refactor: Replace temp-file + setTimeout pattern with positron.runtime.evaluateCode()

[atsyplenkov]

Summary

After exploring the Positron source code (posit-dev/positron), I identified a significant architectural improvement for how the extension communicates with the R runtime. The current approach of writing to temporary files and polling with setTimeout is fragile and race-condition prone. Positron provides a purpose-built API for this exact use case: positron.runtime.evaluateCode().

Current Pattern (Anti-pattern)

Multiple files (src/makeExplicit.js, src/fixLint.js, src/checkR.js) follow this flow:

1. Write data to a temp file.
2. Call positron.runtime.executeCode('r', ...).
3. await new Promise(resolve => setTimeout(resolve, 100)).
4. Read the temp file back.
5. Parse string prefixes like Error_001, Error_002.

This is unreliable because:

• 100ms may not be enough on slow systems.
• It leaves orphaned temp files if an error occurs before cleanup.
• It requires string-based error signaling instead of proper exception handling.

Recommended Pattern

Positron exposes evaluateCode() specifically for silent, promise-based evaluation:

export function evaluateCode(
    languageId: string,
    code: string,
    cancellationToken?: vscode.CancellationToken,
    sessionId?: string
): Thenable<EvalResult>;

export interface EvalResult {
    result: any;
    output: string;
}

Example: checkR.js

Before:

const r_check_command = `writeLines(as.character(...), con = '${normalizedPath}')`;
await positron.runtime.executeCode('r', r_check_command, false, false, positron.RuntimeCodeExecutionMode.Silent);
await new Promise(resolve => setTimeout(resolve, 100));
const output = await fs.readFile(tempFilePath, { encoding: 'utf8' });

After:

const evalResult = await positron.runtime.evaluateCode(
    'r',
    'as.character(c(nchar(system.file(package = "pedant")) != 0, nchar(system.file(package = "remotes")) != 0, nchar(system.file(package = "pak")) != 0))'
);
const [pedant, remotes, pak] = evalResult.result;

Other Improvements

1. Use finally blocks for temp file cleanup to prevent orphaned files.
2. Add try/catch around evaluateCode to handle R execution failures gracefully.
3. Use vscode.window.withProgress when running pedant::explicitize() on large selections.
4. Add @types/positron reference in jsconfig.json for better IntelliSense.

References

• Positron evaluateCode API definition: src/positron-dts/positron.d.ts
• Positron runtime tests: extensions/vscode-api-tests/src/singlefolder-tests/positron/runtime.test.ts