[development] Making a whole site location aware - a design question

[Scott McLewin]

Dave Cohen wrote:
> Consider...
>
> Write a module which stores this data however you see fit.  (Could be a table 
> you create that joins the users table on uid).
>
> Then, use hook_db_rewrite_sql to filter out any nodes which have location 
> information which is not within the users radius.  This way, your work 
> applies to any node with location information, not just events.  And the 
> event (or any other) module does not have to change at all.
>
> If you take this approach, your module is essentially a location-based access 
> control module.  Read up on how node.module uses hook_db_rewrite_sql to 
> filter nodes based on the node_access table and emulate that.
>
>
>   
The use of hook_db_rewrite_sql did what I wanted for making displays 
only return nodes that are near the visitor's location.  I use the term 
visitor here carefully to distinguish from $user - even anonymous users 
are given a lat/lon location based on their IP, a configured site 
default or self-selected.  Every visitor has a valid lat/lon.  Checking 
for anonymous users to control the hook_db_rewrite_sql behavior is not 
an option for me in this circumstance.

There are a handful of side effects to using db_rewrite_sql that I am 
working through.  Some are rather subtle (at least to me).  They include 
the following cases:

While I want displays such as event calendars and taxonomy lists to 
restrict display on location aware node type to nodes which are 
geographically close to the user, I do want search to search the entire 
site.  Without exception logic, search results are restricted only to 
those nodes that are within the geographic search radius of the current 
visitor.

For search to work, the indexer also needs to be able to see all nodes.  
Even hook_cron() triggered page loads end up with a location (the site's 
configured default). Thus the indexer never sees nodes that are outside 
of the default search radius from the site's default location.


For a module that I don't intend to contribute back I'd twiddle the 
order of my module in the system table so that it loaded first, then use 
the module's hook_cron() and hook_search() functions to disable my 
hook_db_rewrite_sql() for that page load (I've got a global to control 
that).  I intent to contribute the location based filter module.

What I am looking for are early entry points in the call path where I 
can disable the location filter. 

For example, to handle data entered into the search form, I tried 
hook_form_alter to add a hidden element with a validation function that 
just disabled the location filter for that page load.  That didn't do 
the trick as it appears that the page reloads after my validation 
function is called (and the page reload resets the global that disables 
location filtering for the page load).

Are there more general purpose entry points that will allow me to 
disable the hook_db_rewrite_sql logic that filters by location in cases 
such as cron jobs and search form entry (I'm certain that there will be 
others), or are there status/state information tidbits within Drupal 
that I can peek at to know, for example, that cron is running?

Scott