# Copyright 2017 The Bazel Authors. All rights reserved. # # 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. """Unit tests for dicts.bzl.""" load("//lib:dicts.bzl", "dicts") load("//lib:unittest.bzl", "asserts", "unittest") def _add_test(ctx): """Unit tests for dicts.add.""" env = unittest.begin(ctx) # Test zero- and one-argument behavior. asserts.equals(env, {}, dicts.add()) asserts.equals(env, {"a": 1}, dicts.add({"a": 1})) asserts.equals(env, {"a": 1}, dicts.add(a = 1)) asserts.equals(env, {"a": 1, "b": 2}, dicts.add({"a": 1}, b = 2)) # Test simple two-argument behavior. asserts.equals(env, {"a": 1, "b": 2}, dicts.add({"a": 1}, {"b": 2})) asserts.equals(env, {"a": 1, "b": 2, "c": 3}, dicts.add({"a": 1}, {"b": 2}, c = 3)) # Test simple more-than-two-argument behavior. asserts.equals( env, {"a": 1, "b": 2, "c": 3, "d": 4}, dicts.add({"a": 1}, {"b": 2}, {"c": 3}, {"d": 4}), ) asserts.equals( env, {"a": 1, "b": 2, "c": 3, "d": 4, "e": 5}, dicts.add({"a": 1}, {"b": 2}, {"c": 3}, {"d": 4}, e = 5), ) # Test same-key overriding. asserts.equals(env, {"a": 100}, dicts.add({"a": 1}, {"a": 100})) asserts.equals(env, {"a": 100}, dicts.add({"a": 1}, a = 100)) asserts.equals(env, {"a": 10}, dicts.add({"a": 1}, {"a": 100}, {"a": 10})) asserts.equals(env, {"a": 10}, dicts.add({"a": 1}, {"a": 100}, a = 10)) asserts.equals( env, {"a": 100, "b": 10}, dicts.add({"a": 1}, {"a": 100}, {"b": 10}), ) asserts.equals(env, {"a": 10}, dicts.add({"a": 1}, {}, {"a": 10})) asserts.equals(env, {"a": 10}, dicts.add({"a": 1}, {}, a = 10)) asserts.equals( env, {"a": 10, "b": 5}, dicts.add({"a": 1}, {"a": 10, "b": 5}), ) asserts.equals( env, {"a": 10, "b": 5}, dicts.add({"a": 1}, a = 10, b = 5), ) # Test some other boundary cases. asserts.equals(env, {"a": 1}, dicts.add({"a": 1}, {})) # Since dictionaries are passed around by reference, make sure that the # result of dicts.add is always a *copy* by modifying it afterwards and # ensuring that the original argument doesn't also reflect the change. We do # this to protect against someone who might attempt to optimize the function # by returning the argument itself in the one-argument case. original = {"a": 1} result = dicts.add(original) result["a"] = 2 asserts.equals(env, 1, original["a"]) return unittest.end(env) add_test = unittest.make(_add_test) def _omit_test(ctx): """Unit tests for dicts.omit.""" env = unittest.begin(ctx) # Test empty dict, empty list. asserts.equals(env, {}, dicts.omit({}, [])) # Test empty dict, nonempty list. asserts.equals(env, {}, dicts.omit({}, ["a"])) # Test nonempty dict, empty list. asserts.equals(env, {"a": 1}, dicts.omit({"a": 1}, [])) # Test key in dict. asserts.equals(env, {}, dicts.omit({"a": 1}, ["a"])) # Test key not in dict. asserts.equals(env, {"a": 1}, dicts.omit({"a": 1}, ["b"])) # Since dictionaries are passed around by reference, make sure that the # result of dicts.omit is always a *copy* by modifying it afterwards and # ensuring that the original argument doesn't also reflect the change. We do # this to protect against someone who might attempt to optimize the function # by returning the argument itself in the empty list case. original = {"a": 1} result = dicts.omit(original, []) result["a"] = 2 asserts.equals(env, 1, original["a"]) return unittest.end(env) omit_test = unittest.make(_omit_test) def _pick_test(ctx): """Unit tests for dicts.pick.""" env = unittest.begin(ctx) # Test empty dict, empty list. asserts.equals(env, {}, dicts.pick({}, [])) # Test empty dict, nonempty list. asserts.equals(env, {}, dicts.pick({}, ["a"])) # Test nonempty dict, empty list. asserts.equals(env, {}, dicts.pick({"a": 1}, [])) # Test key in dict. asserts.equals(env, {"a": 1}, dicts.pick({"a": 1}, ["a"])) # Test key not in dict. asserts.equals(env, {}, dicts.pick({"a": 1}, ["b"])) # Since dictionaries are passed around by reference, make sure that the # result of dicts.pick is always a *copy* by modifying it afterwards and # ensuring that the original argument doesn't also reflect the change. We do # this to protect against someone who might attempt to optimize the function # by returning the argument itself. original = {"a": 1} result = dicts.pick(original, ["a"]) result["a"] = 2 asserts.equals(env, 1, original["a"]) return unittest.end(env) pick_test = unittest.make(_pick_test) def dicts_test_suite(): """Creates the test targets and test suite for dicts.bzl tests.""" unittest.suite( "dicts_tests", add_test, omit_test, pick_test, )