On Friday, May 6, 2016 at 3:55:03 PM UTC+2, Cédric Krier wrote:
>
> On 2016-05-06 05:41, edbo....@gmail.com <javascript:> wrote:
> > Hello,
> >
> > I think I'm using the Model. __post_setup__ class not correctly, but how
> > I'm using it is working in Tryton-3.6.5 but not in Tryton-4.0.0 and
> maybe
> > there are other ways to do it.
> >
> > What I'm trying to do, is to build an IOT-platform into Tryton where
> agents
> > can connect to and update several values they allowed to change. To
> > minimize database-queries and updating database-records or inserting new
> > ones, I build a dict which lives in memory. The agents send the new
> values
> > in and Tryton will update the values in the dict, and a cron-job will
> > archive those values into the database at certain moments. The new
> values
> > are also used to trigger events like sending mail, executing scripts
> etc.
> >
> > To create the dict in memory I'm using the __post_setup__ class to
> populate
> > the dict with the several values.
> >
> > @classmethod
> > def __post_setup__(cls):
> > super(Trigger, cls).__post_setup__()
> >
> > # get the triggers
> > triggers = cls.search( [('id', '>', 0 )] )
> >
> > if triggers:
> > # Build the datastructure in memory
> > for trigger in cls.browse( triggers ):
> >
> > sensors = []
> > on_activate = []
> > on_deactivate = []
> >
> > # get the sensors and conditions
> > for sensor in trigger.sensor:
> > if len(sensor.condition) > 2:
> > sensors.append({
> > "id" : sensor.id,
> > "sensor" : sensor.sensor.id,
> > "condition" : sensor.condition,
> > })
> >
> > # only add the trigger to the list when there are
> > # sensors and conditions
> > if len(sensors):
> > MEM_TRIGGERS.append({
> > "id" : trigger.id,
> > "name" : trigger.name,
> > "sequence" : trigger.sequence,
> > "min_delay" : trigger.minimum_time_delay,
> > "min_active" : trigger.minimum_time_active,
> > "when" : datetime.datetime.utcnow(),
> > "activated" : trigger.activated,
> > "enabled" : trigger.enabled,
> > "sensors" : sensors
> > })
> >
> > I have extended the ir.trigger with
> > sensor = fields.One2Many('iot.triggersensors', 'trigger', 'Sensor')
> >
> > And the iot.triggersensors
> >
> > class TriggerSensor(ModelSQL, ModelView):
> > "TriggerSensor"
> > __name__ = 'iot.triggersensors'
> >
> > name = fields.Char('Name')
> > trigger = fields.Many2One('ir.trigger', 'Trigger',
> > domain=[('on_sensor','=',True)], ondelete='CASCADE')
> > condition = fields.Char('Condition')
> >
> >
> > This is working in Tryton version 3.65, but in 4.0.0 I get
> >
> > File "/home/ed/ERP/iot/trytond-4.0.0/trytond/pool.py", line 229, in
> setup
> > cls.__post_setup__()
> > File "/home/ed/ERP/iot/trytond-4.0.0/trytond/modules/iot/iot.py", line
> > 1173, in __post_setup__
> > for sensor in trigger.sensor:
> > File "/home/ed/ERP/iot/trytond-4.0.0/trytond/model/modelstorage.py",
> line
> > 1352, in __getattr__
> > read_data = self.read(list(ids), ffields.keys())
> > File "/home/ed/ERP/iot/trytond-4.0.0/trytond/model/modelsql.py", line
> > 744, in read
> > getter_result = field.get(ids, cls, fname, values=result)
> > File
> "/home/ed/ERP/iot/trytond-4.0.0/trytond/model/fields/one2many.py",
> > line 86, in get
> > field = Relation._fields[self.field]
> > AttributeError: type object 'iot.triggersensors' has no attribute
> '_fields'
> >
> > So my question is, are there other methods to accomplish this (after the
> > Tryton-server is started create a dict into memory)? or is this a bug in
> > Tryton-4.0.0?
>
> It is definitely not a bug in 4.0. Your code was working before by
> chance. The ORM should never be used during the initialization of the
> Pool. Your code doesn't work any more probably because of the
> optimization of the pool initialization.
>
> Hmm, ok, but the documentation says "Setup the class AFTER added into the
trytond.pool.Pool". So IMO this means that the class is added into the pool
and then __post_setup__ will be called.
> Also you should be careful with global variable, your structure is
> probably not thread-safe.
>
I don't use global. In my iot.py I have
MEM_SENSORS = {} # store sensor values in memory
just below values like
__metaclass__ = PoolMeta
logger = logging.getLogger(__name__)
STATES = {
'readonly': ~Eval('active', True),
}
DEPENDS = ['active']
So I think this is not global and I have written an alternative
write-function for the agents, so the agents are just like clients but very
restricted in their possibilities. But you are right about not being
thread-safe. That's why I'm thinking to put Tryton-celery in between. But I
have to dig deeper into that and maybe I can solve this issue with that.
> Finally, are you sure you need this? Of course database query is
> expensive but included in all the stack of a request, I don't think it
> has so much effect.
>
I'm running the system on a BeagleBone Black and it's running surprisingly
well! even better then I have thought of. I have one agent which also runs
on the BB and collects data from 8 sensors at several intervals (minimal
interval 2 sec, maximum interval 5 minutes). The memory consumption of
postgresql, tryton-server and agent is around 30MB and the CPU is running
around 30% load. Connecting with the client to get the archived values will
take some seconds, but that's not a problem.
I'm switched to the memory-part because the BB couldn't keep up with
updating the records in the database and starts to swap. I also tried to
use sqlite in memory, but the same happened. So I need this kind of
behavior and it's just working well.
> --
> Cédric Krier - B2CK SPRL
> Email/Jabber: cedric...@b2ck.com <javascript:>
> Tel: +32 472 54 46 59
> Website: http://www.b2ck.com/
>
--
You received this message because you are subscribed to the Google Groups
"tryton-dev" group.
To view this discussion on the web visit
https://groups.google.com/d/msgid/tryton-dev/e1239559-ef17-4009-9a23-905debb7e6ed%40googlegroups.com.
