Rebuilding tagwatch

I'm going to be rebuilding the tagwatch feature, which was implemented in a way that wasn't scaling well.

So, since I'm going to be doing it from the ground up... what do you want?

I think you want:

1. To show books recently tagged X on a tag page
2. To watch books tagged any number of user-specified tags from your home page

Anything else? Any caveats?

If you're opening this up to little things:

Ideally, I'd like a way to view tagwatch without books I already have in my library (or an option to do it with and without) since I mostly used tagwatch to find new books.

If not, tagwatch didn't have checkmarks to show whether or not I'd cataloged the books. If that can be added, it would be really nice.

Other nice to haves:

- On the tagwatch page only show me the tag(s) I was watching that triggered the display of the book and then give me an option to see the rest of the tags the user added. Some of the more compulsive taggers (I'm not throwing rocks in glass houses) would really give me a lot more info than I ever wanted.

- Can you show me the first instance of a book with said tag for display that day (with an option to see anyone else's if I am so inclined)? When a new book was released and had the tag listed my multiple users, I'd often see 5-10 of the same book. I had zombies up as a lark for a tag, but whenever a new trade for The Walking Dead would drop, my tagwatch would be flooded with nothing but the new trade...

Yes to both 1 and 2.

I liked having it on the home page, and it would be very cool if it would either not show books I already have in my catalog, or put check-marks on them. As stephmo said, its most interesting use is to find new books.

P.S. Yay! I've missed tagwatch.

Sorry, I do not remember what tagwatch did. Though I assume from the name that it involved watching tags.

Yeah ! I miss tagwatch. It was the best way to find new books.

1. In the past you could enter in the tags you want to watch, which I would still like to do. I don't want it to just use the list on my homepage.

2. I would like to have only one copy of the book listed, even if multiple people use the same tag and add it.

2a. Not sure if just adding, maybe people moving it to Currently Reading too ?

3. I would like to be able to enter a list of tags to ignore like: Children's Book, Comic, Graphic Novel, and others.

4. I would like it to only show me books in English if that is possible, or a way to set language if others on the .com site want multiple languages.

5. In the old TW, it listed the tags on the book - PLEASE, PLEASE, PLEASE set it so that it only displays X number of tags with a show more link. Several times I got books with 17-32 PAGES of tags, and it made the listing unusable until they went away.

6. I would like to update regularly. Before it was taken down it often stalled on the same books for days/weeks.

7. Showing a check next to those I have in my Library would be nice.

Thank you for working on it, and I hope to see it soon.

Tag watch sounds like something I'd use to find recommendations. I like most of #6s suggestions but confess to ignorance about quite a few of them based on my not recalling what it was like before.

And yeah, I'd like to use it to see whose reading what within a tag. That would be cool.

This is great - I've really missed tagwatch.

I'd second everything FicusFan said in #6. I'd especially like to be able to exclude tags, have only one copy listed, and see a check for ones I have cataloged.

Like stephmo, I used tag watch a lot to find books I'd never have heard of otherwise.

So has this project been kicked to the curb in favor of a small reworking Add Books ? Is this going to be picked up again ?

http://www.librarything.com/topic/95558

Hmmm... tagmashwatch?

Question is, would it be worth it (site-development-wise, I mean) to "watch" for individual entries that had a particular combination of tags? (It would be nice to be notified when a work previously tagged A by 26 users had now acquired the tag B from someone who recently added it, or went on a tagging spree, but for all I know it may be near-impossible to code, or cause the site to slow down too much.)

>10 235711:

I am thinking along same lines, but can't recall if tagwatch did this already in the first incarnation.

I'd like the ability to watch single tags along with specified tagmashes, perhaps signified A, B, C, A+B such that I would not see results for any A+C or B+C mashups.

11>

I don't think I understand what you want when you say "not see results for A+C".

So, you would want to see a book tagged A, or one tagged C, but not one tagged with both?

(edited to add: reply to msg 11:)

It didn't.

Maybe it'd have to be changed to putting each tag or tagmash on an individual line:

A
B
C
A, B

FicusFan summed up what I'd like to see in message 6. I want to emphasize her request for limiting the number of tags displayed (with "more" link). I found "tag spam" very annoying, but otherwise enjoyed tag watch, and look forward to a new and improved version.

12:

I think the focus was on the way things-to-be-watched-for would have to be separated (how the system reads commas), not on excluding those combinations.

>12 lorax:

Yes, I wanted to see a book tagged A, or one tagged C, but not one tagged with both.

The key concept here is that I'd want to be able to set the combinations that do return hits, each time I search. I'm not saying I never want to see the results of A and C, simply trying to point out that in a search with 3 terms, I'm interested in some of the combinations but not all of them.

So I'm proposing the nomenclature of A, B, C, A+B. Typing in "A+B" ensures I don't automatically get a tagmash result set for all the possible permutations of entered terms, it only gives tagmash results for the combinations I specify ... in this example, A+B but not A+C or B+C.

That nomenclature won't give you want you want, though.

What you really want is

A,
B,
C,
-(A,C)
-(B,C)

so that you are explicitly excluding anything that includes both A and C. Otherwise just having A and C listed individually will automatically include works tagged with both.

Ah, that works. And after I mulled your original question, even if my proposal worked I could see that it would prove annoying to those who did want every permutation.

Your solution is elegant, and probably not any more difficult to implement or explain. I lend it my imprimatur.

I agree with Ficus Fan 100%. I sorely miss tag watch as it was one of my favorite and easiest ways to find new and interesting book and libraries. My book purchases have dropped off considerably since tag watch has been out of commission. Please even it you put the old tag watch back up til you have the new one ready??? It would be so appreciated. Thx :-)

I agree with Ficus Fan 100%. I sorely miss tag watch as it was one of my favorite and easiest ways to find new and interesting book and libraries. My book purchases have dropped off considerably since tag watch has been out of commission. Please even it you put the old tag watch back up til you have the new one ready??? It would be so appreciated. Thx :-)

Bump. I want tagwatch! It was a great way to find books of interest.

(And can we still have our old settings when the rebuild is done?)

Me too, it was a great way to find both interesting books and interesting libraries.

Yes, I really REALLY miss tagwatch.

Bump, bump, bump

*perks* Did someone mention tagwatch?

Yes, I miss tagwatch too!

I miss it, too - anxious to get it back.

A great way to find both interesting books and interesting libraries.

My only complaint about it was that I often saw the same titles over and over when they were found in several different libraries. Could there be a way to red-X a title I don't want to see again?

Missing it too.

I;ve never seen it so I guess I don't know what I'm missing.

A lot. Please, please bring it back.

I miss it too. We've been working on performance issues—going down three nights in a row, and going over queries and structures with a fine-toothed comb—precisely so that I can even think about bringing it back. We made some big progress in the last few days, working the problem non-stop. When it comes back, it will be stable, fast and not a risk to the speed and stability of LT generally.

I do think it's worth noting that Tagwatch isn't particularly high up on the list of user priorities that I put together from the polling data--#68 as opposed to Tag Mirror at #18.

Thanks for the data. I hadn't looked.

Here's the whole list as of a few days ago.

>32 _Zoe_: Many of us have problems differentiating between tag "watch" and "mirror". It's the same package and a vote for one is a vote for both.

If I must choose, I'd personally say "watch" is the most useful.

It's the same package and a vote for one is a vote for both.

I don't think that's a safe assumption at all.

@35,36

No it is probably not a safe assumption. At least I don't care much for tag watch but would love to get a tag mirror.

And it's just the reverse for me.

p.s.: Nothing personal of course. To each his/her own.

If you're into looking closely at things, like person A uses this tag, likes this book or has such and such a collection, you are probably more likely to miss tag watch. And if you're into aggregated stuff, you're more likely to want tag mirror. (If I haven't misunderstood what tag mirror was, since I haven't actually used it before.)

If tagwatch came back, it would give me a reason to use my homepage again.

For what it's worth, even if tagwatch ranks lower, the "will I do it?" calculus is the product of how much something is wanted AND how hard it will be. Tag mirror is a performance nightmare! I am honestly unsure how to bring it back in a sustainable way. Tag watch is more of a performance problem. Problem can be solved. It was implemented wrong last time, and did performance damage. I can implement it so that it does no damage whatsoever.

For the tag mirror problem I can see where that would be a performance hit. My best suggestion would be to only ever calculate it on request (so only those who use it will ever cause it to be generated) and cache it for a week or so, thus it is always available for those who want it and it reduces the calc as much as possible. Of course I'm not in a position to know if this would be enough to make it sustainable but I suspect it's on the road to the best option.

I'd be happy even with a longer caching period (like a month, unless the library size changes by more than X%) for tag mirror, if it meant getting it back.

or even a "please add me to the queue to recalculate when you have time" button...

No, the solution can't involve real-time production. If someone has 1,000 books, that means surveying the tags on 1,000 books. If you have a fairly common library, that can mean examining 20 or 30 million records. That's just not sustainable.

The solution is probably to calculate it offline. We have one box set aside for long-running queries like this--ones that don't impact site speed at all and can be off by seconds or even hours without serious problem. The trick is, we have 1 million users--or let's says tens of thousands with libraries over 1,000 books. Pretty soon, it's the only thing running against that box, or the data never gets updated at all. We could perhaps make it so you only get it if you ask for it--and then it comes in the next day. And we can probably run a number of users at the same time--pick users with similar libraries to cut down on duplicate requesting.

Anyway, you see the problem.

>44 TimSharrock:

Right.

I would be more than happy to see tag mirror on-request, and with a delay. I for one don't have much interest in a tag mirror, although I can see why others do, and I'm an active user who'd be happy to leave mine turned off.

I'm sure all the inactive users wouldn't mind, either. ;) It'd certainly cut down on the number of libraries it needs to calculate.

I wouldn't mind having to wait for my mirror to be calculated or it not being properly updated. But I sure can see the logic behind doing the tag watch now and not focusing on the mirror.

Just agreeing with other people here: Tag Mirror on request, with a delay, would be fine. And then after you've calculated it once, just keep showing that one until you get a request to recalculate. Limit the recalculations either by time (no more than once a week? once a month?) or by change in library (x new books, x percent change, whatever), or some combination. And show the date when it was last calculated.

On the subject of difficulty vs. wantedness: can you just go ahead and make the change to the Blurbers description, so that people know what it's for? No programming required at all. http://www.librarything.com/topic/95046

Yes, sounds fair enough.

Probably not fair enough unfortunately: Users will probably cause way too many recalculate requests.

I'm wondering: If you have a pretty stable library just adding a few books every month, then isn't it the case that tag mirror actually hardly changes for you given the large number of tags in the system? If that's so, then you bascially need to be able to detect (and recalculate) the libraries in which large numbers of changes are taking place and all other recalculations can be spread out over several months.

Since tagwatch was kind of my idea you could enable it just for me and everyone else can have their icky tag mirror. ;)

No. I don't want the icky Tag Mirror, nor would I know where to find it if it did exist.

I want Tag Watch too, not some Tag Substitute.

no, hell no, i want tag watch. that was useful to me every single time i logged on. tag mirror was an occasional thing to look at for fun. i never really got the point of it. but tag watch was .... so good.

Me too; tag mirror was interesting, and if we can get it to work acceptably that will be nice, but tag watch was really useful, and I looked at it every day, and often followed up on what showed up there.

Probably not fair enough unfortunately: Users will probably cause way too many recalculate requests.

But Tim can just disallow "too many" recalculate requests, for whatever "too many" is. The recalculations can even be disabled entirely at times. It would be basically the same system we have now, except instead of showing nothing, it would show the old tag mirrors that had already been calculated long ago.

I miss TagWatch. Seems like I've been seeing "TagWatch will return shortly" on my LT home page for months and months.

Maybe there isn't much demand for tag watch because nobody knows what it is because it's been gone so long. Or at least I have no idea what it is. Just a broken box on my home page.

If you watched a tag you would get notified if anybody entered a book so tagged. I found out about some rather nice books that way.

Thank you for the suggestion, Matt, but I don't think I can change my LT settings to "watch" particular tags - because Tag Watch has been unavailable for several months. If I am mistaken, please let me know.

Sorry, I was merely explaining to lemontwist (#58) what Tag Watch used to do, and why I liked it.

Well, back in September 2010 Tim said (among other things):
...the "will I do it?" calculus is the product of how much something is wanted AND how hard it will be...Tag watch is more of a performance problem. Problem can be solved. It was implemented wrong last time, and did performance damage. I can implement it so that it does no damage whatsoever.(Post 41)

So let's keep on confirming that Tag Watch still IS wanted. More so for every month that passes without it.

> 62 Fair enough.

I want Tag Watch back.

And I!

Well... I have reservations. I'm not sure I could afford it coming back!

Ah, yes, good point. Some form of budget control could be handy.

65: Only look at it when you have spare cash? :)

No, no, I can't have LT looking at my bank account. If it kept telling me how many books I could buy every day I'd never make it to the grocery :)

I want Tag Watch back, please.

I want Tag Watch back, please

I want Tag Watch back.

I want Tag Watch back.

And can we get tags on Talk topics, and watch those too? :-)

#73 YES!

>65 majkia:-68 I think that TW could be improved with the addition of a symbol similar to the coloured check marks for ownership.

Say we use a currency symbol in a circle, like this - ($)

Then, after checking up on you online, it could colour the background of the symbol as follows:

Green:- You can afford this
Red:- This is way out of your league
and the intermediate stage could be,
Purple:- Your cash on hand won't cover this but if you sell some non-book assets (Grandma's china dinner set, Grandma, a kidney, children etc) you could afford it. Go on, you know you want to.

Furthermore, I think Carthage Tag Watch must be destroyed return.

75>

Furthermore, I think Carthage Tag Watch must be destroyed return.

That's wonderful, and I'm very tempted to steal it.

That's fine, I think Cato the Elder's copyright has run out!

#77 Given the new copyright laws, don't be so sure....

Cato was decidedly before 1923, but translations are copyrightable.

Well, if justjim translated that himself, all bets remain off...

Furthermore, I think Carthage Tag Watch must be destroyed return.

Well said, justjim!

I would REALLY like Tag Watch back!

Another 8 months, LibraryThing has become faster in several respects, flashy new search functions abound. Time to clamour for the return of TagWatch again.

I still miss it. I still think it was the second most useful LibraryThing function ever, right after catalogueing itself.

yeah, i totally loved tagwatch, and found many, many books that way. and many possible new connections, too!

alas, I've given up hope...

85>
Well, back in 2010 Tim described TagMirror as a nightmare while TagWatch was a mere problem that could be solved. TagMirror is back, so there's always hope...

... Which sort of annoys me.

Tim *started* this thread about rebuilding tagwatch, and hasn't posted in it since September of last year, almost an entire year ago.

If TagMirror was a "nightmare" and yet it's been reimplemented, why is TagWatch being ignored?

They sped up the basic tag system, with the coincidental result that Tag Mirror was no longer a nightmare. That change unfortunately didn't help Tagwatch.

Okay, so we still have a ways to go before the "two weeks" is completed...

88>

We know it's two different tasks. But they solved the nightmare task. Only the "mere problem" remains, it being a beast or not :-)

Please bring Tag Watch back.

I'm watching for Tagwatch.

Please, please, please.

...with sugar on top. :)

(Personally this isn't a need for me. Rather see the tags applied without checkmarks issue looked at but I couldn't resist.)

I still miss tagwatch. And its now 2013. Seemed fine the way it was. Can't you just add it back in?

re Tag Watch...

I'm just fiddling with this for the first time. Would it make sense to have it exclude one's own use of the tags in question? For instance, I entered a not very common tag which I use a lot so that I could see who else uses it -- and I have to click through pages and pages and pages of my own entries in order to find the others. I already know that I use it, so that info is just noise.

97 > I think I answered you on another thread.

The short answer is, Tag watch is to alert you to new instances of the tag being used. If you want to see who else is using the tag already, the Tag page for that tag should give you that info. Then tag watch can let you know if someone uses it again.

You did, but I didn't see it until after I had already posted this here. I do want to monitor the latest instances of use -- but I still don't see the point of being shown my own recent uses of it. As I said, I already know about those. Ideally, I'd like to see Tagwatch as it's intended with myself filtered out. But maybe that's not possible.

Thx! Lost in time & space as usual here.

99>

Unless you're entering a whole bunch of books with that tag all at once, though, what you see should mostly be other people's books. (Unless it's sufficiently obscure that it has to reach back months and months to find enough usages for the module, I suppose!)

Both of those are often the case with the tags I enter in tagwatch...