Oh hai. So I finally took the time to push the updates that were done two months ago online. This should cover quite a few of the feature requests so far. The news:
Size filter tag:
==========
It's now possible to filter on image size, using size tag.
Examples searches:
Resolutions:
http://anisearch.net/?q=claymore+-size%3Asmall
Advanced search:
=========
It's now possible to negate search terms. Adding -uniform to the search "school", will remove all images with the tag "school uniform".
Test: http://anisearch.net/?q=school+-uniform
Support for OR queries. Searching for "toradora ~mittens" will return all images tagged with mittens, even though they're not tagged with toradora.
http://anisearch.net/?q=toradora+~mittens
Query expansion / Aliased searches:
=========
Searches are now expanded using alias list. Searching "neko mimi" will now also give you results for "cat ears". Internally this is converted to "(neko OR cat) AND (mimi OR ear)".
http://anisearch.net/?q=neko+mimi
I think that's all, i forget. Bug reports/comments are welcome. :3
great. great indeed though I can't see any reason why you didn't post it two months ago. but it's great to see it finally online! big big big thx 2 u!
however, that thing about aliases - is it a complete list of expanders available?
hmm, it all looks nice as far as I see with the exception of one thing: was it too hard to implement standard screen dimensions like 1600x1200 etc.? I believe it's a matter of one line of code but it's not very convenient without it :(
thank you again!
There also was a good idea about Cooliris. Any chances?
don't you find strange this:
* Only large images: claymore size:large
* All but small images: claymore size:large -size:smallwhat's the point in refining small images since they weren't included anyway because size:large already returns "Only large images"?
I guess the example should be just "-size:small" without "size:large"
>>3
Glad you like the update :)
I use 1byte to store image size information in the index, so there are 8 flags available.
4 flags are in use (small, medium, large, unknown), so 4 are still available. Should I add size:x-large (or size:huge even) ? What minimum resolution should that be?
>>4
That requires (probably minor) database changes, altering of the crawling code and full recrawl of metadata for all the sites.
Full recrawl needs to be done, as some results are beginning to become outdated. I will add cooliris when I do this.
This is a boring task though, so no time frame.
tl;dr version: it will be added eventually :)
Is this twit news a good idea? http://bildr.no/view/363901
People I've asked seem to think so, but like "Did you know?" that's already on the front page better.
Twitter account is http://twitter.com/dawl0
Wow, fast reply.
>I use 1byte to store image size information in the index, so there are 8 flags available.
>4 flags are in use (small, medium, large, unknown), so 4 are still available. Should I add size:x-large (or size:huge even) ? What minimum resolution should that be?
I see. Well, if we think about 'size' as a tool for searching wallpapers (I believe it will be 90% used for this task) than I'd recommend changing 'small' since images below 1024x768 (but not including) don't make sense imho thus having 5 flags available we can do something like this (I keep in mind standard 4:3 screen dimensions):
# tiny: 800x600 or less
# small: 1024x768 down to tiny
# medium: 1280x800 down to small
# medium-wide: 1280x1024 down to medium-wide
# large-wide: 1680x1050 down to large
# large: 1600x1200 down to large-wide
# huge: 1920x1200 down to large
# unknown.
Well, if we look deeper than the best idea may be to do 'exact sizes' since if you search for a wall you don't need some weird size like 1508x1187 but it'll still clutter up the search. so it's like
# tiny: 800x600 or smaller
# small: 1024x768
# medium: 1280x800
# medium-wide: 1280x1024
# large-wide: 1680x1050
# large: 1600x1200
# huge: 1920x1200 or larger
# unknown - all other sizes greater than 800x600 but smaller than 1920x1200.
In this case maybe names should reflect the size they match to. It seems to me like a perfect solution.
It'd be great to see this working :)
>Is this twit news a good idea?
Well, my opinion is that the present system looks better by design. "what's new" where it is now also looks better to me than twitter. it somehow doesn't correspond to the design imho.
Oh, I just thought of another thing. Currently it's not possible to search without specifying any tags. Since we got size I think a good idea is to remove this limitation.
thank you for your work.
p.s: if, by any chance, you'll need some help with the site I might assist you in php/mysql/apache/js. just drop a letter to mailme .at. smtp dot ru.
>>9
Thanks for your thoughts :). First off, things need to be kept simple for the lazy average user (like me). Small, medium and large are simple keywords to remember, so these should be used to cover most users needs. Is 1024x768 really considered to be small, and not medium? If that's the case, I don't mind changing that, and adding tiny.
I don't think exact sizes are a good idea, it filters out to many images. I for one don't use the size filter for wallpapers, but mostly to filter out small images.
I need the unknown tag for sites that don't provide information about image dimensions, it can't be used for anything else.
So we have the basic tags small, medium, large and unknown. That leaves 4 tags that can be used by more advanced/demanding users.
So should these be at least medium-wide or large-wide, as you suggested?
small-wide: ???
medium-wide: 1280x1024 down to small-wide
large-wide: 1680x1050 down to medium-wide
tiny or huge: ???
> Oh, I just thought of another thing. Currently it's not possible to search without specifying any tags. Since we got size I think a good idea is to remove this limitation.
This would actually not be removing a limitation, but adding a rather large feature. I'd have to create a separate index to look up images by filters, rather than just looking up tags in the inverted index and running the filter on them.
I think I'll drop the `twit news' from the front page indeed.
Thanks for offering assistance, but I think coordinating work would be more of an effort than just hacking on the site once in a while. :) Btw, this site is written mostly in perl + some data structures and the inner loop of search written in C.
Hi!
>Is 1024x768 really considered to be small, and not medium? If that's the case, I don't mind changing that, and adding tiny.
Sure it is :) I last saw a 1024x768 display about 3 years ago and I believe it's a relic now :)
>I need the unknown tag for sites that don't provide information about image dimensions, it can't be used for anything else.
Ah, I see. I'll keep that in mind.
>So should these be at least medium-wide or large-wide, as you suggested?
I think not, I suggested that thinking of 8 tags. Now I think 4 extra tags should cover:
* (tiny: 800x600 or less)
* (small: 1024x768 to tiny)
* (other (or big or ...?): larger than 1024x768)
* (unknown)
* medium-wide: 1280x800 to small
* medium: 1280x1024 to medium-wide
* large: 1600x1200 to medium
* huge: larger than 1600x1200
This is how I see it. So we have 'other' as an alias for medium-wide thru huge. I think this is how you see it.
Also, what the engine will do if it encouters 1680x1050 size? 1680 is larger than large but 1050 is smaller. Will it be considered huge?
>I'd have to create a separate index to look up images by filters, rather than just looking up tags in the inverted index and running the filter on them.
I see. Well, if you have index based on tags than maybe a pseudotag like 'random' can work? It may either pick a random tag and work as usual or it may pick a random row an go filter it till the end.
>I think I'll drop the `twit news' from the front page indeed.
I really like the design you've made and twitter was somehow breaking it imho.
>Btw, this site is written mostly in perl + some data structures and the inner loop of search written in C.
Wow, that's up my level. I know C a bit but prefer Delphi for win apps and php seems much friendly for me than perl :)
>Thanks for offering assistance, but I think coordinating work would be more of an effort than just hacking on the site once in a while. :)
Well, just keep me in mind in case you need some extra help (maybe it's another interesting project).
Thanks.
Btw, there's another problem with wallpapers and random searching. If you don't care what anime the wall will be from than currently you may click the "random tag" and enter the desired size etc. but the problem is that most of tags have very few images (less than 20 I guess and the number is less if you're searching for a specified size etc.) so to find a good wall you're forced either to enter some common tag or repeat the above process quite a number of times. Maybe it's a good idea to check whether a search query consists of 'random' tag and than use 'numimgs' or other filter to make a result with exactly 'numimgs' images without caring what the tag was.
Sorry if I suggest too much sometimes :)
p.s: well, there was also not too bad idea about random sorting the result. Maybe it can be implemented too? I believe it's not so hard to do...
Thanks!
Come on dawl0, don't give up at the point it's almost complete...
What?! Who said anything about giving up? And giving up on what..?
Have you got a full list of aliases?
@dawl0
Sizes still don't work so it's like you're giving up. You haven't replied rather.
>>15
http://board.anisearch.net/kareha.pl/1237998353/
>>16
size filter works fine. I just haven't added any of the suggestions in this thread yet though.
>>12
How about a "random with > 200 images"-link (or another number)?
>I just haven't added any of the suggestions in this thread yet though.
That's why I said you were giving up :)
>How about a "random with > 200 images"-link (or another number)?
Will it select a tag which has > 200 images? If so, I don't think there are many tags with so many images. Am I wrong?
Besides, it just could add images to result based on 'tag = random()' until its size meets 200. This will give a unique result every time.
>>18
My thesis is due 1 June, so hobby projects are a bit on hold until then. I'm not `giving up' :).
> Besides, it just could add images to result based on 'tag = random()' until its size meets 200
That's not really search any more, rather just picking out random images. How about creating a search special:random that picks out random images from the index, and runs the filters (size, rating etc.) on them?
The challenge is to find a good (as in fast) way of picking out lots of random image id's.
The id's are stored as keys in a GHashTable *doc_hash held in RAM if you're up to the brainstorming.
>My thesis is due 1 June, so hobby projects are a bit on hold until then. I'm not `giving up' :)
Ah sorry didn't know that. At December you just disappeared without a word...
>How about creating a search special:random that picks out random images from the index, and runs the filters (size, rating etc.) on them?
Positive :)
>The challenge is to find a good (as in fast) way of picking out lots of random image id's.
I don't think there's another way rather than just walking the array (assuming db won't be rebuilt of course). Maybe it's better to random sort it at first (not sorting the whole index but rather creating an array with random values which represent the ids in doc_hash). I presume ids in your index are continuous. After that the array with ids can be walked and filtered as usual.
Regarding site load, it might be useful not to create such random arrays on every request but watch server's load and at idle times (at night etc.) or by cron make new arrays. Thus on page request you'll need to just pick a random id of that array (I think 10-20 arrays should be enough for (half) a day).
Well, another crazy idea :) maybe rebuilding index in background on site idle. I mean creating a separate index of sizes etc. Thus the whole search process will be boosted.
That's what I think.
Another idea: 'shared' arrays. Like when a user requests a search, random array is only built when the user has seen all other arrays. Built array is stored in ram so when a user who hasn't seen that rarray searches, new array won't be built but an old one will be used instead. Seen arrays indexes can be stored in cookie.
Hey dawl0, are you alive with your thesis yet? :)
only 20 days left till deadline, fuck x_x
also, only 20 days left, yay \o/
Haha, I see, just like me :) Good luck, all the users of Anisearch will pray for you! :)
thanks o/
Is it at all possible to filter out duplicate images? =/
Most definitely. Sorry if this isn't an appropriate place for it, but yes, the ability to filter duplicate images would be absolutely wonderful.
> would be absolutely wonderful
absolutely wonderful would be filtering images by size, as Anonymous have already said
> dawl0 got ate by his mentor... Mouhahahaha XD
thesis delivered~ Can finally get back to hacking
don't forget to tell us when sizes feature is ready =)
ah come on man, is there any activity yet?
hey dude is there going to be some update whatsoever?
am i the only one who wants to enter specific sizes, rather than pre-defined tags?
size:800x600 - returns images that have the specified dimensions.
sizebiggerthan:800x600 - returns images with resolutions bigger than the specified dimensions.
sizesmallerthan:800x600 - returns images with resolutions smaller than the specified dimensions.
wouldn't something like this be most useful? i mean, it wouldn't just be limited to wallpapers, it could also be used if you want to print a poster in double A0 size and need to search for resolutions bigger than 10000 px, for example.
So I'm not alone here!! Yes bud I'm also eager to search by specific size. dawl0 just pretend he doesn't read this thread.
Btw, searching for 3 sizes only will be stiff, more convenient is to use 8 flags as suggested in
>>11
Just updated filters small, medium and large to reflect >>11.
Also added size:tiny and size:huge
The medium-wide thing confused me, so I added filters size:wide and size:regular if the image fits a widescreen resolution and regular respectively. It should be possbie to combine them, as in "size:large size:wide". Haven't tester properly yet, so it might be buggy.
Thanks for the suggestions and nagging ;p.
Off to sleep!
Here's the code snippet for classifying size, open for suggestion on improvement http://paste.kameleon.ws/paste/438/
OH MAN YOU ROCK!
I have tested the sizes: tiny, small, medium, large and huge work, as well as wide and regular (btw, that was a great idea to filter them by additional flag which applies for all size bits).
the snippet seems perfect, we even have two extra size bits left? the only thing I can't get from that snippet is this: what will it do if a size is neither regular nor wide? IMO it shouldn't show such images if any of 2 size:wide or size:regular flags was specified. is this so?
thank you again, dawl0, you're really cool, although a bit slow sometimes but that's fine :)
there are some weird things though, like having 1280x1024 both in size:wide and size:regular. but in general it works fine :)
there's one thing about drop-down list in the search field - why can't it load URL once clicked/chosen (using keyboard) an item? I'm always clicking on "search" after that... at worst a setting could be made. it's a small fix, don't you think so?
actually it's nice how it doesn't auto search because often I want more tags then just listed in the drop down and I can quickly add them with it. Besides, you can just press "Enter" on your keyboard .
okay, I just didn't know how other people search. let's leave it as it is.
Mmm... say, dawl0, can we have a very small addition like opening images in a new window? At least it could be configurable in Settings. It's a bit annoying to use middle button all the time you know :) And I can't use back button because it'll mess up with JavaScript pages thing.
>>45
Since sane browser open up new windows as tabs anyway, that's a reasonable request. Done! :)
Also, I like your ID.
Great! Big thanks to you, dawl0 :D This site is definitely getting wonderful :)
p.s: nice coincidence with the ID.
How about a way to search for images containing just a certain tag? E.G a Search for red_hair shows all the images that have Red_hair in their tags.
Sorry for spamming, but what's with Anisearch? I keep getting a 403 error whenever I try to enter.