed the Author(s) value to 'Adams, Douglas' has it listed as such on the content server which seems to conform that the server is looking at the 'Author(s)' rather than the 'Author sort' field.

Is there an easy way to change the behaviour back to the previous way of displaying book when sorted by author?

Thanks in advance,

Robert

[kovidgoyal]

This change is deliberate. The content server now reflects the way authors are displayed in the Tag Browser. So no, there's no way to change it.

[Worldwalker]

Why do we want our books sorted by the author's first name instead of last name?

[kovidgoyal]

The books are not sorted by authors first name, the authors are sorted by authors first name

[TarquiniusX]

Thanks for your very fast reply.

I understnat that this is your app and you are developing it, but I can not think of any logical reason to sort authors via their first name.

So right now I'm looking at having to re-do my entire ebook collection (over 2000 titles) just so they disply correctly. I mostly have academic pdf and I rarely bother to remember the authors first name. I've always dealt with family names as I suspect everyone else has too.

Which brings us back to why there are two fields now. Calibre currents automatically changes Douglas Adams to Adams, Douglas and then sorts via the second in the app on the computer, but sorts via the first in the content server. Thus we have an inconsistant method of displaying books.

Also...I've just discovered that while attempting to select several titles and edit meta data individually, it still opens the edit in bulk window.

[kovidgoyal]

Given the display name of an author, come up with an algorithm that unfailingly converts it into lastname, other names and works across all types of author names from all parts of the world. If you do come up with such an algorithm, I will be more than happy to have calibre sort authors by lastname.

Indeed, there are plenty of cultures where people have no last names and some names are sorted differently depending on what culture is doing the sorting.

So I have decided not to open a hornets nest and simply display and sort authors by their display name. Get used to it.

The author_sort field is not associated with authors, it is associated with books and it is meant to be hand correctable. In other words you can enter whatever you think is the best way to sort (possibly multiple) authors associated with a book.

[TarquiniusX]

I appreciate your replies, I really do, but...

Perhaps there is a misunderstanding of the basic issue here.

In previous versions of Calibre (prior to the 0.7.x release I mean) when you add a new ebook to the collection you could edit the 'Author(s)' value and Calibre would automatically set the 'Author sort' value. It still does that; I just checked.

In the app itself on the computer (for me Windows 7), when the collection is sorted by the 'Author(s)' column, the column displays 'First name, Family name', but the books are actually sorted via 'Family name, First name' as found in the 'Author sort' in the meta data. Thus the value for 'Author(s)' was displayed, but sorting was carried out employing the value in the Author sort entry of the meta data.

The issue is with the order when sorted on the content server. It is sorting via first name, family name not family name, first name as it did in previous versions. I don't see this as an issue of an algorithm that converts names correctly, nor a cultural issue of name order. I live in Japan I know all about that. These things have already been decided before the content server is generated. The change has come in which value the content server is using to sort via author and First Name, Family Name is a system that quite frankly is not in common use.

I feel my only option now is to downgrade to a 0.6.x so that it worked in a logical manner.

I just added a new book to my library. It correctly found the Author's name (in this case Peter Stone). The Author(s) has been set to Peter Stone and the author sort has been set to Stone, Peter. When I sort in Calibre via the Author(s) coulmn, Peter Stone appears with all the other authors whose falimy names start with an 'S'. When I look for this book in the content server however it is listed with all the books written by authors whose first name start with 'P'. This is what I mean by inconsistance.

If I manually edit the meta data so that the Author is 'Stone, Peter' Calibre attempts to change this to 'Peter, Stone,'.

[kovidgoyal]

Once again:

author sort is associated with a book not with an author

Maybe you need an example

The book Nightfall is written by Isaac Asimov and Robert Silverberg

It has (in my library) the author sort value of Asimov, Isaac

Now obviously while asimov, issac is the correct author sort for isaac asimov, it is not for robert silverberg.

0.6 presented authors sorted by last name by using a simple hack that worked for people from some cultures but did not for others. I decided to get rid of that hack.

[TarquiniusX]

Ok, I understand your example, and I see what you have done. I simply disagree that it is the best solution when you already have the sort value in the app.

I don't feel that looking for Nighfall under 'I' for Isaac is nessessarily an improvement on what we had before, particularly when you perfer to have it sorted as Asimov, Isaac.

I'll stop bothering you on this issue and simple re-edit my library.

[Worldwalker]

How about a config checkbox to do it either way?

[Starson17]

I use the content server a lot, but I had not seen this change. It's probably because I usually search for a specific set of books (by the same author) or I sort by title or series, not author.

Nonetheless, this change does seem odd. If I sorted on author in the content server, I suppose I'd expect it to use the author_sort field, as it does in the GUI. I don't think I'm sorting authors when I select that field, I think I'm sorting books, and I've come to think of that sorting as being controlled by the author_sort field. In Kovid's example of Isaac Asimov and Robert Silverberg, I don't expect the author sorting to show anything under R or S. Previously I would have expected it to sort under A, now it seems to sort under I. It doesn't sort under both authors.

Respectfully, I don't understand this change either. If the author_sort field is available to the content server, then it's simply a matter of which field is used to sort when the author field is clicked. I'd prefer the author_sort, as that's what I expect.

If there was a "tweak" allowing the content server author sorting to be by author_sort, or a checkbox option, I'd use it.

[chaley]

Originally Posted by Starson17

Respectfully, I don't understand this change either. If the author_sort field is available to the content server, then it's simply a matter of which field is used to sort when the author field is clicked. I'd prefer the author_sort, as that's what I expect.

If there was a "tweak" allowing the content server author sorting to be by author_sort, or a checkbox option, I'd use it.

No can do.

I will repeat what Kovid said in different words and somewhat more technically. When calibre produces a list of authors, it does so by scanning the authors table. At this point we have authors, but no books, so we don't have author_sort. Sorting will be by author, as it is written. Now assume you want the books by an author. Calibre looks up all the books that reference that author. We now have a list of books, each with an author sort, but what can we do with it? Nothing, because the list is book within author, not author within book.

There are only two ways to solve this problem correctly: 1) store authors so that they produce the order you want naturally (this is what I do), or store an editable author sort field for each author *in addition to* each book. The first the users can do, the second takes development.

We could argue about whether the hack that inverted author names should have been left in. Although he didn't ask me (and there is no reason he should), I happen to agree with Kovid that the hack shouldn't be included. It assumes that names are 'fn ln', which is incorrect for a sizable piece of the world.

I'll stop now, or I will yet again get into a rant about name order assumptions and difficulties ...

[Starson17]

Originally Posted by chaley

I will repeat what Kovid said in different words and somewhat more technically.

Great. That's what I'm looking for - understanding, not a discussion of what it should or shouldn't do, but an understanding of the technical constraints.

When calibre produces a list of authors, it does so by scanning the authors table.

At what point would it be doing this? Say the content server has a search active for all books with title:Nightmare. Won't it find those book ids first? Then suppose the user selects the author column to sort. Doesn't it know the author_sort for those book ids? I assumed step 1 is to always find book ids, not authors.

At this point we have authors, but no books, so we don't have author_sort. Sorting will be by author, as it is written. Now assume you want the books by an author. Calibre looks up all the books that reference that author. We now have a list of books, each with an author sort, but what can we do with it? Nothing, because the list is book within author, not author within book.

I assumed it has the book ids for all matches in the author search. This seems like a simple SQL query. Obviously I'm missing something.

There are only two ways to solve this problem correctly: 1) store authors so that they produce the order you want naturally (this is what I do), or store an editable author sort field for each author *in addition to* each book. The first the users can do, the second takes development.

If Calibre was doing it this way in the content server, then the first step would be to search the authors table. Let's say that we search for author:tom and we have a book that has authors = 'Tom Jones & Tom Smith' We'd find both authors, then we'd want the books for those authors, and we'd find the same book twice. But that's not what I see in the content server. Sorting by author doesn't give you two references to the same book merely because you get two matches on the two author names. That implies to me that you start with the book ids, not the authors and you sort the books.

What am I missing? Do you need to point me to code where this is done?

Let me ask it another way - simpler - if it's done in the main GUI, why not in the content server? Is there a hack in the GUI that's been removed from the CS?

I'll stop now, or I will yet again get into a rant about name order assumptions and difficulties ...

Polite and informative rants are always welcome

[chaley]

I think I am confused. I thought we were talking about the authors list available in the OPDF section of the server, where we can choose to see authors, then books by that author. Are we talking about the main book display? If so, then my arguments are not relevant.

[Starson17]

Originally Posted by chaley

I think I am confused. I thought we were talking about the authors list available in the OPDF section of the server, where we can choose to see authors, then books by that author. Are we talking about the main book display? If so, then my arguments are not relevant.

I'm talking about the main display of the content server. I don't use OPDF and I haven't even looked at the /mobile server format. I just realized this behavior had changed when I read this post. It's not a huge thing for me, as I usually search for a small subset of books, but I've been considering a switch to your LN, FN tweaked format for a while. If there are technical reasons for this limitation that can't be easily overcome, it might be enough to push me to make the switch.

Edit: the /mobile server also sorts by FN LN.