Interest match module - how about this?

Well, that sounds like a great idea, and a nice application for a module like this. I was unaware of the location module, and will surely look at it. One of the things I'm "strugling" with is the case where a user want to search for users with similar interests and for instance living in a certain area or being in a particular age group.

One of the reasons I'd like this as a standalone module, is to make it fit in different applications. Getting the match value of two users and placing it somewhere wouldn't be too hard, but to integrate the order of interest in a query on the other hand, might be difficult.

Thanks.

When it comes to apis, I believe a better approach when it comes to searching all users would be to have a usersearch module where this one hooks in, given the administrator and user allows it. The user search could ask for a SQL array from all modules and this would with the current approach provide something like:

array (
 'user_id_join' =>'currentinterests.uid',
 'fields' => array('SUM(ABS(interests.rating - currentinterest.rating)) AS matchvalue'),
 'tables' => array('interests', 'interests AS currentinterests'),
 'whereclauses' => array('interests.uid = currentinterests.uid'),
interest.field_id=currentinterests.field_id'),
 'groupings' => array('currentinterests.uid')
 'orderings' => array('matchvalue ASC')
);

Note: not thought through.

But I really like the idea of modules making apis available. Maybe this could do so, allowing interests ratings to be hooked in where modules want it? And thereby adding to their orderings?

Maybe it should be easy to add a little button "I find this interesting" different places. A boolean value would probably improve search speed, but might make it less easy to gain overview (a list of users interests for instance). And there's the problem of "instant interests" and maybe more fine-grained interest searches.

Yes, there must surely be good applications for this if implemented nicely. It has been suggested by this nice fellow sami to also add to the users interest terms which he/she submits to a site, which is similar to what you suggest, though people of course read more than they write and hence reading would more easily and probably more accurately indicate the users interests. This leads to another problem with this approach only; it isn't "instantly" available. A user can't sign up, rate his/her interest in some subjects and then go about some business requiring interest knowledge.

These things may however make it more difficult, and certainly slower to search for users with particular interests, especially some total interest type search which is currently implemented. Again, caching would be important. Maybe there is a way to say "Yes please, I'll have it both ways!"?

Thanks.

Note: I probably did not understand everything (no CS-guy per now), so please excuse me if some comments are very affected of this.

As I understand it: Nodes and taxonomies may have properties (of integer type preferrably, text not interesting directly, dates, float and booleans of some interest). These may be calculated based on something. For instance, which users use them. If a user with a certain interest profile (to use that as an example) visits a node, the node participate in a certain term and so on interesting properties in this perspective of that node may change. The same could go for vocabularies and terms. They may be similar in different respects and these properties are there to pick that up. With these interesting properties, new ways of ordering and finding nodes becomes available. There may be different methods such as one based on a certain type of user. This particular type of searching could be defined as a closeness measure. This would add a dynamic way of organizing content. And it would be possible to a large extend to organize the content based on the users interests for instance, which could be based on other users with similar interests behavior on the site.

Such a system could really add an interesting gravityforce in Drupal's nature, and it seems very interesting to me (add interest). But could it not lead to excessive configuration. Would one be able to add general closeness measures providing features better suited to do the ordering than the current taxonomy system? "Doing it yourself" with the taxonomy might also provide desired control over having some closeness mesaurement do the job. Not only similarity is interesting when it comes to ordering. Things like what the site administrator wants users to see also play a role. And say one wish to list similar nodes from the same term or vocabulary under one node. To do that, what measure would be used? Maybe hints could be given by looking at the words used, so wouldn't that be better implemented by and suited for the search engine?

Which properties would be useful? I suppose this is an important question, as the gravity, or different gravities would be defined by it.

Where would it be appropriate?