You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 

215 lines
9.9 KiB

# -*- coding: utf-8 -*-
##############################################################################
#
# Daniel Reis
# 2011
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License as
# published by the Free Software Foundation, either version 3 of the
# License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU Affero General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#
##############################################################################
import os
import sys
from datetime import datetime
from osv import fields, osv
import logging
_logger = logging.getLogger(__name__)
_loglvl = _logger.getEffectiveLevel()
class import_odbc_dbtable(osv.osv):
_name="import.odbc.dbtable"
_description = 'Import Table Data'
_order = 'exec_order'
_columns = {
'name': fields.char('Datasource name', required=True, size=64),
'enabled': fields.boolean('Execution enabled'),
'dbsource_id': fields.many2one('base.external.dbsource', 'Database source', required=True),
'sql_source': fields.text('SQL', required=True, help='Column names must be valid "import_data" columns.'),
'model_target': fields.many2one('ir.model','Target object'),
'noupdate': fields.boolean('No updates', help="Only create new records; disable updates to existing records."),
'exec_order': fields.integer('Execution order', help="Defines the order to perform the import"),
'last_sync': fields.datetime('Last sync date', help="Datetime for the last succesfull sync. Later changes on the source may not be replicated on the destination"),
'start_run': fields.datetime('Time started', readonly=True),
'last_run': fields.datetime('Time ended', readonly=True),
'last_record_count': fields.integer('Last record count', readonly=True),
'last_error_count': fields.integer('Last error count', readonly=True),
'last_warn_count': fields.integer('Last warning count', readonly=True),
'last_log': fields.text('Last run log', readonly=True),
'ignore_rel_errors': fields.boolean('Ignore relationship errors',
help = "On error try to reimport rows ignoring relationships."),
'raise_import_errors': fields.boolean('Raise import errors',
help = "Import errors not handled, intended for debugging purposes."
+ "\nAlso forces debug messages to be written to the serevr log."),
}
_defaults = {
'enabled': True,
'exec_order': 10,
}
def _import_data(self, cr, uid, flds, data, model_obj, table_obj, log):
"""Import data and returns error msg or empty string"""
def find_m2o(field_list):
""""Find index of first column with a one2many field"""
for i, x in enumerate(field_list):
if len(x)>3 and x[-3:] == ':id' or x[-3:] == '/id':
return i
return -1
def append_to_log(log, level, obj_id = '', msg = '', rel_id = ''):
if '_id_' in obj_id:
obj_id = '.'.join(obj_id.split('_')[:-2]) + ': ' + obj_id.split('_')[-1]
if ': .' in msg and not rel_id:
rel_id = msg[msg.find(': .')+3:]
if '_id_' in rel_id:
rel_id = '.'.join(rel_id.split('_')[:-2]) + ': ' + rel_id.split('_')[-1]
msg = msg[:msg.find(': .')]
log['last_log'].append('%s|%s\t|%s\t|%s' % (level.ljust(5), obj_id, rel_id, msg))
_logger.debug( data )
cols = list(flds) #copy to avoid side effects
err = list()
msg = str()
if table_obj.raise_import_errors:
model_obj.import_data(cr, uid, cols, [data], noupdate=table_obj.noupdate)
else:
try:
model_obj.import_data(cr, uid, cols, [data], noupdate=table_obj.noupdate)
except:
msg = str(sys.exc_info()[1])
if err and not table_obj.ignore_rel_errors:
#Fail
append_to_log(log, 'ERROR', data, msg )
log['last_error_count'] += 1
return False
if err and table_obj.ignore_rel_errors:
#Warn and retry ignoring many2one fields...
append_to_log(log, 'WARN', data, msg )
log['last_warn_count'] += 1
#Try ignoring each many2one (tip: in the SQL sentence select more problematic FKs first)
i = find_m2o(cols)
if i >= 0:
#Try again without the [i] column
del cols[i]
del data[i]
self._import_data(cr, uid, cols, data, model_obj, table_obj, log)
else:
#Fail
append_to_log(log, 'ERROR', data, 'Removed all m2o keys and still fails.' )
log['last_error_count'] += 1
return False
return True
def import_run(self, cr, uid, ids=None, context=None):
db_model = self.pool.get('base.external.dbsource')
actions = self.read(cr, uid, ids, ['id', 'exec_order'])
actions.sort(key = lambda x:(x['exec_order'], x['id']))
#Consider each dbtable:
for action_ref in actions:
obj = self.browse(cr, uid, action_ref['id'])
if not obj.enabled: continue #skip
_logger.setLevel(obj.raise_import_errors and logging.DEBUG or _loglvl)
_logger.debug('Importing %s...' % obj.name)
#now() microseconds are stripped to avoid problem with SQL smalldate
#TODO: convert UTC Now to local timezone (http://stackoverflow.com/questions/4770297/python-convert-utc-datetime-string-to-local-datetime)
model_name = obj.model_target.model
model_obj = self.pool.get(model_name)
xml_prefix = model_name.replace('.', '_') + "_id_"
log = {'start_run': datetime.now().replace(microsecond=0),
'last_run': None,
'last_record_count': 0,
'last_error_count': 0,
'last_warn_count': 0,
'last_log': list()}
self.write(cr, uid, [obj.id], log)
#Prepare SQL sentence; replace "%s" with the last_sync date
if obj.last_sync: sync = datetime.strptime(obj.last_sync, "%Y-%m-%d %H:%M:%S")
else: sync = datetime.datetime(1900, 1, 1, 0, 0, 0)
params = {'sync': sync}
res = db_model.execute_and_inspect(cr, uid, [obj.dbsource_id.id], obj.sql_source, params)
#Exclude columns titled "None"; add (xml_)"id" column
cidx = [i for i, x in enumerate(res['colnames']) if x.upper() != 'NONE']
cols = [x for i, x in enumerate(res['colnames']) if x.upper() != 'NONE'] + ['id']
#Import each row:
for row in res['rows']:
#Build data row; import only columns present in the "cols" list
data = list()
for i in cidx:
#TODO: Handle imported datetimes properly - convert from localtime to UTC!
v = row[i]
if isinstance(v, str): v = v.strip()
data.append(v)
data.append( xml_prefix + str(row[0]).strip() )
#Import the row; on error, write line to the log
log['last_record_count'] += 1
self._import_data(cr, uid, cols, data, model_obj, obj, log)
if log['last_record_count'] % 500 == 0:
_logger.info('...%s rows processed...' % (log['last_record_count']) )
#Finished importing all rows
#If no errors, write new sync date
if not (log['last_error_count'] or log['last_warn_count']):
log['last_sync'] = log['start_run']
level = logging.DEBUG
if log['last_warn_count']: level = logging.WARN
if log['last_error_count']: level = logging.ERROR
_logger.log(level, 'Imported %s , %d rows, %d errors, %d warnings.' % (
model_name, log['last_record_count'], log['last_error_count'] ,
log['last_warn_count'] ) )
#Write run log, either if the table import is active or inactive
if log['last_log']:
log['last_log'].insert(0, 'LEVEL|== Line == |== Relationship ==|== Message ==')
log.update( {'last_log': '\n'.join(log['last_log'])} )
log.update({ 'last_run': datetime.now().replace(microsecond=0) }) #second=0,
self.write(cr, uid, [obj.id], log)
#Finished
_logger.debug('Import job FINISHED.')
return True
def import_schedule(self, cr, uid, ids, context=None):
cron_obj = self.pool.get('ir.cron')
new_create_id = cron_obj.create(cr, uid, {
'name': 'Import ODBC tables',
'interval_type': 'hours',
'interval_number': 1,
'numbercall': -1,
'model': 'import.odbc.dbtable',
'function': 'import_run',
'doall': False,
'active': True
})
return {
'name': 'Import ODBC tables',
'view_type': 'form',
'view_mode': 'form,tree',
'res_model': 'ir.cron',
'res_id': new_create_id,
'type': 'ir.actions.act_window',
}
import_odbc_dbtable()