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.
 
 
 

186 lines
6.4 KiB

# Copyright (C) 2019-Today: GRAP (<http://www.grap.coop/>)
# @author: Sylvain LE GAL (https://twitter.com/legalsylvain)
# License AGPL-3.0 or later (http://www.gnu.org/licenses/agpl.html).
import os
import pygount
from pathlib import Path
import logging
from odoo import api, fields, models
from odoo.tools.safe_eval import safe_eval
from odoo.modules.module import get_module_path
_logger = logging.getLogger(__name__)
class IrModuleModule(models.Model):
_inherit = 'ir.module.module'
author_ids = fields.Many2many(
string='Authors', comodel_name='ir.module.author', readonly=True,
relation='ir_module_module_author_rel')
module_type_id = fields.Many2one(
string='Module Type', comodel_name='ir.module.type', readonly=True)
python_code_qty = fields.Integer(
string='Python Code Quantity', readonly=True)
xml_code_qty = fields.Integer(
string='XML Code Quantity', readonly=True)
js_code_qty = fields.Integer(
string='JS Code Quantity', readonly=True)
css_code_qty = fields.Integer(
string='CSS Code Quantity', readonly=True)
# Overloadable Section
@api.model
def _get_analyse_settings(self):
"""Return a dictionnary of data analysed
Overload this function if you want to analyse other data
{
'extension': {
'data_to_analyse': 'field_name',
},
}, [...]
extension: extension of the file, with the '.'
data_to_analyse : possible value : code, documentation, empty, string
field_name: Odoo field name to store the analysis
"""
return {
'.py': {
'code': 'python_code_qty',
},
'.xml': {
'code': 'xml_code_qty',
},
'.js': {
'code': 'js_code_qty',
},
'.css': {
'code': 'css_code_qty',
},
}
@api.model
def _get_clean_analyse_values(self):
"""List of fields to unset when a module is uninstalled"""
return {
'author_ids': [6, 0, []],
'module_type_id': False,
'python_code_qty': False,
'xml_code_qty': 0,
'js_code_qty': 0,
'css_code_qty': 0,
}
@api.model
def _get_module_encoding(self, file_ext):
return 'utf-8'
# Overload Section
@api.model
def update_list(self):
res = super().update_list()
if self.env.context.get('analyse_installed_modules', False):
self.search([('state', '=', 'installed')]).button_analyse_code()
return res
@api.multi
def write(self, vals):
res = super().write(vals)
if vals.get('state', False) == 'installed':
self.button_analyse_code()
elif vals.get('state', False) == 'uninstalled'\
and 'module_analysis' not in [x.name for x in self]:
self.write(self._get_clean_analyse_values())
return res
# Public Section
@api.multi
def button_analyse_code(self):
IrModuleAuthor = self.env['ir.module.author']
IrModuleTypeRule = self.env['ir.module.type.rule']
rules = IrModuleTypeRule.search([])
cfg = self.env['ir.config_parameter']
val = cfg.get_param('module_analysis.exclude_directories', '')
exclude_directories = [x.strip() for x in val.split(',') if x.strip()]
val = cfg.get_param('module_analysis.exclude_files', '')
exclude_files = [x.strip() for x in val.split(',') if x.strip()]
for module in self:
_logger.info("Analysing Code for module %s ..." % (module.name))
# Update Authors, based on manifest key
authors = []
if module.author and module.author[0] == '[':
author_txt_list = safe_eval(module.author)
else:
author_txt_list = module.author.split(',')
author_txt_list = [x.strip() for x in author_txt_list]
author_txt_list = [x for x in author_txt_list if x]
for author_txt in author_txt_list:
authors.append(IrModuleAuthor._get_or_create(author_txt))
author_ids = [x.id for x in authors]
module.author_ids = author_ids
# Update Module Type, based on rules
module_type_id = rules._get_module_type_id_from_module(module)
module.module_type_id = module_type_id
# Get Path of module folder and parse the code
module_path = get_module_path(module.name)
# Get Files
analysed_datas = self._get_analyse_data_dict()
file_extensions = analysed_datas.keys()
file_list = self._get_files_to_analyse(
module_path, file_extensions, exclude_directories,
exclude_files)
for file_path, file_ext in file_list:
file_res = pygount.source_analysis(
file_path, '',
encoding=self._get_module_encoding(file_ext))
for k, v in analysed_datas.get(file_ext).items():
v['value'] += getattr(file_res, k)
# Update the module with the datas
values = {}
for file_ext, analyses in analysed_datas.items():
for k, v in analyses.items():
values[v['field']] = v['value']
module.write(values)
# Custom Section
@api.model
def _get_files_to_analyse(
self, path, file_extensions, exclude_directories, exclude_files):
res = []
for root, dirs, files in os.walk(path, followlinks=True):
if set(Path(root).parts) & set(exclude_directories):
continue
for name in files:
if name in exclude_files:
continue
filename, file_extension = os.path.splitext(name)
if file_extension in file_extensions:
res.append((os.path.join(root, name), file_extension))
return res
@api.model
def _get_analyse_data_dict(self):
res_dict = self._get_analyse_settings().copy()
for file_ext, analyse_dict in res_dict.items():
for analyse_type, v in analyse_dict.items():
analyse_dict[analyse_type] = {
'field': v,
'value': 0
}
return res_dict