09ba30ef29
Previously one could set a flag in a `TEST_CASES.json` file to exclude the test-cases from being run if the input files were being compiled partially and then linked. There are also scenarios where one might want to exclude test-cases from "full compile" mode test runs. This commit changes the compliance test tooling to support a new property `compilationModeFilter`, which is an array containing one or more of `"full compile"` and `"linked compile"`. Only the tests whose `compilationModeFilter` array contains the current compilation mode will be run. PR Close #39939 |
||
---|---|---|
.. | ||
full | ||
linked | ||
partial | ||
test_cases | ||
test_helpers | ||
README.md |
README.md
Compliance test-cases
This directory contains rules, helpers and test-cases for the Angular compiler compliance tests.
There are three different types of tests that are run based on file-based "test-cases".
- Full compile - in this test the source files defined by the test-case are fully compiled by Angular. The generated files are compared to "expected files" via a matching algorithm that is tolerant to whitespace and variable name changes.
- Partial compile - in this test the source files defined by the test-case are "partially" compiled by Angular to produce files that can be published. These partially compiled files are compared directly against "golden files" to ensure that we do not inadvertently break the public API of partial declarations.
- Linked - in this test the golden files mentioned in the previous bullet point, are passed to the Angular linker, which generates files that are comparable to the fully compiled files. These linked files are compared against the "expected files" in the same way as in the "full compile" tests.
This way the compliance tests are able to check each mode and stage of compilation is accurate and does not change unexpectedly.
Defining a test-case
To define a test-case, create a new directory below test_cases
. In this directory
- add a new file called
TEST_CASES.json
. The format of this file is described below. - add an empty
GOLDEN_PARTIAL.js
file. This file will be updated by the tooling later. - add any
inputFiles
that will be compiled as part of the test-case. - add any
expected
files that will be compared to the files generated by compiling the source files.
TEST_CASES.json format
The TEST_CASES.json
defines an object with one or more test-case definitions in the cases
property.
Each test-case can specify:
- A
description
of the test. - The
inputFiles
that will be compiled. - Additional
compilerOptions
andangularCompilerOptions
that are passed to the compiler. - Whether to exclude this test-case from certain tests running under certain compilation modes (
compilationModeFilter
). - A collection of
expectations
definitions that will be checked against the generated files.
Note that there is a JSON schema for the TEST_CASES.json
file stored at test_cases/test_case_schema.json
.
You should add a link to this schema at the top of your TEST_CASES.json
file to provide
validation and intellisense for this file in your IDE.
For example:
{
"$schema": "../test_case_schema.json",
"cases": [
{
"description": "description of the test - equivalent to an `it` clause message.",
"inputFiles": ["abc.ts"],
"expectations": [
{
"failureMessage": "message to display if this expectation fails",
"files": [
{ "expected": "xyz.js", "generated": "abc.js" }, ...
]
}, ...
],
"compilerOptions": { ... },
"angularCompilerOptions": { ... }
}
]
}
Input files
The input files are the source file that will be compiled as part of this test-case.
Input files should be stored in the directory next to the TEST_CASES.json
.
The paths to the input files should be listed in the inputFiles
property of TEST_CASES.json
.
The paths are relative to the TEST_CASES.json
file.
If no inputFiles
property is provided, the default is ["test.ts"]
.
Note that test-cases can share input files, but you should only do this if these input files are going to be compiled using the same options. This is because only one version of the compiled input file is retrieved from the golden partial file to be used in the linker tests. This can cause the linker tests to fail if they are provided with a compiled file (from the golden partial) that was compiled with different options to what are expected for that test-case.
Expectations
An expectation consists of a failureMessage
, which is displayed if the expectation check fails,
a collection of expected files
pairs and/or a collection of expectedErrors
.
Each expected file-pair consists of a path to a generated
file (relative to the build output folder),
and a path to an expected
file (relative to the test case).
The generated
file is checked to see if it "matches" the expected
file. The matching is
resilient to whitespace and variable name changes.
If no files
property is provided, the default is a a collection of objects {expected, generated}
,
where expected
and generated
are computed by taking each path in the inputFiles
collection
and replacing the .ts
extension with .js
.
Each expected error must have a message
property and, optionally, a location
property. These are
parsed as regular expressions (so .
and (
etc must be escaped) and tested against the errors that
are returned as diagnostics from the compilation.
If no failureMessage
property is provided, the default is "Incorrect generated output."
.
Expected file format
The expected files look like JavaScript but are actually specially formatted to allow matching with the generated output. The generated and expected files are tokenized and then the tokens are intelligently matched to check whether they are equivalent.
- Whitespace tolerant - the tokens can be separated by any amount of whitespace
- Code skipping - you can skip sections of code in the generated output by adding an ellipsis (…) to the expectation file.
- Identifier tolerant - identifiers in the expectation file that start and end with a dollar
(e.g.
$r3$
) will be matched against any identifier. But the matching will ensure that the same identifier name appears consistently elsewhere in the file. - Macro expansion - we can add macros to the expected files that will be expanded to blocks
of code dynamically. The following macros are defined in the
test_helpers/expected_file_macros.ts
file:- I18n messages - for example:
__i18nMsg__('message string', [ ['placeholder', 'pair] ], { meta: 'properties'})
. - Attribute markers - for example:
__AttributeMarker.Bindings__
.
- I18n messages - for example:
Running tests
The simplest way to run all the compliance tests is:
yarn test-ivy-aot //packages/compiler-cli/test/compliance/...
If you only want to run one of the three types of test you can be more specific:
yarn test-ivy-aot //packages/compiler-cli/test/compliance/full
yarn test-ivy-aot //packages/compiler-cli/test/compliance/linked
yarn test-ivy-aot //packages/compiler-cli/test/compliance/test_cases/...
(The last command runs the partial compilation tests.)
Updating a golden partial file
There is one golden partial file per TEST_CASES.json
file. So even if this file defines multiple
test-cases, which each contain multiple input files, there will only be one golden file.
The golden file is generated by the tooling and should not be modified manually.
When you first create a test-case, with an empty GOLDEN_PARTIAL.js
file, or a change is made to
the generated partial output, we must update the GOLDEN_PARTIAL.js
file.
This is done by running a specific bazel rule of the form:
bazel run //packages/compiler-cli/test/compliance/test_cases:<path/to/test_case>.golden.update
where to replace <path/to/test_case>
with the path (relative to test_cases
) of the directory
that contains the GOLDEN_PARTIAL.js
to update.
Debugging test-cases
The full and linked compliance tests are basically jasmine_node_test
rules. As such, they can be
debugged just like any other jasmine_node_test
. The standard approach is to add --config=debug
to the Bazel test command.
For example:
yarn test-ivy-aot //packages/compiler-cli/test/compliance/full --config=debug
yarn test-ivy-aot //packages/compiler-cli/test/compliance/linked --config=debug
To debug generating the partial golden output use the following form of Bazel command:
yarn bazel run //packages/compiler-cli/test/compliance/test_cases:generate_partial_for_<path/to/test_case>.debug
The path/to/test_case
is relative to the test_cases
directory. So for this TEST_CASES.json
file at:
packages/compiler-cli/test/compliance/test_cases/r3_view_compiler_directives/directives/matching/TEST_CASES.json
The command to debug the test-cases would be:
yarn bazel run //packages/compiler-cli/test/compliance/test_cases:generate_partial_for_r3_view_compiler_directives/directives/matching.debug
Focusing test-cases
You can focus a test case by setting "focusTest": true
in the TEST_CASES.json
file.
This is equivalent to using jasmine fit()
.
Excluding test-cases
You can exclude a test case by setting "excludeTest": true
in the TEST_CASES.json
file.
This is equivalent to using jasmine xit()
.