Can we ban *pixiv.net/member_illust.php* source URLs? There are 59 pages of source=medium, which is what the URL rewriter serves up anyway, but there are also 23 pages of source=big, which redirects to the pixiv homepage in many instances. In either case, IIRC it was decided by consensus that the source URL of pixiv images should be exactly the source of the image file, not in small part because that URL contains an artist handle which can be used to identify artists.
This is probably a bit difficult, but could those 82 pages of pixiv posts be automatically resourced somehow?
Also, IMO the text for the source link on the post view page should give the real source field in the db, not the rewritten URL, which should be reserved for the href attribute of the anchor tag.
Also in that vein, when adding Pixiv artists, we need to be careful to add both their profile's url as well as the url of one of their images, otherwise the artist-autodetect won't work. I've seen several artists like this.
For the last bit, about 4 months ago I wrote a Greasemonkey script to auto-rewrite the pixiv link to get you to the friendlier "member_illust" page rather than the useless 403 error. You can grab it here if you like:
Your script rewrites the 403 error to the correct page in the href attribute of the source URL link on the post page. I actually used your script for a while, but albert implemented it directly, and took it a step further by replacing the TEXT of the link with the "member_illust.php" link. I was just suggesting that the actual text of the link be retained, with only the actual destination URL rewritten to member_illust.php (which is actually how your script works, IIRC). Otherwise, we need to click Edit to see the database field for "source URL".
Just so we're clear, I do want the "useless 403 error" URL to be in the database. I'm pretty sure this was agreed upon, it just seems that many well-meaning contributors are entering the member_illust.php?mode=medium, or worse, the member_illust.php?mode=big page instead, not realizing that danbooru autoredirects pixiv pages anyway, and to do it manually is not consistent with the inherent semantics.
I agree, the source URL must be the actual jpg link. It would probably help if Danbooru could, for pixiv images, add a Referer (sic) header to the HTTP request, so we can upload images directly by the image URL instead of getting a 403 when trying.
Otherwise, people are encouraged to save the image to disk to upload it, and then enter the URL of the page they have open in the upload form. This will be the =big URL.
To fix the existing ones, it wouldn't be hard to write a script to do that automatically, using the API. Though you have to make one PHP request to pixiv for each image to get the real URL. Not sure if it's possible to query the artist DB at the same time and add the artist tag automatically.
Shinjidude said: Also in that vein, when adding Pixiv artists, we need to be careful to add both their profile's url as well as the url of one of their images, otherwise the artist-autodetect won't work. I've seen several artists like this.
Yes, very very yes. This goes for every artist. Way too many people add artists without a sample image, which means Find Artist doesn't work right.
It's especially useful for Danbooruup, which is remarkably accurate in finding the right artist... IF someone put a sample URL in. I find myself having to add this to artists all the time.
jxh2154 said: Yes, very very yes. This goes for every artist. Way too many people add artists without a sample image, which means Find Artist doesn't work right.
Please help me get this straight. When adding an artist, we have to add the URL to a random one of their images to their artist page? I would never have guessed.
Soljashy said: Please help me get this straight. When adding an artist, we have to add the URL to a random one of their images to their artist page? I would never have guessed.
Yes, the "Find artist" link on the upload form (and apparently danbooruup) uses that information to figure out the artist tag from the URL. If there is no image URL in the artist DB entry, it won't work.
