• Replace manual requirements.txt parsing with tree-sitter for better accuracy
• Add async/await support to provider methods for proper Promise handling
• Implement tree-sitter-based requirement parsing with version extraction
• Add web-tree-sitter and tree-sitter-requirements dependencies

New tree-sitter based requirements.txt parser module

src/providers/requirements_parser.js

Refactor to use tree-sitter parsing and async methods

src/providers/python_controller.js

Convert provider methods to async and use tree-sitter

src/providers/python_pip.js

Description

getDependencies() calls the now-async #installingRequirementsOneByOne() without awaiting it, so
dependency discovery and cleanup can run before installs finish. This can yield incomplete/incorrect
dependency graphs and pip install/uninstall races.

Code

src/providers/python_controller.js[R153-156]

				this.#installingRequirementsOneByOne()
			}
		}
-		let dependencies = this.#getDependenciesImpl(includeTransitive)
+		let dependencies = await this.#getDependenciesImpl(includeTransitive)

Evidence

The best-effort installation method was changed to async (it awaits requirement parsing), but the
caller does not await it and immediately proceeds to compute dependencies and then uninstall,
creating a race condition.

src/providers/python_controller.js[124-176]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`Python_controller.getDependencies()` calls `#installingRequirementsOneByOne()` without `await`, but that method is `async` and awaits parsing before installing packages. This introduces a race where dependency tree calculation and uninstall can run before installation completes.

## Issue Context
This happens only in the `TRUSTIFY_DA_PYTHON_INSTALL_BEST_EFFORTS=true` path.

## Fix Focus Areas
- src/providers/python_controller.js[129-176]

## Suggested fix
- Change `this.#installingRequirementsOneByOne()` to `await this.#installingRequirementsOneByOne()`.
- (Optional hardening) Replace `requirements.forEach(...)` with `for (const req of requirements) { ... }` to ensure deterministic sequential installs and clearer error propagation.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

python_pip.provideStack() calls createSbomStackAnalysis() twice, doubling runtime and repeating
heavy side effects like environment creation and pip operations. This can also produce inconsistent
results if installs/uninstalls happen twice.

Code

src/providers/python_pip.js[R45-52]

+async function provideStack(manifest, opts = {}) {
+	const sbom = await createSbomStackAnalysis(manifest, opts)
+	console.log(`WHAT ${JSON.stringify(sbom, null, '\t')}`)
	return {
		ecosystem,
-		content: createSbomStackAnalysis(manifest, opts),
+		content: await createSbomStackAnalysis(manifest, opts),
		contentType: 'application/vnd.cyclonedx+json'
	}

Evidence

provideStack computes the SBOM once into a variable (then logs it) but returns the result of a
second call. createSbomStackAnalysis invokes Python_controller.getDependencies which may
install/uninstall packages, so duplication is expensive and risky.

src/providers/python_pip.js[45-52]
src/providers/python_pip.js[186-201]
src/providers/python_controller.js[136-165]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`python_pip.provideStack()` calls `createSbomStackAnalysis()` twice. This doubles execution time and repeats expensive side effects (pip install/uninstall and/or virtualenv operations).

## Issue Context
`createSbomStackAnalysis()` constructs a `Python_controller` and calls `getDependencies()`, which may install dependencies and then uninstall them.

## Fix Focus Areas
- src/providers/python_pip.js[45-52]
- src/providers/python_pip.js[186-202]

## Suggested fix
- Replace `content: await createSbomStackAnalysis(...)` with `content: sbom`.
- Remove the `console.log(...)` line, or gate it behind `TRUSTIFY_DA_DEBUG === "true"` and log a bounded message (not full SBOM).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

The added GitHub dependency resolves in package-lock.json to a git+ssh URL, which commonly fails in
CI/user environments without GitHub SSH keys configured. This can block installation entirely.

Code

package.json[R56-57]

+		"tree-sitter-requirements": "github:Strum355/tree-sitter-requirements",
+		"web-tree-sitter": "^0.26.6",

Evidence

package.json adds a GitHub shorthand dependency; the lockfile shows npm resolved it via
git+ssh://git@github.com/..., which requires SSH-based git access and will often fail for public CI
runners or users without SSH setup.

package.json[46-58]
package-lock.json[7507-7512]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
The dependency `tree-sitter-requirements` is declared via GitHub shorthand. In `package-lock.json` it resolves to `git+ssh://git@github.com/...`, which frequently fails without SSH keys.

## Issue Context
This affects every `npm ci` / `npm install` consumer of this package.

## Fix Focus Areas
- package.json[56-57]
- package-lock.json[7507-7511]

## Suggested fix
- Update `package.json` to use an HTTPS git URL pinned to a commit/tag, e.g. `"tree-sitter-requirements": "git+https://github.com/Strum355/tree-sitter-requirements.git#<sha>"` (or a published semver npm package).
- Re-run `npm install` (or `npm ci` with regeneration as appropriate) to update `package-lock.json` so `resolved` uses `git+https` not `git+ssh`.
- Verify CI can install without SSH configuration.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

requirements_parser helpers each call init(), repeatedly running Parser.init() and Language.load()
which reloads WASM and the language multiple times per request. This adds avoidable latency and
could cause duplicate initialization behavior.

Code

src/providers/requirements_parser.js[R7-34]

+async function init() {
+	await Parser.init({
+		locateFile() {
+			return require.resolve('web-tree-sitter/web-tree-sitter.wasm')
+		}
+	});
+	return await Language.load(require.resolve('tree-sitter-requirements/tree-sitter-requirements.wasm'));
+}
+
+export async function getParser() {
+	const language = await init();
+	return new Parser().setLanguage(language);
+}
+
+export async function getRequirementQuery() {
+	const language = await init();
+	return new Query(language, '(requirement (package) @name) @req');
+}
+
+export async function getIgnoreQuery() {
+	const language = await init();
+	return new Query(language, '((requirement (package) @name) @req . (comment) @comment (#match? @comment "^#[\\t ]*exhortignore"))');
+}
+
+export async function getPinnedVersionQuery() {
+	const language = await init();
+	return new Query(language, '(version_spec (version_cmp) @cmp (version) @version (#eq? @cmp "=="))');
+}

Evidence

The module defines init() that performs expensive WASM initialization and language loading; every
exported accessor calls init() independently. Callers often request multiple accessors (parser +
multiple queries), multiplying the work.

src/providers/requirements_parser.js[7-34]
src/providers/python_controller.js[46-56]
src/providers/python_pip.js[92-96]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`requirements_parser.js` re-runs `Parser.init()` and `Language.load()` every time `getParser()` / `get*Query()` is called. Callers often call multiple helpers per request, causing redundant WASM work.

## Issue Context
Both `Python_controller` and `python_pip` call multiple helpers (parser + queries).

## Fix Focus Areas
- src/providers/requirements_parser.js[7-34]
- src/providers/python_controller.js[46-56]
- src/providers/python_pip.js[92-96]

## Suggested fix
- Add a module-level memoized `languagePromise` (or `initPromise`) so initialization is performed once.
- Implement `getLanguage()` that returns the cached promise.
- Update `getParser()` and all `get*Query()` functions to call `await getLanguage()` instead of `await init()`.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

Placeholder console logs (e.g., "BANANA"/"WHAT") are left in production code, and multiple tests are
commented out, reducing signal and coverage for key scenarios. This makes failures noisier and
increases regression risk for the new parsing logic.

Code

src/providers/python_controller.js[R102-114]

+	async #parseRequirements() {
+		const content = fs.readFileSync(this.pathToRequirements).toString();
+		const tree = (await this.parser).parse(content);
+		return Promise.all((await this.requirementsQuery).matches(tree.rootNode).map(async (match) => {
+			const reqNode = match.captures.find(c => c.name === 'req').node;
+			const name = match.captures.find(c => c.name === 'name').node.text;
+			const versionMatches = (await this.pinnedVersionQuery).matches(reqNode);
+			const version = versionMatches.length > 0
+				? versionMatches[0].captures.find(c => c.name === 'version').node.text
+				: null;
+			console.log(`BANANA ${name} ${version}`)
+			return { name, version };
+		}));

Evidence

The code prints requirement names/versions unconditionally during parsing and prints the full stack
SBOM in provideStack; meanwhile, python_pip tests disable several suites/cases instead of fixing
them or marking them skipped with rationale.

src/providers/python_controller.js[98-114]
src/providers/python_pip.js[45-51]
test/providers/python_pip.test.js[50-117]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
Production code contains unconditional debug `console.log(...)` statements, and multiple tests are commented out, lowering test coverage for important scenarios.

## Issue Context
The new tree-sitter parsing behavior is complex and needs strong regression coverage.

## Fix Focus Areas
- src/providers/python_controller.js[98-114]
- src/providers/python_pip.js[45-51]
- test/providers/python_pip.test.js[50-117]

## Suggested fix
- Remove the `console.log('BANANA ...')` and `console.log('WHAT ...')` statements, or wrap them with `if (process.env.TRUSTIFY_DA_DEBUG === 'true')` and avoid logging full SBOM payloads.
- Re-enable the commented-out tests; if they are flaky/slow, convert them to `test.skip(...)` or `suite.skip(...)` with a comment explaining why and how to re-enable, rather than commenting blocks out.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools