Danbooru

[Prototype] User Report Ver 6.3

Posted under General

Well, if you use the ID or not is the same difference. But ok, then I'll look every number up manually :3.
But you have done these tables: What do you want to express with them? I can only see that you want to complain about sniping other users so that user A gets the upload and not user B.

Not exactly the same... having to look something up manually is usually enough of a deterrent to all but the most determined.

The tables are also presented as is... whether they're useful or not is really more in the eye in the beholder. As I mentioned in the first post, if there's in interest in them, I'll continue to collect and finish up the tables. Else I'll just leave that sole post as a demonstration of the data available to collect.

Loiter Time was the total time spent on all of the uploads counted as "Snipes", measured from the timestamp on the Uploads record to the timestamp on the Posts record.

Loiter/Snipe is Loiter Time time divided by the number of Snipes, i.e. the average time on those "Snipes" that were taken before the final tag change was recorded by the system.

Note that uploads not counted as "Snipes" did not count toward Loiter Time nor Loiter/Snipe. (Although that data could be counted separately if so desired...)

So you mean I upload an image with these tags at 17:30:01
"1girl solo artist copyright character looking_at_viewer"
That would count as a snipe. Now I tag this image completely (let's say to 23 gentags) and the time is 17:30:51.
So that'd mean I spend 50 seconds on that post. And these 50 seconds are the loiter time then?

Yeah, you've got the idea. Also, the name "Snipe" is just tongue-in-cheek like I mentioned on the original post. Calling them "posts that were minimally tagged on the upload but properly tagged afterwards" is a bit long... ;)

Well, I'm not really against this idea to use them.
But as I said: "Sniping" is nothing that would cause a bad or neutral feedback. So we don't have a policy against them. It's like "feel free and steal posts from other users that are tagging before".

We could start a discussion again and it was also discussed in topic #8824. But there wasn't a real conclusion after that discussion. But recently I was told that as long as you're tagging afterwards, it isn't perceived as an issue.
As for me: I hate tagging before uploading. With Shift+E it is much, much more confortable.

Yeah, that's right. That's why as I mentioned they are just an interest item and not something to act on. It's also why I didn't include them in the first post in the thread.

However, some might find the data interesting, which is why I posted it. If there is no interest in continuing it, I'll just leave that lone post up there as an example.

Well, it is interesting at least to see how common the tag-after-upload workflow is. I tag after upload because the upload UI sucks, as brought up in the Danbooru 2 thread.

I don't care about sniping, but if you're really interested, you'd probably have to factor in age of the source. Uploading a Pixiv post that's 20 minutes old with no tags may be considered sniping, but uploading some post from 2014 isn't really beating anyone to the punch.

Heh... :p Yeah, it was more of the former then the latter. There's not really a succinct word to use for that kind of activity though, so "Snipe" it is...

I plan to eventually get hooked up with the Pixiv API, but my idea is to use it to help tag artists and find lost golden nuggets...

Edit:

Now that I think about it, the tables could be used as a sort-of-proof for the need to overhaul the Uploads interface...

Tagging after upload is a behaviour I developed avert getting sniped one time too many. Yesterday was a particularly telling reminder for me, having about 8 posts sniped with one or two tags, and even this morning.

I would not mind changes to the upload bookmarklet as long as they don't interfere with use on mobile devises.

Jarlath said:

Tagging after upload is a behaviour I developed avert getting sniped one time too many. Yesterday was a particularly telling reminder for me, having about 8 posts sniped with one or two tags, and even this morning.

I would not mind changes to the upload bookmarklet as long as they don't interfere with use on mobile devises.

Well, just got sniped again this morning, and yesterday, by selecting the bare minimum tags - only to get sniped by someone who selected ONE tag and hit upload. Given this, I'm it sure the report here is accurate, even for that one day.
I kind of wish the system enforced a minimum of four or five tags to help finding uploads (as recommended in the howro:upload wiki), but what can one do?

This is I think tangential to the subject at hand. Could there be any reports to see who is utilizing certain tags? Over the past few weeks I've been watching day grow over a thousand posts and I can't tell besides Provence and I who is pushing it. It would be nice to see who is putting in the effort to get tags to pick off. It would also so people's contributions (for instance some people's enormous contribution to breast tagging).

This was touched upon briefly in the original topic. The top N tags being added through tag gardening could be calculated per person. It didn't get a lot of traction though, so I never implemented it. If there's interest though, I can work on it, and probably have something up by tomorrow.

If you're asking for the tracking of a particular tag, especially on-demand, that might be a little more difficult. I don't know how long that would take server-side, but client side it takes about an hour or so to generate a report for all post/tag changes over the last month. The data could perhaps be duplicated in an alternate table that would optimize it for that type of search, but I'm not sure the amount of coding effort that would require...?

Latest Update

Ver 4.4 -> Ver 4.5

Google Drive

Note: The data only includes tag gardening and not uploads.

I also uploaded a JSON file with the data instead of a CSV file, otherwise the file would be over 100 MB.

The JSON data takes the following format:

Ex: [{"id":1234,"tags":{"solo":45,"1girl"20}},{"id":245,"tags":{"breasts":45,"standing"20}}]

So the "id" field is singular, and the "tags" field is variable length.

@evazion

From forum #118827, I tried importing the data into https://jqplay.org/, but the site just died since the file is about 1MB. I also tried using the Win-32 version, but couldn't get the command-line to work... :(

Edit:

Nevermind, I figured it out...

Updated

For those using JQ on windows and with the JSON file from above...

jq-win32 "sort_by((.tags.day)) | reverse | map(select(.tags.day) | {url: \"http://danbooru.donmai.us/users/\(.id)\",count: (.tags.day)})" usertags.json
... will produce the following output ...

[
{
"url": "user #463832",
"count": 107
},
{
"url": "user #32756",
"count": 26
},
{
"url": "user #460797",
"count": 20
},
{
"url": "user #166417",
"count": 5
},
{
"url": "user #351189",
"count": 4
},
{
"url": "user #444105",
"count": 1
},
{
"url": "user #423742",
"count": 1
},
{
"url": "user #479523",
"count": 1
},
{
"url": "user #467044",
"count": 1
},
{
"url": "user #449675",
"count": 1
},
{
"url": "user #490356",
"count": 1
}
]

It shows all of the users that added the day tag over the last month (via tag gardening, not uploads) sorted by count. Just replace all of the ".tags.day" in the above statement with ".tags.<tagname>", where <tagname> is the name of the tag you're interested in.

Updated

@BrokenEagle98: I'm playing with your JSON dump right now. Thanks for that, there's lots of interesting data in here. For starters I'm using this to get the full list of tags used by a person (same as your tag gardening table in the OP, except a full list):

jq '.[] | select(.id == 460797) | .tags | to_entries | sort_by(.value) | from_entries' usertags.json

Would it be useful to have in addition to the above, JSON files for both the uploads tags and the total amount of tags (tag gardening + uploads)?

The only reason I excluded the uploads originally is because a lot of the standard tags would dominate, such as 1girl, solo, looking at viewer, etc, and so would be useless for a Top 10 Tags table. However, with only the tag gardening JSON, it's impossible to see who's adding all of a particular tag, such as the day tag that chodorov is interested in (since a portion would be added through uploads).

1 2 3 4 5 6 7 8 9 15