96e6d7e403
* Clear separation for problems that can be handled so that the program can continue and problems that can't handled automatically due to major issues in configuration, environment or code itself * Split YAQL exceptions into two types: grammar exception and evaluation exception * General NotFoundException is replaced with more specific DBEntryNotFoundException for better consistency with other DB exceptions and more clear semantics * Fixed corresponding tests Change-Id: I07f495ab316b0f164caece78b1f101219199e68c Implements: blueprint mistral-engine-error-handling
330 lines
9.9 KiB
Python
330 lines
9.9 KiB
Python
# Copyright 2013 - Mirantis, Inc.
|
|
# Copyright 2015 - StackStorm, Inc.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
from mistral import exceptions as exc
|
|
from mistral import expressions as expr
|
|
from mistral.tests.unit import base
|
|
|
|
DATA = {
|
|
"server": {
|
|
"id": "03ea824a-aa24-4105-9131-66c48ae54acf",
|
|
"name": "cloud-fedora",
|
|
"status": "ACTIVE"
|
|
},
|
|
"status": "OK"
|
|
}
|
|
|
|
SERVERS = {
|
|
"servers": [
|
|
{'name': 'centos'},
|
|
{'name': 'ubuntu'},
|
|
{'name': 'fedora'}
|
|
]
|
|
}
|
|
|
|
|
|
class YaqlEvaluatorTest(base.BaseTest):
|
|
def setUp(self):
|
|
super(YaqlEvaluatorTest, self).setUp()
|
|
|
|
self._evaluator = expr.YAQLEvaluator()
|
|
|
|
def test_expression_result(self):
|
|
res = self._evaluator.evaluate('$.server', DATA)
|
|
self.assertEqual({
|
|
'id': "03ea824a-aa24-4105-9131-66c48ae54acf",
|
|
'name': 'cloud-fedora',
|
|
'status': 'ACTIVE'
|
|
}, res)
|
|
|
|
res = self._evaluator.evaluate('$.server.id', DATA)
|
|
self.assertEqual('03ea824a-aa24-4105-9131-66c48ae54acf', res)
|
|
|
|
res = self._evaluator.evaluate("$.server.status = 'ACTIVE'", DATA)
|
|
self.assertTrue(res)
|
|
|
|
def test_wrong_expression(self):
|
|
res = self._evaluator.evaluate("$.status = 'Invalid value'", DATA)
|
|
self.assertFalse(res)
|
|
|
|
self.assertRaises(
|
|
exc.YaqlEvaluationException,
|
|
self._evaluator.evaluate,
|
|
'$.wrong_key',
|
|
DATA
|
|
)
|
|
|
|
expression_str = 'invalid_expression_string'
|
|
res = self._evaluator.evaluate(expression_str, DATA)
|
|
self.assertEqual(expression_str, res)
|
|
|
|
def test_select_result(self):
|
|
res = self._evaluator.evaluate(
|
|
'$.servers.where($.name = ubuntu)',
|
|
SERVERS
|
|
)
|
|
item = list(res)[0]
|
|
self.assertEqual({'name': 'ubuntu'}, item)
|
|
|
|
def test_function_string(self):
|
|
self.assertEqual('3', self._evaluator.evaluate('str($)', '3'))
|
|
self.assertEqual('3', self._evaluator.evaluate('str($)', 3))
|
|
|
|
def test_function_len(self):
|
|
self.assertEqual(3, self._evaluator.evaluate('len($)', 'hey'))
|
|
data = [{'some': 'thing'}]
|
|
|
|
self.assertEqual(
|
|
1,
|
|
self._evaluator.evaluate('$.where($.some = thing).len()', data)
|
|
)
|
|
|
|
def test_validate(self):
|
|
self._evaluator.validate('abc')
|
|
self._evaluator.validate('1')
|
|
self._evaluator.validate('1 + 2')
|
|
self._evaluator.validate('$.a1')
|
|
self._evaluator.validate('$.a1 * $.a2')
|
|
|
|
def test_validate_failed(self):
|
|
self.assertRaises(exc.YaqlGrammarException,
|
|
self._evaluator.validate,
|
|
'*')
|
|
|
|
self.assertRaises(exc.YaqlGrammarException,
|
|
self._evaluator.validate,
|
|
[1, 2, 3])
|
|
|
|
self.assertRaises(exc.YaqlGrammarException,
|
|
self._evaluator.validate,
|
|
{'a': 1})
|
|
|
|
def test_json_pp(self):
|
|
self.assertEqual('"3"', self._evaluator.evaluate('json_pp($)', '3'))
|
|
self.assertEqual('3', self._evaluator.evaluate('json_pp($)', 3))
|
|
self.assertEqual(
|
|
'[\n 1,\n 2\n]',
|
|
self._evaluator.evaluate('json_pp($)', [1, 2])
|
|
)
|
|
self.assertEqual(
|
|
'{\n "a": "b"\n}',
|
|
self._evaluator.evaluate('json_pp($)', {'a': 'b'})
|
|
)
|
|
self.assertEqual(
|
|
'"Mistral\nis\nawesome"',
|
|
self._evaluator.evaluate(
|
|
'json_pp($)', '\n'.join(['Mistral', 'is', 'awesome'])
|
|
)
|
|
)
|
|
|
|
|
|
class InlineYAQLEvaluatorTest(base.BaseTest):
|
|
def setUp(self):
|
|
super(InlineYAQLEvaluatorTest, self).setUp()
|
|
|
|
self._evaluator = expr.InlineYAQLEvaluator()
|
|
|
|
def test_multiple_placeholders(self):
|
|
expr_str = """
|
|
Statistics for tenant "<% $.project_id %>"
|
|
|
|
Number of virtual machines: <% $.vm_count %>
|
|
Number of active virtual machines: <% $.active_vm_count %>
|
|
Number of networks: <% $.net_count %>
|
|
|
|
-- Sincerely, Mistral Team.
|
|
"""
|
|
|
|
result = self._evaluator.evaluate(
|
|
expr_str,
|
|
{
|
|
'project_id': '1-2-3-4',
|
|
'vm_count': 28,
|
|
'active_vm_count': 0,
|
|
'net_count': 1
|
|
}
|
|
)
|
|
|
|
expected_result = """
|
|
Statistics for tenant "1-2-3-4"
|
|
|
|
Number of virtual machines: 28
|
|
Number of active virtual machines: 0
|
|
Number of networks: 1
|
|
|
|
-- Sincerely, Mistral Team.
|
|
"""
|
|
|
|
self.assertEqual(expected_result, result)
|
|
|
|
def test_function_string(self):
|
|
self.assertEqual('3', self._evaluator.evaluate('<% str($) %>', '3'))
|
|
self.assertEqual('3', self._evaluator.evaluate('<% str($) %>', 3))
|
|
|
|
def test_validate(self):
|
|
self._evaluator.validate('There is no expression.')
|
|
self._evaluator.validate('<% abc %>')
|
|
self._evaluator.validate('<% 1 %>')
|
|
self._evaluator.validate('<% 1 + 2 %>')
|
|
self._evaluator.validate('<% $.a1 %>')
|
|
self._evaluator.validate('<% $.a1 * $.a2 %>')
|
|
self._evaluator.validate('<% $.a1 %> is <% $.a2 %>')
|
|
self._evaluator.validate('The value is <% $.a1 %>.')
|
|
|
|
def test_validate_failed(self):
|
|
self.assertRaises(exc.YaqlGrammarException,
|
|
self._evaluator.validate,
|
|
'The value is <% * %>.')
|
|
|
|
self.assertRaises(exc.YaqlEvaluationException,
|
|
self._evaluator.validate,
|
|
[1, 2, 3])
|
|
|
|
self.assertRaises(exc.YaqlEvaluationException,
|
|
self._evaluator.validate,
|
|
{'a': 1})
|
|
|
|
|
|
class ExpressionsTest(base.BaseTest):
|
|
def test_evaluate_complex_expressions(self):
|
|
data = {
|
|
'a': 1,
|
|
'b': 2,
|
|
'c': 3,
|
|
'd': True,
|
|
'e': False,
|
|
'f': 10.1,
|
|
'g': 10,
|
|
'h': [1, 2, 3, 4, 5],
|
|
'i': 'We are OpenStack!',
|
|
'j': 'World',
|
|
'k': 'Mistral',
|
|
'l': 'awesome',
|
|
'm': 'the way we roll'
|
|
}
|
|
|
|
test_cases = [
|
|
('<% $.a + $.b * $.c %>', 7),
|
|
('<%($.a + $.b) * $.c %>', 9),
|
|
('<% $.d and $.e %>', False),
|
|
('<% $.f > $.g %>', True),
|
|
('<% $.h.len() >= 5 %>', True),
|
|
('<% $.h.len() >= $.b + $.c %>', True),
|
|
('<% 100 in $.h %>', False),
|
|
('<% $.a in $.h%>', True),
|
|
('<% ''OpenStack'' in $.i %>', True),
|
|
('Hello, <% $.j %>!', 'Hello, World!'),
|
|
('<% $.k %> is <% $.l %>!', 'Mistral is awesome!'),
|
|
('This is <% $.m %>.', 'This is the way we roll.'),
|
|
('<% 1 + 1 = 3 %>', False)
|
|
]
|
|
|
|
for expression, expected in test_cases:
|
|
actual = expr.evaluate_recursively(expression, data)
|
|
|
|
self.assertEqual(expected, actual)
|
|
|
|
def test_evaluate_recursively(self):
|
|
task_spec_dict = {
|
|
'parameters': {
|
|
'p1': 'My string',
|
|
'p2': '<% $.param2 %>',
|
|
'p3': ''
|
|
},
|
|
'publish': {
|
|
'new_key11': 'new_key1'
|
|
}
|
|
}
|
|
|
|
modified_task = expr.evaluate_recursively(
|
|
task_spec_dict,
|
|
{'param2': 'val32'}
|
|
)
|
|
|
|
self.assertDictEqual(
|
|
{
|
|
'parameters': {
|
|
'p1': 'My string',
|
|
'p2': 'val32',
|
|
'p3': ''
|
|
},
|
|
'publish': {
|
|
'new_key11': 'new_key1'
|
|
}
|
|
},
|
|
modified_task
|
|
)
|
|
|
|
def test_evaluate_recursively_arbitrary_dict(self):
|
|
context = {
|
|
"auth_token": "123",
|
|
"project_id": "mistral"
|
|
}
|
|
|
|
data = {
|
|
"parameters": {
|
|
"parameter1": {
|
|
"name1": "<% $.auth_token %>",
|
|
"name2": "val_name2"
|
|
},
|
|
"param2": [
|
|
"var1",
|
|
"var2",
|
|
"/servers/<% $.project_id %>/bla"
|
|
]
|
|
},
|
|
"token": "<% $.auth_token %>"
|
|
}
|
|
|
|
applied = expr.evaluate_recursively(data, context)
|
|
|
|
self.assertDictEqual(
|
|
{
|
|
"parameters": {
|
|
"parameter1": {
|
|
"name1": "123",
|
|
"name2": "val_name2"
|
|
},
|
|
"param2": ["var1", "var2", "/servers/mistral/bla"]
|
|
},
|
|
"token": "123"
|
|
},
|
|
applied
|
|
)
|
|
|
|
def test_evaluate_recursively_environment(self):
|
|
environment = {
|
|
'host': 'vm1234.example.com',
|
|
'db': 'test',
|
|
'timeout': 600,
|
|
'verbose': True,
|
|
'__actions': {
|
|
'std.sql': {
|
|
'conn': 'mysql://admin:secrete@<% env().host %>'
|
|
'/<% env().db %>'
|
|
}
|
|
}
|
|
}
|
|
|
|
context = {
|
|
'__env': environment
|
|
}
|
|
|
|
defaults = context['__env']['__actions']['std.sql']
|
|
applied = expr.evaluate_recursively(defaults, context)
|
|
expected = 'mysql://admin:secrete@vm1234.example.com/test'
|
|
|
|
self.assertEqual(expected, applied['conn'])
|