= MultiClassification =

== Requirements ==

 * unlimited number of classifications for one document (task)
 * unlimited number of assigned aspect values in one aspect

== Solution ==

Separate '''[[Classifications]]''' records are used for every classification which is set. No empty records used, a document has as many '''[[Classifications]]''' records as many classifications are set.


== Issues ==

 1. How to handle auto-classification from the drawing code? E.g. do you create a new (additional) classification, or (all?) replace 
    existing classifications?


=== Detail ===

No problems determined until now. A smarty loop goes through all '''[[Classifications]]''' records to display them line to line.

=== New/Update ===

For '''every''' aspects defined in table '''[[Aspects]]''' at least one line appears even if it is empty. At the right side of all line there is a ''new'' button to open additional line for entering classification in the same aspect, and a ''delete'' button to remove that classification. If there is no more classification in an aspect (after deleting the last one) an empty line appears.

|| aspect X || code || AX_L1_0 || AX_L2_0 || ... || from - to || ''new/delete'' ||
||          || code || AX_L1_1 || AX_L2_1 || ... || from - to || ''new/delete'' ||

=== Search ===

Searching is a performance issue. Because of all '''[[Classifications]]''' are different records a large joins could be used to perform searching, which can be very slow. The following optimization steps will be implemented:

 1. determine first the number of '''[[Classifications]]''' records involved in one aspect value search
 1. perform a nested select with the smallest records set in the inner most level to limit the involved records to the minimum. The result should be saved into a temp table in the database. This temp table should be deleted at every new search or list.
 1. read DrawingsRevisionsJoinSet with the id set got from the select above

E.g.: assumed that we are looking for documents with aspect codes 22, 345 and 42, the steps above:

{{{
 1. select code, count(*) from Classifications where code in ( '22', '345', '42' ) order by 2;

 +------+----------+
 | code | count(*) |
 +------+----------+
 |  345 |       10 |
 |   42 |      234 |
 |   22 |     1410 |
 +------+----------+

 2. select objectId from Classifications where objectTypeId = ... and code = '22' and object Id in (
      select objectId from Classifications where objectTypeId = ... and code = '42' and object Id in (
	select objectId from Classifications where objectTypeId = ... and code = '345'
      )
    )
}}}

=== List ===

Displaying of classifications in a document list is also a performance issue. To be able to display classification information quickly a static database table will be used to store some information about classified documents:

'''AspectCache Table'''

|| Object Id || Object Type ||<-7> aspect 0 ||...||<-7> aspect 9 ||
|| || || min code || max code  || full code || hoverText || count || from || to ||...|| min code || max code  || full code || hoverText || count || from || to ||

There is one record for one object (Drawing, Task). The field ''full code'' is a comma separated list of all aspect values in the appropriate aspect assigned to the object. On a document list with classifications the ''min code'', ''max code'' can appear according the sorting order. On a non-sorted list the ''full code'' can appear or it can be a site setting parameter.

'''Issue'''

The AspectCache table can be updated:

 1. database procedures on regularly basis (cron?)
 1. fired by triggers on the tables Drawings, Comments and Classifications 
 1. by application transactions


'''Listing Multiple Classification Values'''

Three cases are possible:

 1. One classification, with a single from-to pair (E.g. one chainage range on a single alignment)
    * Display: '''Code x->y''' -- or code only, with no values?
    * Hover:   '''Name x->y'''

 1. Multiple classifications on same classification, with different from-to pairs (e.g. multiple chainage ranges on a single alignment)
    * Display: '''Code min-x -> max-y...'''  
    * Hover:   '''Name min-x -> max-y (multiple)'''  

 1. Multiple classifications on different classification, with different from-to pairs (e.g. multiple chainage ranges on multiple alignments)
    * Display: '''Code-mix, Code-max...''' -- Or can we display something more sensible?
    * Hover:   '''Name-min, Nam-max...(multiple)'''

We don't need to make a decision now, as the display only changes the update that fills the aspect cache table.