Get length of a many2many or one2many ids

Wow thank you ! that was fast, this is very clear for me now.

So if I use list(myobject.item_ids), does that mean that only one SELECT for all IDs is made or it only means that all SELECT are made in one pass ? (note that when I write one SELECT, I know that there are a lot of more SELECT behind this for translations, access rules, etc. )

About the read method, maybe a workaround could be added to the browse method with two parameters:
- a field restriction array (also used for 2many fields)
- a depth level (to restrict 2many fields level )

for eg:
   sale = sale_order.browse([1], ['name', 'partner_id', 'order_line_ids','product_id'], 2][0]
could return an object where only name, partner_id and order_line_ids could be accessed by dot notation.
product_id would be also be available via dot notation on order_line_ids
product_id.name too
but not sale_order.order_line_ids.product_id.seller_ids because of the depth level (and also because it does not exists in the field restriction list)

Hum, no. If you have 100 records/IDs in your "myobject.item_ids" relation field, then behind the scene a "list(myobject.item_ids)" statement will generate 100 calls to the 'read()' RPC method, one by record (and then construct a list from these results). That is not optimal :)

There is no way to know if a "list()" is used upon "myobject.item_ids" or not. The iteration of a "_ids" relation has been made this way to begin to iterate a large collection of records without having to wait that all records are fetched in local, we just request them one by one (it is often more confortable).

But the call to the 'read()' method below generates only one RPC request:

>>> object_obj.read([myobject_id])[0]['item_ids']
[5, 6, 74, 12, 6]
>>> len(object_obj.read([myobject_id])[0]['item_ids'])
5

Or, if you prefer, from your 'myobject' record:

>>> myobject.__data__['raw_data']['item_ids'] # __data__['raw_data'] contains all data requested during the instantiation of 'myobject'
[5, 6, 74, 12, 6]
>>> len(myobject.__data__['raw_data']['item_ids'])
5

About the parameters added to 'browse()', unfortunatly this will not solve the number of RPC requests made. OERPLib already avoid to auto-fetch sub-relations of a record, it only fetch them when the user accesses the relation field (the list of IDs behind your 'item_ids' field is fetch from OpenERP when you access this field on the first time):

>>> object_obj = oerp.get('my.object')
>>> myobject = object_obj.browse(1) # Only 'char', 'float' and other normal fields are fetched here. No relational field values
>>> list(myobject.item_ids) # IDs contains in 'item_ids' are requested to OpenERP. This allow an unprivileged user to browse 'myobject', and receive an exception if it access a relation such as 'item_ids' on which he has no access right

But yes, there are surely some improvements to do about browsable records in OERPLib, maybe an option in OERP.config to change the default behaviour:

>>> object_obj = oerp.get('my.object')
>>> object_obj.browse(IDS) # Like now, returns a generator to iterate records one by one => [read(ID) for ID in IDS]
>>> oerp.config['browse_all_at_once'] = True
>>> object_obj.browse(IDS) # Do only one call to the 'read()' method => read(IDS)

(about a new parameter on "browse()", I just avoid to differ too much from the standard method signatures of OpenERP, but I may be wrong, don't know :)