123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990 |
- #
- # Metrix++, Copyright 2009-2013, Metrix++ Project
- # Link: http://metrixplusplus.sourceforge.net
- #
- # This file is a part of Metrix++ Tool.
- #
- # Metrix++ is free software: you can redistribute it and/or modify
- # it under the terms of the GNU General Public License as published by
- # the Free Software Foundation, version 3 of the License.
- #
- # Metrix++ 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 General Public License for more details.
- #
- # You should have received a copy of the GNU General Public License
- # along with Metrix++. If not, see <http://www.gnu.org/licenses/>.
- #
- import core.api
- import core.cout
- import re
- class Plugin(core.api.Plugin, core.api.Child, core.api.IConfigurable):
-
- def declare_configuration(self, parser):
- parser.add_option("--std.suppress", "--ss", action="store_true", default=False,
- help="If set (True), suppression markers are collected from comments in code. "
- "Suppressions are used by post-processing tools, like limit, to remove false-positive warnings. "
- "Suppressions should be in the first comment block of a region (function/class/interface). "
- "Format of suppressions: 'metrix++: suppress metric-name'. "
- "For example: 'metrix++: suppress std.code.complexity:cyclomatic'. "
- " [default: %default]")
-
- def configure(self, options):
- self.is_active = options.__dict__['std.suppress']
-
- def initialize(self):
- fields = []
- if self.is_active == True:
- fields.append(self.Field('count', int, non_zero=True))
- fields.append(self.Field('list', str))
- core.api.Plugin.initialize(self, fields=fields)
-
- if len(fields) != 0:
- core.api.subscribe_by_parents_interface(core.api.ICode, self, 'callback')
- # suppress pattern
- pattern = re.compile(r'''metrix[+][+][:][ \t]+suppress[ \t]+([^ \t\r\n\*]+)''')
- def callback(self, parent, data, is_updated):
- is_updated = is_updated or self.is_updated
- if is_updated == True:
- text = data.get_content()
- for region in data.iterate_regions():
- count = 0
- list_text = ""
- last_comment_end = None
- for marker in data.iterate_markers(
- filter_group = data.get_marker_types().COMMENT,
- region_id = region.get_id(),
- exclude_children = True):
-
- if last_comment_end != None and marker.get_offset_begin() > last_comment_end + 2:
- # check continues comment blocks
- # stop searching, if this comment block is far from the last
- break
- last_comment_end = marker.get_offset_end()
-
- matches = self.pattern.findall(text, marker.get_offset_begin(), marker.get_offset_end())
- for m in matches:
- namespace_name, field = m.split(':')
- namespace = self.get_plugin_loader().get_database_loader().get_namespace(namespace_name)
- if namespace == None or namespace.get_field_packager(field) == None:
- core.cout.notify(data.get_path(), region.get_cursor(),
- core.cout.SEVERITY_WARNING,
- "Suppressed metric '" + namespace_name + ":" + field +
- "' is not being collected",
- [("Metric name", namespace_name + ":" + field),
- ("Region name", region.get_name())])
- continue
- count += 1
- list_text += "[" + m +"]"
- if count > 0:
- region.set_data(self.get_name(), 'count', count)
- region.set_data(self.get_name(), 'list', list_text)
|