We're going to be working on this while Matt is in town this week. Our backend team has said that there shouldn't be any major obstacles to serving "full" User objects (that is, what's returned from /users/show) throughout the API.
On Mon, Mar 9, 2009 at 16:52, Torgny <[email protected]> wrote: > > This is an excellent idea, Blair! This would make it so much easier to > get this information without any additional calls. > > On Mar 9, 6:17 pm, blairblends <[email protected]> wrote: >> When requested a list of friends/followers via /status/friends and / >> status/followers, I've always found it limiting that only the >> "followers_count" is provided for each user, while "friends_count", >> "favourites_count", and "statuses_count" is left off when I consider >> this to be basic user information, not extended. >> >> I understand that we need not know people's theme, style, and >> notification preferences in a "basic" object, but I think simple >> numeric counts like these is vital. >> >> With many of the applications I use, I specifically want the >> application to know this information. As it stands now, acquiring it >> is a total boondoggle and waste of resources. I have to first get the >> list I want, then queue up and individually, one by one, request / >> users/show to get the extended object for each user, when I don't even >> need most of the "extended" information -- just three extra variables. >> >> Not only does this take much, much longer, but it burns tens of >> thousands of API calls a day. Even though my applications politely >> queue it out to stay WELL below 20,000 an hour, it's literally the >> difference between 10 API calls to get that many pages of friends or >> followers, or 10 PLUS 1,000 API calls to get the lists, then make the >> 10,000 individual /user/show requests. On my LIMITED scale, I could >> probably save you 50 to 60,000 API calls a day. >> >> My point is two fold: (a) if followers_count is basic information, >> then I think it's fair to include friend/status/favorite counts as >> basic information, too. Why should one qualify and the other three >> not? (b) for practicality, why not make it simpler for those of us >> who are caching data [which is implicit in the REST philosophy] >> instead of requiring tens of thousands of unnecessary API calls an >> hour. That is, to say, if what now takes 50,000 API calls could be >> done in 500, *shouldn't it*? >> >> Isn't adding three simple numeric values into the basic user object >> WORTH the reduction in API calls by a factor of 1,000? And that's >> just from me! >> >> There may be other solutions, too. Maybe a new call such as /statuses/ >> friends/counts/id.xml which would be the basic object plus these other >> data. Perhaps the CURRENT API calls should default to the basic user >> object for each person in the list, but an option could be added to >> request the extended object? Or, even better (from a developer's >> standpoint), perhaps the current /statuses/friends and /statuses/ >> followers return the basic user object, but you are allowed to request >> that other variables (such as the three I'm adamant about) be returned >> in tandem. >> >> I'm thrilled to have such a robust API, but I'm frustrated on a >> personal level that I have to do such elaborate queuing and thousands >> of extra "calls" to accomplish something that could be nearly >> instant. It's slowing down my potential as a developer, and I can't >> imagine it's very kind to your servers. >> >> Thanks for hearing me out. > -- Alex Payne - API Lead, Twitter, Inc. http://twitter.com/al3x
