Hello every one.
Sorry : I’m not friend in English and not in Ruby 
I failed to build a complex Arel query.
I have 3 tables : One is a “Set” table, One is a “Element” table, and one “elements_sets” is the join table for the relation. I need only these join table for my Arel query :
mysql> describe elements_sets;
±----------------±-----±-----±----±--------±------+
| Field | Type | Null | Key | Default | Extra |
±----------------±-----±-----±----±--------±------+
| element_id | int | NO | PRI | NULL | |
| set_id | int | NO | PRI | NULL | |
±----------------±-----±-----±----±--------±------+
For a given set , les say my_set_id = 33
I want to find the 10 most closer other sets ids in this tables.
I have 3 definitions of my metric :
proximity : size of the intersection of (my_set and a_set) / size of the union of (my_set and a_set)
inclusion of my_set : size of the intersection / size of my_set
inclusion of a_set : size of the intersection / size of a_set
I was able to construct the arel query for the second one, because the denominator is a constant,
and I can order the result based on the size of the intersection.
Here is my query :
es1 = Arel::Table.new :elements_sets
es2 = Arel::Table.new :elements_sets
my_set_id = 33 # for example.
query = es1
.project( :set_id, es1[:element_id].count )
.where(
es1[ :set_id ].not_eq( my_set_id )
.and( es1[:element_id].in(
es2.project(:element_id).where(es2[:set_id].eq( my_set_id ) )))
)
.group( es1[:set_id] )
.having(es1[:element_id].count.gt(10) )
.order(es1[:element_id].count.desc)
.take(10)
Maybe this query, for the second metric, is not optimal, but it’s work.
For the first and last metric, I have no idea how to build the query, because the denominator depend of the size of each “a_set”.
I need to write an expression in the order clause, like
.order( “size of the intersection” / “size of the union” )
Thanks for help me … 