Danbooru

Let's stop using bold / italics / underline to indicate tag counts

Posted under General

Danielx21

Case in point: New color lists like list of blue tags, which are formatted with bold / italics / underline all over.

I'm pretty sure at some point years ago I liked using this kind of formatting. This was discussed in 2012 (topic #8229) and 2014 (topic #10591).

But in hindsight I have changed my mind, that does not look so good in my opinion. Some people in those discussions seemed to dislike it as well.

These are some of the problems I found with that kind of formatting:

  • I think it's really distracting. It's jarring to see that much random formatting together.
  • It's easy to get outdated fast. (more on that below)
  • We don't need it. The tag count is literally in all posts, not to mention the tag search feature.
  • The automated tag count found everywhere displays the exact number, not something vague like bold = between 501 and 5,000 posts.
  • Speaking of vague numbers, it does not seem very hard to have a vague guess of how much a tag is used. For instance, I think we already know about the most used tags. I don't need blue eyes to be formatted with bold+italic+underline (blue eyes) which tells me it has more than 50,000 posts. I already know it's a pretty common tag because characters with any given eye color are everywhere.
  • The top tier formatting does not tell us much: blue eyes is formatted as bold+italic+underline (blue eyes), which means that tag has more than 50,000 posts. We use it because we haven't come up with how to format tags with more 500,000 posts. (blue eyes has 687,527 posts at the moment)
  • It does not seem common or intuituive. Are there many major websites or lists using combinations of bold / italics / underline to express vague quantities of stuff? I know Danbooru itself just lists tags everywhere without formatting them like this.
  • It's a bit intimidating to edit. If I add a new tag to a list formatted like that, it seems I will have to check what is the proper bold/italics/underline combination every time.
  • It may be confusing at first glance. People need to scroll down to the end of the page to read the legend and understand what's going on, or else it just looks like random formatting for no reason.

About outdated formatting, let's see tag group:hands which was formatted like this in October 2017. It has all those mistakes after only a year and a half:

To be fair, that list could be updated today, but it also means that all pages with that kind of formatting would need to be updated constantly.

List of blue tags was created recently and seems to be up to date. But for instance blue shorts and blue nails are approaching 5,000 posts. It seems they will soon have to be changed from bold (blue shorts and blue nails) to bold+italic (blue shorts and blue nails).

I'd like to completely remove that kind of formatting from pages like these.

Updated

Reply
iridescent slime

I don't mind so much the bolding of especially prominent tags as long as it's kept well maintained (though it never is, TBH). Anything more than that is overkill, though. I don't know whose idea it was to introduce extra levels of gradation (italics, bold and italics, underlines) but all the added visual noise makes these pages a headache to read.

What's worse than all the extraneous formatting, I think, is the use of curly braces for links to missing wiki pages. This has the effect of shifting the text which is a major problem in a multi-level list because you can no longer tell at a glance which level each item is on. To use the unreadable mess that is tag group:sports as an example, take a look at the entry for baseball cap and notice how the links directly beneath it appear to be indented as though they are subsets of this tag. How is this supposed to help anyone find what tags to use?

Reply
NWF Renim

iridescent_slime said:

I don't know whose idea it was to introduce extra levels of gradation (italics, bold and italics, underlines) but all the added visual noise makes these pages a headache to read.

Can’t say for sure, since it was years ago, but it was probably initiated by me. I think I experimented with one or two character lists by adding this kind of formatting, but it was expanded and altered later to other lists by others.

If my memory is right, in the initial one I experimented on, there were 3 tiers. Italics were used for characters that didn’t really have art and their tag only existed at the time due to being featured on some relationship charts with small profile pictures (in short it was to indicate a tag that exists but without real art). Normal and bold were used to indicate having art, but some arbitrary post population difference between a high count and low count.

Reply
BrokenEagle98

I think I've probably done most of the recent such formatting, though I have tried to restrict its use to meta-wikis only.

As for myself, I kind of like the extra formatting for the tag size, as it gives a certain weight to the tags. Otherwise, all of the tags look the same, without any way to know which ones are more heavily used. Also, for the majority of such lists/groups, it is not as simple as as a search blue_* on the tags page.

Beyond the formatting used for tag size, which is admittedly more optional, the primary use was to identify fixup items, such as tags which didn't have wikis, had wikis but no posts, or had neither a wiki or posts.

The curly brace thing isn't that big a deal to me, but I can see how it might be to others. To be honest, I originally tried adding a No wiki to the right of the tags to indicate such, but at the time the use of the [tn] forced a newline and looked awful. However, that is no longer the case, so I could go back to using No wiki and No posts instead of the surrounding characters.

Example at: List of Blue Tags

Also, as far as the maintenance of those meta-wikis, I hadn't set up any recurring tasks for such, though it shouldn't be too hard to do so. I'm guessing that monthly should be often enough to keep them current?

Updated

Reply
fossilnix

I'd be in favor of getting rid of the formatting because it's hard to understand and quickly out of date. If we keep it, it should be mechanically updated (it's far too much work for users to do by hand). and maybe instead of text styling, it'd be easier to read plain old numbers

Reply
Danielx21

BrokenEagle98 said:

The curly brace thing isn't that big a deal to me, but I can see how it might be to others. To be honest, I originally tried adding a No wiki to the right of the tags to indicate such, but at the time the use of the [tn] forced a newline and looked awful. However, that is no longer the case, so I could go back to using No wiki and No posts instead of the surrounding characters.

Maybe Danbooru could do what Wikipedia does:

If a wiki does not exist, links to that wiki could be red instead of blue.

I mean the software itself would automatically present links like this as needed.

Reply
BrokenEagle98

Danielx21 said:

Maybe Danbooru could do what Wikipedia does:

If a wiki does not exist, links to that wiki could be red instead of blue.

I mean the software itself would automatically present links like this as needed.

Dynamic rendering has been attempted in the past, but the overhead is so intense that it makes some pages take forever to rend, and it would cause the server to slow down to mush for everybody else.

fossilnix said:

I'd be in favor of getting rid of the formatting because it's hard to understand and quickly out of date. If we keep it, it should be mechanically updated (it's far too much work for users to do by hand). and maybe instead of text styling, it'd be easier to read plain old numbers

Extra text for numbers would be far more distracting than text styling.

Tables could be used...

sometag1000
someothertag2000

...however with tables you lose the easy ability to do hierarchies...

sometag1000
subtag1000
someothertag2000

...plus it looks terrible.

Reply
GabrielWB

Another way would be to use emoji. If you dig beyond the usual Laughing Face and Poop emojis, there are a number of colourful shapes that could be used to indicate the category a tag belongs to.

For example:
❇️ = 0-500 posts
🔺 = 501-5000 posts
🔶 = 5001-50000 posts
🔷 = 50001-... posts

The first entries of the list_of_blue_tags page would then look like this:

...

And hierarchies are still possible.

...

And the great thing is that if we ever get dynamically updated post numbers, these symbols could be moved to that system so they would change dynamically.

...

Reply
BrokenEagle98

GabrielWB said:

...

I like the way you're thinking, however:

1. All emoji don't have universal support in all browsers.
2. No matter how much I stare at the legend, the symbols still mean nothing to me.
3. Consideration must be taken for users which may be partially or wholly color blind.
4. Those yellows are murder on my eyes... and that red-almost-black triangle is irksome...pick a color!

That's the problem with those geometric symbols... they don't carry a lot of meaning.

As an alternative, I propose using numbers based off a 10-based logarithmic scale (starting at 1).

Those symbols are instantly recognizable, and it won't take long to get used to how much weight each number has. Also, for those worrying about running out of numbers, well that won't happen until Danbooru gets over 1000000000 posts, which is a little over 290 times more posts than Danbooru currently has. Hopefully in 1000 years from now, they'll have fast enough machines to facilitate a better scheme.

Edit:

A deemphasis on the rating numbers, removing the tag counts, plus "No wiki"/"No tags" markers.

...or...

Optionally, "No tags"/"No wiki" could be shortened somehow, like "NT"/"NW", "†"/"‡", etc.

Updated

Reply
Danielx21

BrokenEagle98 said:

Dynamic rendering has been attempted in the past, but the overhead is so intense that it makes some pages take forever to rend, and it would cause the server to slow down to mush for everybody else.

If it's not too hard to find, I'd be interested in seeing any old discussions about that. I tried to search the forum for "dynamic rendering" or just "dynamic" but couldn't find it.

Wikipedia and Danbooru could potentially have different circumstances (budget, server issues, how the software works or whatever), but somehow Wikipedia makes the red / blue links work.

Maybe this was discussed before, but I guess the red / blue links could be cached (instead of checking for blue / red links every time someone sees a page). When someone edits or creates a wiki page such as list of blue tags, the software checks for red or blue links upon saving, and keeps the style saved until the page is edited again. It could potentially be out of date sometimes, but that seems easily fixed by editing the page or just clicking "Submit" without doing any changes. The idea of having blue / red links has the added benefit of being easy to ignore in case people are not interested on whether a wiki exists or not.

In my opinion, there's this problem with bold / italics / underline, which is the same problem with emoji and numbers suggested above. I think they draw too much attention. Some people may care about the info presented but others may not, in which case the bold / italics / underline / emoji / numbers may be considered a nuisance rather than something helpful.

Updated

Reply
BrokenEagle98

Danielx21 said:

If it's not too hard to find, I'd be interested in seeing any old discussions about that. I tried to search the forum for "dynamic rendering" or just "dynamic" but couldn't find it.

@evazion did something similar, trying to colorize the links based upon the tag category. It was discussed on either Discord/GitHub.

Wikipedia and Danbooru could potentially have different circumstances (budget, server issues, how the software works or whatever), but somehow Wikipedia makes the red / blue links work.

They have the budget AND the manpower AND the servers. Danbooru pretty much is maintained by one guy and the budget only pays for a scant amount of servers.

Maybe this was discussed before, but I guess the red / blue links could be cached (instead of checking for blue / red links every time someone sees a page). When someone edits or creates a wiki page such as list of blue tags, the software checks for red or blue links upon saving, and keeps the style saved until the page is edited again. It could potentially be out of date sometimes, but that seems easily fixed by editing the page or just clicking "Submit" without doing any changes. The idea of having blue / red links has the added benefit of being easy to ignore in case people are not interested on whether a wiki exists or not.

Gauging by the lack of activity on GitHub overall and especially recently, it's very likely that any kind of "feature requests" will end up by the wayside.

In my opinion, there's this problem with bold / italics / underline, which is the same problem with emoji and numbers suggested above. I think they draw too much attention. Some people may care about the info presented but others may not, in which case the bold / italics / underline / emoji / numbers may be considered a nuisance rather than something helpful.

That's your opinion, so that's apparently your vote. I'm guessing that most people don't care, given the lack of protest in the several years such a system has been in effect.

Reply
7HS

Wouldn't this be an ideal use case for a bot? Wiki pages don't need to list tag counts that are up-to-date to the minute (especially if we go with some symbolic representation of popularity that doesn't include exact numbers,) they could be gone through and updated as infrequently as every week, and the updates could even be staggered - this page on Mondays, this page on Tuesdays, etc. - to reduce server impact.

Also, I think BrokenEagle98's suggestions (in the form shown in the edit) look quite sound and unobtrusive.

I sort of wonder if people would expect smaller numbers to represent more popular tags, though, as if ranking votes in a contest/election. Perhaps all tags in each subsection of a tag list should just be ranked from highest (#1) to lowest (#2, #3, etc...) post counts, so there's less ambiguity about what the numbers represent.

Reply
BrokenEagle98

7HS said:

Wouldn't this be an ideal use case for a bot? Wiki pages don't need to list tag counts that are up-to-date to the minute (especially if we go with some symbolic representation of popularity that doesn't include exact numbers,) they could be gone through and updated as infrequently as every week, and the updates could even be staggered - this page on Mondays, this page on Tuesdays, etc. - to reduce server impact.

That was the idea. I already have something that can process and stylize the links as they currently are.

I sort of wonder if people would expect smaller numbers to represent more popular tags, though, as if ranking votes in a contest/election. Perhaps all tags in each subsection of a tag list should just be ranked from highest (#1) to lowest (#2, #3, etc...) post counts, so there's less ambiguity about what the numbers represent.

Not sure about smaller numbers... but the following is how such would look, using the same Log 10 scale.

If you're talking about using numbers to order them by tag count, then there would be no need for numbers at all. However, such a scheme would drastically impinge the human lookup time for tags, as there's a reason that dictionaries are in alphabetical order. Also, how would one handle subtags?

Reply
Shinjidude

BrokenEagle98 said:

That was the idea. I already have something that can process and stylize the links as they currently are.

Not sure about smaller numbers... but the following is how such would look, using the same Log 10 scale...

I haven't really been following this thread closely, but I will say this last styling looks so much better than anything else in the thread so far.

Reply
evazion

I really don't think there needs to be any kind of special formatting. I agree with everything said in the OP. It's just plain hard to read.

I don't think tag size is that important to begin with. Big tags don't necessarily deserve more emphasis than small tags.

I recently wanted to update tag group:attire, but the wiki was so hard to read, let alone edit, that I ended up removing all the formatting just so I could make my change. Other kinds of formatting would be even harder to edit. Their readability is a bit better, but still not great compared to plain text.

Danielx21 said:

If it's not too hard to find, I'd be interested in seeing any old discussions about that. I tried to search the forum for "dynamic rendering" or just "dynamic" but couldn't find it.

There was some discussion in issue #2755. In the extreme case you can have 1000+ tags on a single page (see list of original characters or topic #13868). Handling these worst case scenarios efficiently proved to be challenging.

That said, my Danbooru EX userscript does have a feature that marks broken links. The way it does it though (using Javascript to add the formatting after the page loads) is kind of hacky and wouldn't be the best way for Danbooru itself to do it.

Reply
GabrielWB

BrokenEagle98 said:

I like the way you're thinking, however:

1. All emoji don't have universal support in all browsers.
2. No matter how much I stare at the legend, the symbols still mean nothing to me.
3. Consideration must be taken for users which may be partially or wholly color blind.
4. Those yellows are murder on my eyes... and that red-almost-black triangle is irksome...pick a color!

While I'm of the personal opinion that people who use devices/operating systems/browsers that don't support Unicode 6.0 from 2010 should really consider upgrading, I do agree with those other points. I didn't consider readability of colours for some people.

Using a number scale for scalability is pretty ingenious, but I do have to ask if it's needed to differentiate between the larger numbers (say beyond 100000 posts)
1girl is the largest tag on Danbooru with 2392256 posts. It will break ten million eventually, and it would certainly be a major milestone, but I don't think that it would be valueable to create a new category just for that when it comes to tag counts.

The great value of pages like list of blue tags is that it shows tags that are underused and I think that the focus should be put there. In other words:

evazion said:

I don't think tag size is that important to begin with. Big tags don't necessarily deserve more emphasis than small tags.

I would argue that tag size is important, but for the inverse reason, namely that small tags need more attention than larger tags to make them more used.

I still think that using symbols is far more readable than numbers, so I would like to propose this version by stealing taking inspiration from an idea posted by @Shinjidude back in 2012 in topic #8229

Shinjidude said:
A symbol to the right or left? Maybe one similar to those used for ranking by Consumer Reports? (○◒◓●) or (○◔◑◕●) A numeric value of some sort?

No more colours, and original symbols from Unicode revision 1.1 that should be supported everywhere.
It's pretty much a small pie-chart that fills up as a tag grows.

◎ - 0 ... 100 posts
◔ - 101 ... 1000 posts
◑ - 1001 ... 10000 posts
◕ - 10001 ... 100000 posts
◉ - 100001 and beyond

Which results in:

And the de-emphasized version:

Reply
BrokenEagle98

GabrielWB said:

...

Yeah, I like it. Those symbols are intuitive, and I agree that several of the number categories are probably unneeded. I also like the bounding "|" on both sides when it occurs, as it still draws the eye to the tag without too much distraction.

You gave a good argument for small tags, but I also have one for larger tags. The larger a tag is, the less acceptable it is to have no wiki. Highlighting this helps draw attention to this fact where it is most needed. Granted one could just use the tags endpoint for that, however that entails a large never-ending quest which would put off the most dedicated users. However, taking care of a few entries in one of the List_of/Tag_group wikis is a lot more doable, especially in the spur of the moment, and gives a sense of accomplishment once finished.

I'd like to do up several wikis in that style, so that people can get a good idea of how it works for them. Sometimes it just takes getting used to something before the value of using it is discovered.

Reply
BrokenEagle98

I did up about a dozen wikis, using a deemphasized and a regular style. I added the regular style since my poor poor coder's eyes were having to squint a lot to see some of those small symbols clearly. It's fine to deemphasize them when they are numbers since those are distinct and the human eye is trained to see them in most cultures, but those circular symbols are harder to discern.

Deemphasized style

Regular style

My own preference would be the regular style since they're easier to see, and because they aren't numbers, they don't conflict with the alphanumeric tags making them easy to ignore/filter out if desired.

Reply
Shinjidude

Personally, I actually like the de-emphasized style better, as the "regular style" ones draw my eye automatically which makes it hard to read the text. It takes longer to read the de-emphasized ones, but to me the content of the text is more important than the magnitude of the tag being linked to. It'd be nice if somehow we could suppress the bullets in the lists and replace them with the symbols. I'd also suggest adding commas to the numbers in the legend just for ease of reading there too.

Reply
BrokenEagle98

Shinjidude said:

Personally, I actually like the de-emphasized style better, as the "regular style" ones draw my eye automatically which makes it hard to read the text. It takes longer to read the de-emphasized ones, but to me the content of the text is more important than the magnitude of the tag being linked to.

After thinking about it for a bit, it probably would be better to put the de-emphasized version there by default, since that throws in HTML elements that allow the user to adjust those elements as desired with Custom CSS.

For instance, the following restores the size of the symbols but leaves the grey color, while also shrinking the size of the follow-on markers if they exist.

/*Adjust this to modify the symbols*/
#wiki-page-body li > .tn:first-of-type {
    font-size: 1em;
}
/*Adjust this to modify the follow-on text*/
#wiki-page-body li > .tn:nth-of-type(2) {
    font-size: 0.7em;
}

Both the symbols and the follow-on text can be removed entirely by setting the font-size to 0, or the color can be set to white so that it only shows up when the text is selected.

It'd be nice if somehow we could suppress the bullets in the lists and replace them with the symbols.

That would also be possible with ID'd headers, for instance "[nodtext]h4#stylized-other-text.[/nodtext]". All applicable header IDs need to start with the same ID text, such as "stylized", which translates to "dtext-stylized" when rendered by Danbooru.

/*Adjust the margin-left to move the position*/
#wiki-page-body [id^="dtext-stylized-"] + ul li {
    list-style-type: none;
    margin-left: -0.5em;
}

I did up the Tag Group:Technology wiki that way for anybody that wants to test. The "See also" section at the end shows the regular bullets, which demonstrates that it can be limited by section.

One limitation though is that no elements can lie between the header and the bullet list. I had to move a couple of "See also" text lines to the end of the bullet group, but they should probably have been there anyways.

I'd also suggest adding commas to the numbers in the legend just for ease of reading there too.

I'm guessing something like the following?

Legend:
◎ - 0 ... 100 posts
◔ - 101 ... 1,000 posts
◑ - 1,001 ... 10,000 posts
◕ - 10,001 ... 100,000 posts
◉ - 100,001 and beyond

Edit:

After thinking about it a bit more, I thought maybe putting the vertical bars and spaces inside the deemphasized sections would be better. That way users could remove all styling in its entirety if they wanted to with the following.

#wiki-page-body li > .tn {
    font-size: 0;
    line-height: 1em;
}

I did up the List of Animals wiki in that adjusted style. That style still works for me, so I'd be fine with doing that.

Updated

Reply
1 2