Series deleted when combining works

when I combine 2 works, the series of the less popular work is deleted

It had been the case since the new series were implemented… so I suspect it may be on purpose.

And even before that, there was a warning that it would happen.

I think it's undesirable that series info gets deleted. But yes, what >2 AnnieMod: said.

>3 MarthaJeanne: Yes, but it didn't actually happen before - as far as I'm aware

>4 Nicole_VanK: Understandable is not the same thing as desirable.

And yes, it didn't happen with the old series system, despite the warning that it would.

I've complained about this several times to no avail. It's pretty annoying, especially when both works only have one copy, so LT's pick is which one is less popular is totally arbitrary.

>6 Stevil2001: Yeah. I’ve learned to check after combinations like that. Annoying is one way to call it. :)

Not a bug, but I'm making note of the feedback here so the team is aware! Sorry for the trouble.

>1 supersidvicious: when I combine 2 works, the series of the less popular work is deleted

The less popular work ceases to exist. That's what combination means.

>9 timspalding: Sure. But when the two works you are trying to combine have 1 copy each (or 2 each and so on), you can lose the series attribution altogether... Or if the series is set only on the smaller work.

>9 timspalding: PS: And that may mean losing a lot of publisher series - if the less popular work is the German edition of a book and has the German publishing series, it will now be lost when it gets merged into the big work... Before the new implementation, both series and pub series were merged into the now surviving record.

>9 timspalding: Why would that necessarily mean series data ceases to exist too? CK is not lost when we combine works and that is a good thing

And neither are Other Authors deleted.

It has to either delete it, or add them together. Is that what I should do?

>14 timspalding: If they are different, add them together. If they are the same, only the one remains. Pretty much how CK works for this.

Add them together (without duplicating). Yes please.

So if work a is in series a and work b is in series b and you combine them, the resulting work should be in both series a and series b.

Or if work a is not in any series, but work b is in series b and when you combine the resulting work is work a, it should be in series b.

None of the series data should be lost - it should be combined. :-)

In addition/related to this, the combine warning message (advising that combining will lose the series of the lesser work, but this can be added back afterwards) doesn't really do a good enough job now in the era of new series.

A listing within a series now is more than just the series name and the item label that it was in the CK-series days, there is also the Group and the Position (for a series with custom ordering). Unless the combiner takes especial care to realise, and go and note this info before combining, they are left guessing after the combination, or retread research effort that's already been done once.

Ideally, combining would seamlessly transfer all the series of all the components into the combined work (cf. how all the CK data transfers), but if this isn't feasible, the warning message could be improved by also providing the group and position of the lesser item, or at a bare minimum should warn the combiner to go and note this info down before combining since it will be lost in the combination.

Example:
You are combining A Wonderful Novel (1 copies) into A Wonderful Novel (354 copies).
A Wonderful Novel (1 copies) is in the publisher series ABC-Books, label 24, group "Other Novels", position 44 of 45 items

If the combiner doesn't go and make a note of this info, all they can do after the combination is link to series ABC-Books with label 24 but they don't know what group to select, or where in the series to drag it to. Didn't matter much with CK-series because the only choice was between Core and Collection/selection, but with more detailed groups and custom ordering, we would benefit from improvements in how combining handles the series, to ensure this info isn't being needlessly shed.

Edit to add: yes, if an option to add both series together on the combined work, yes please :-)

>14 timspalding: Yes, please

>14 timspalding: Yes, yes, please combine!

>14 timspalding: Adding my vote to yes, please!

>14 timspalding: Yes, all series information should be kept! Along with order labels and order.

>14 timspalding: yes please keep all the information

>14 timspalding: bump!
series still deleted when combining 2 works

>23 supersidvicious: Please note >8 kristilabrie:. NOT A BUG.
Already been closed.

It's been closed, but the last comment from staff was a request for how it should be handled. This has been answered, so I don't think it's unreasonable to bump it in hopes that he looks at the answers.

It may not be a bug, but it's extremely undesirable behaviour, as detailed by Nevov above.
Please, please change it so that all the data is retained!

bump

Reopening this since @timspalding never got back to his question at >14 timspalding:. Will ping him about it, but I wouldn't expect much movement within the next month or so (staff summer vacations are cropping up).. please ping again in another month or two!

>28 kristilabrie: ok enjoy your holidays

>28 kristilabrie: Are the summer vacations over yet?

bump

I don't really see a problem I can solve here. When you combine works you're deciding that a given work and all of its attendant information is not a stand-alone work, but really part of something else, and that the old work should be killed. Any amount of information is going to be involved in that old work but there's no good way to know which ought to somehow be replicated in the new work and which is as wrong as the dead work itself.

About the best I could do would be to list more information about the soon-to-be-dead work's series information, as suggested in >14 timspalding: and maybe a note about being diligent, which will be ignored by those who aren't already diligent.

>32 timspalding: What we usually lose are publisher series -- a new user adds the book, adds to a publisher series and then it get folded into its actual work. People are supposed to look for the warning and follow up but... it does not happen. Especially when the combiner is a newish member who does not even understand what the warning says let alone how to fix it - they just followed the proposed combinations link on a book for example or found a ratty author list and decided to help.

And unlike stranded links and CK on authors, there is nothing someone can fix later because there is no indication anywhere that something needs fixing. The data is just lost.

And then there is the problem with newish works and works with a very low number of copies (often in a not very popular language but pure English also suffers) where someone does a lot of work on adding series to books and what's not and then the combination loses all the information because the ones without series become the main work.

Why can't we just combine the records for the series: if A is combined with B, then the new combined work (being it A or B) gets the series from both A and B?

>33 AnnieMod:

I assume if work A and B both belong to series X, I ignore B, right?

>34 timspalding: That series for B, yes - if they both belong, the result will still stay in the work. And if they have different labels, there is a bigger problem somewhere.

But if A belongs to X and Y and B belongs to X and Z, the result should belong to X, Y and Z (with the label and grouping from A in X).

>32 timspalding: Any amount of information is going to be involved in that old work but there's no good way to know which ought to somehow be replicated in the new work and which is as wrong as the dead work itself.

You can't just replicate all of the series information for both works and let users remove anything that isn't correct? It's extremely frustrating when combining two works that have one copy each and the series information gets completely erased based on, I guess, random chance.

thank you for clarifying the bug

I just combined the 2 following works

https://www.librarything.com/work/28973793/editions

Take a Chance on Me: Camp Firefly Falls Book 12/Zoe York (no current copies was listed in the Series Camp Firefly Falls as #2

as soon as I combined it with Take a Chance on Me (Camp Firefly Falls #12)/York, Zoe (1 copy separate)

the series disappeared and I had to assign the work again to the series.

Today I did it, tomorrow I could forget it

>35 AnnieMod: Yes, I agree. The assumption that one of the series designations is bad data doesn't really make sense to me. A lot of times when I add a new Doctor Who audio drama, I find that there is a one-copy work created by me, another one-copy work, and a two-copy work. So I combine them all, but a lot of the time it's the person who created the other one-copy work who went through the trouble of adding series data, so that gets wiped out and I have to redo it. I don't see why we assume that the data attached to the one-copy work is bad.

>33 AnnieMod: AnnieMod:

>35 AnnieMod: AnnieMod:

I strongly agree with this.

There is no reason to suppose that the data from either set is bad.
And if it is, it is much easier to subsequently delete the bad data, than it is to add back the deleted good data.

This is especially true, now that it is no longer simply a matter of adding back the deleted series, but potentially also the position of the work within a custom sort, and membership of a group within the series. None of which may have been noted before the deletion is made.

bump

>32 timspalding: "When you combine works you're deciding that a given work and all of its attendant information is not a stand-alone work, but really part of something else, and that the old work should be killed"

Except we have no choice which work gets "killed". LT chooses, apparently at random.

>41 DuncanHill:
>apparently at random

Combining goes by which one has more copies. Not sure what factor decides it on a tie. Lower workID? But does look random when combining 1-copy with 1-copy, as sometimes the series is on the one being kept, great! then other times it's not.

Occasionally the copy numbers make it possible to take advantage, by combining in stages, eg. 4-copy, 3-copy with series, 2-copy. Combining all in one will lose the series, but do the 3+2 first for a 5-copy with series, and this will win versus the 4-copy. But rare it presents conveniently like that.

>32 timspalding: When you combine works you're deciding that a given work and all of its attendant information is not a stand-alone work, but really part of something else

No, it's not "part of something else" (that would be a relationship of some sort); you've decided that the two works being combined are exactly the same. So anything - CK, series, authors - that applied to either original work should apply to the new combined work.

You shouldn't throw data away just because it belongs to the "losing" work.

CK from both is kept and merged into the new work, authors are kept and merged into the new work. The bug is that series aren't merged from both into the new work, but instead the series information from one is thrown away.

Bump. Purely by chance it's exactly a year since the last message in the thread.

>44 r.orrison:

Seconded.
I'm wasting too much of my life, restoring deleted series data.

>44 r.orrison: >45 kleh: Heard. We're busy prepping for the PLA Conference next week, mostly, but feel free to bump again in a few weeks so we can keep this floating! Thanks for your patience.

Bump https://www.librarything.com/topic/362759

Bump https://www.librarything.com/topic/368058#8774733

I don't really understand. There's a warning message about what will happen. We can't keep all the information, because you'd get all sorts of strange duplicates. So we keep only the series information on the winner and tell you this before you combine.

>50 timspalding: Inexperienced combiners do not read notices - or when they read them, they either do not understand them or just ignore them.

Plus it is annoying to have to redo series when you combine works with minimal numbers - i.e. publisher series on non-English editions are almost always lost when someone is tidying up an author's page and need to be redone manually post combination.

Can't we make this a bit more intelligent? If more than one of the works is in a series, sure, keep just the one. But if only the one that gets combined into something else carries the series, just move the series up.

>50 timspalding: Would there really be duplicates? I assume you're talking about if the two works being combined were both already in the same series, but I've seen that happen maybe once in recent memory, and it's easily fixed.

I can only agree with >51 AnnieMod: and >52 norabelle414:. This mechanism hinders more than helps!

You can (and do) merge and remove duplicates for Common Knowledge, why not series?

And even if you can't remove duplicated entries, as >52 norabelle414: points out most of the time this happens the series information is on just one of the works. That's fine if it's on the winning work, but if it's on the losing work it's lost and has to be recreated manually.

Needing to remove duplicates manually would probably happen a lot less often than having to fix deleted information manually currently does.

And if the user doesn't delete the duplicates manually, that's still better than if the user doesn't replace the lost series information the way it currently works.

TLDR: I'd much rather have duplicated entries than lost data.

I am not inexperienced and I have no idea how to combine them and not lose data. It's much easier to clean up if there is duplication of data then it is to figure out what's missing after the fact.

>55 SandraArdnas: I am not inexperienced and I have no idea how to combine them and not lose data. It's much easier to clean up if there is duplication of data then it is to figure out what's missing after the fact.

I'd call myself moderately experienced, and I agree with this. And if LT knows there are conflicts, it can list the conflicts to help us make fixes. That is what would really help: I never know if I'm doing something genuinely damaging or not.

How much extra programming would be required to have the series info retained for whichever work is the first With Series Info?

Obviously wouldn't help if there are multiple works with disparate series. But what I've typically seen is the work showing up second or third in the listing being the one with Series info and none of the others do.

>52 norabelle414:
>53 SimoneA:
>54 r.orrison:
>55 SandraArdnas:
>56 waltzmn:
>57 Avron:

Absolutely.

It is a far more difficult task to restore missing series data than to delete any resulting duplicates.

With the addition of series labels and groups to new series, you now have to take detailed snapshots of the series before combining, which is an incredible waste of time and effort.

I guess most combiners don't bother, and the data is just lost.

It seems a strange procedure, when LibraryThing generally ensures that data is never completely deleted.

The point is that once you have made the combination, you can't reenter the series, because you don't know what they were. They are gone and you have no place to copy them from. This goes double for publishers series in languages you don't know.

I have posted in Recommend Site Improvements about this, in the hope that it will be picked up! See https://www.librarything.com/topic/375056.

Okay, I've changed things a fair amount. When you combine works it lists all the series (and publisher series) in them, allowing you to transfer them or not (with a check box). If 2+ works belong to the same series, the checkbox becomes a radio button, so you can choose the new relationship.

Anyway, I think it's everything you want. Let me know if you have any problems.

>61 timspalding:
Exciting change.

A bug: I'm finding that the label erases, when the lesser work in the combination is in a series, and the larger work is not. The series info shows up fine before combining, on the work combination page Effects section, but then after combining, the work ends up in the series, in the right group, but with a blank label.

Edit: posted this as its own bug: https://www.librarything.com/topic/377665

>61 timspalding:
Also, has this change had an impact on the Too Much Love rule? There's a couple of examples posted in that thread in the couple of hours where it is a big work and a very small work, which should go through combining without triggering the block:
post 120 https://www.librarything.com/topic/370662#9053223 – 374 vs 3 copies
post 121 https://www.librarything.com/topic/370662#9053270 – 10,978 vs 5 copies

>63 Nevov: Oh good, it's not just me.

Not a Happy Family by Shari Lapeña
https://www.librarything.com/work/25490303 (1813 copies)
strays to be combined:
https://www.librarything.com/work/29375919 (0 copies)
https://www.librarything.com/work/32744840 (2 copies)

ETA: >61 timspalding: Cool! Thanks.

>61 timspalding: Oh, hooray! Thank you.

>63 Nevov: Ditto. I've run into this a couple of times.

>61 timspalding: Thank you.

Okay, the "too much love" rule has been changed for now:

The rule is now that it's about how many copies there are in the works EXCEPT the largest one. So if you're combining a 10 copy work into a 1,000 copy work, you will not be stopped. It's only if the lesser works have more than 250 copies that you may be stopped.

To be honest, the rule was originally about copies AND whether your account was free or paid. (Paid members weren't better, but they had more to lose.) Fortunately or unfortunatley, that distinction no longer exists.

I'd like to change the rule to be more flexible. I'm thinking that I check out how many works people have combined, and if more than something, they get free reign. Ideally i'd check how many works they have combined without being disagreed with, but there's no simple way to assess that, as a subsequent separation isn't proof you did it wrong. I think perhaps I'll just add a timing factor--more than X combinations, counting only ones that are two weeks old or older. Most of the problems are with people who come in and go hog wild, and this would stop them.

Thoughts?

See https://www.librarything.com/topic/335019#9054180 for the "too much love" issue. I would love your feedback.

>68 timspalding: I'd like to change the rule to be more flexible. I'm thinking that I check out how many works people have combined, and if more than something, they get free reign.

This risks people doing combinations to up their combination score until they reach the point to do real damage. :-(

I agree with the idea of social responsibility, but I'd also perhaps have a "length of membership" thing (can't do it until you've been on for a couple of years), and maybe let other community service things, like Common Knowledge, count.

Just thoughts; that's not really an algorithm.

>69 timspalding: That links back to here!

Yeah. That was supposed to be posted somewhere else. Sigh. :)

As a newbie, I fully support caution -- *I* don't trust me to do much in the way of big stuff, and I've been trying to pay attention & learn. I'm also finding that splitting/combining is something that I'll do in batches, and that I'll get really good at it, need to do other things for a few months (including hard non-LT things), then need to do more splitting/combining and struggle to get it right again.

I think some combination of >68 timspalding:'s more than X combinations, counting only ones that are two weeks old or older & >70 waltzmn:'s length of membership, perhaps tempering the former into "counting only ones that are two weeks old or older but not more than 1 year old"?

>73 LeslieWx:

I hate having barriers. I might use badges, because they include various factors, like length of membership and activity.

>74 timspalding: I like the idea of using existing badges for that (or even creating a new one much higher than where the biggest one is now). And short of going for a white list of users who can do these, that may be the cleanest way.

Will someone try to trick it so they can do it? Sure. So what? It’s still better than the current way where as long as you are minimally savvy around URLs, you could figure the secret method (my memory is bad for tasks I do not do often and I keep misplacing my notes so when I need one of these, I simply work it out over and over). And it will make it easy for the people that do enough combinations to know better (as opposed to the people who just know how to figure it out).

>74 timspalding: I hate having barriers.
Fair.