This commit applies `strictNullChecks` to the entire codebase to improve
maintainability and type safety. Key changes include:
- Remove some explicit null-checks where unnecessary.
- Add necessary null-checks.
- Refactor static factory functions for a more functional approach.
- Improve some test names and contexts for better debugging.
- Add unit tests for any additional logic introduced.
- Refactor `createPositionFromRegexFullMatch` to its own function as the
logic is reused.
- Prefer `find` prefix on functions that may return `undefined` and
`get` prefix for those that always return a value.
This commit fixes compiler bug where it fails when optional values are
compiled into absent values in nested calls.
- Throw exception with more context for easier future debugging.
- Add better validation of argument values for nested calls.
- Refactor `FunctionCallCompiler` for better clarity and modularize it
to make it more maintainable and testable.
- Refactor related interface to not have `I` prefix, and
function/variable names for better clarity.
Context:
Discovered this issue while attempting to call
`RunInlineCodeAsTrustedInstaller` which in turn invokes `RunPowerShell`
for issue #246. This led to the realization that despite parameters
flagged as optional, the nested argument compilation didn't support
them.
Enable `contextIsolation` in Electron to securely expose a limited set
of Node.js APIs to the renderer process. It:
1. Isolates renderer and main process contexts. It ensures that the
powerful main process functions aren't directly accessible from
renderer process(es), adding a security boundary.
2. Mitigates remote exploitation risks. By isolating contexts, potential
malicious code injections in the renderer can't directly reach and
compromise the main process.
3. Reduces attack surface.
4. Protect against prototype pollution: It prevents tampering of
JavaScript object prototypes in one context from affecting another
context, improving app reliability and security.
Supporting changes include:
- Extract environment and system operations classes to the infrastructure
layer. This removes node dependencies from core domain and application
code.
- Introduce `ISystemOperations` to encapsulate OS interactions. Use it
from `CodeRunner` to isolate node API usage.
- Add a preloader script to inject validated environment variables into
renderer context. This keeps Electron integration details
encapsulated.
- Add new sanity check to fail fast on issues with preloader injected
variables.
- Improve test coverage of runtime sanity checks and environment
components. Move validation logic into separate classes for Single
Responsibility.
- Improve absent value test case generation.
As part of transition to Vue 3.0 and Vite (#230), this commit
facilitates the shift towards building rest of the application using
Vite. By doing so, it eliminates reliance on outdated Electron building
system that offered limited control, blocking desktop builds (#233).
Changes include:
- Introduce Vite with Vue 2.0 plugin for test execution.
- Remove `mocha`, `chai` and other related dependencies.
- Adjust test to Vitest syntax.
- Revise and update `tests.md` to document the changes.
- Add `@modyfi/vite-plugin-yaml` plugin to be able to use yaml file
depended logic on test files, replacing previous webpack behavior.
- Fix failing tests that are revealed by Vitest due to unhandled errors
and lack of assertments.
- Remove the test that depends on Vue CLI populating `process.env`.
- Use `jsdom` for unit test environment, adding it to dependency to
`package.json` as project now depends on it and it was not specified
even though `package-lock.json` included it.
Rework code validation to be bound to a context and not
context-independent. It means that the generated code is validated based
on different phases during the compilation. This is done by moving
validation from `ScriptCode` constructor to a different callable
function.
It removes duplicate detection for function calls once a call is fully
compiled, but still checks for duplicates inside each function body that
has inline code. This allows for having duplicates in final scripts
(thus relaxing the duplicate detection), e.g., when multiple calls to
the same function is made.
It fixes non-duplicates (when using common syntax) being misrepresented
as duplicate lines.
It improves the output of errors, such as printing valid lines, to give
more context. This improvement also fixes empty line validation not
showing the right empty lines in the error output. Empty line validation
shows tabs and whitespaces more clearly.
Finally, it adds more tests including tests for existing logic, such as
singleton factories.
- Unify test data for nonexistence of an object/string and collection.
- Introduce more test through adding missing test data to existing tests.
- Improve logic for checking absence of values to match tests.
- Add missing tests for absent value validation.
- Update documentation to include shared test functionality.
Major refactoring using ESLint with rules from AirBnb and Vue.
Enable most of the ESLint rules and do necessary linting in the code.
Also add more information for rules that are disabled to describe what
they are and why they are disabled.
Allow logging (`console.log`) in test files, and in development mode
(e.g. when working with `npm run serve`), but disable it when
environment is production (as pre-configured by Vue). Also add flag
(`--mode production`) in `lint:eslint` command so production linting is
executed earlier in lifecycle.
Disable rules that requires a separate work. Such as ESLint rules that
are broken in TypeScript: no-useless-constructor (eslint/eslint#14118)
and no-shadow (eslint/eslint#13014).
It allow pipes to be used in nested functions. Before, pipes were added
to a variable before variable content was evaluated/compiled by
another function. This commit ensures that the commits are evaluted in
expected order.
The issue is solved by stopping precompiling functions. It makes code
less complex. It adds to compile time of the script file but nothing
noticable and something optimizable.
The problem was that the call trees we're not executed in expected
order. E.g. let's say we have functionA that outputs something like
"Hello {{ $name| pipe }}", and we have function B calling with "name:
dear {{ $firstName}}", and at last we have a script that's calling
function B with "firstName: undergroundwires". Before, expressions were
evaluated directly, meaning that function A would become:
"Hello Dear {{ $firstName}}", as you see the pipe in function A
is lost here after being applied to function B and not reaching
$firstTime input value. Parsing expressions in the end allows for pipes
etc. to not get lost.
The commit also does necessary name refactorings and folder refactorings
to reflect logical changes. `FunctionCompiler` is renamed to
`SharedFunctionsParser` as precompiling is removed and it just simply
parses now. `/FunctionCall/` is moved to `/Function/Call`.
Finally, it improves documentation and adds more tests.
This commit allows for parameters that does not require any arguments to
be provided in function calls. It changes collection syntax where
parameters are list of objects instead of primitive strings. A
parameter has now 'name' and 'optional' properties. 'name' is required
and used in same way as older strings as parameter definitions.
'Optional' property is optional, 'false' is the default behavior if
undefined. It also adds additional validation to restrict parameter
names to alphanumeric strings to have a clear syntax in expressions.
