AlgoSnap

Algorithm templates that fit your editor context. Works offline in VS Code.

Insert tested Python, C++17, and Java algorithms and data-structure examples from a searchable picker. AlgoSnap suggests parameter names from preceding code and lets you edit linked names with Tab. No account, API key, backend, or network request is needed.

Quick start

1. Open a Python, C++, or Java source file.
2. Put a single cursor on an empty or indented blank line.
3. Run AlgoSnap: Insert Algorithm Template from the Command Palette, or press Ctrl+Shift+A (Cmd+Shift+A on macOS).
4. Select a template; Tab through the parameter names. Escape exits snippet mode.
5. One Undo removes the insertion (later edits have their own undo entries).

Example: with data = [1, 3, 5] and key = 3 above the cursor, the binary search template suggests def binary_search(data, key): and updates all parameter references when you rename a placeholder.

Automatic suggestions

Type binary search, two pointer, sliding window, dfs, bfs, dp, tree, graph, trie, or heap on its own line, then pause. The picker replaces that keyword after you select a variant. Keywords are case-insensitive; aliases such as binary_search, bisect, and flood fill also work. Comments, common string literals, pasted multi-character text, undo/redo, selections, and multi-cursor edits do not open the picker. Moving the cursor, changing documents, or editing while the picker is open cancels it so an outdated selection cannot overwrite new text.

Included templates

There are 68 native-language implementations: 34 Python, 20 C++, and 14 Java. The eight core variants below work in all three languages (24 combinations). The restored catalog contributes 44 more implementations with explicit language support; the picker never falls back to code from another language.

Aliases such as knapsack, lcs, coin change, dijkstra, union find, segment tree, and priority queue select the corresponding variants. Type data structures to browse the collection. Shared variants such as DSU appear only once in the manual picker. See catalog migration for the preservation map and changed contracts.

Lower/upper bound return the first index with value >= / > the target, or the array length. Exact/rotated search return -1 when absent. Two sum returns zero-based indices, or an empty result. Grid DFS supports empty grids and uses an explicit stack to avoid recursive stack overflow.

Placement and prerequisites

• Python: insert at module scope for examples containing imports/classes/sample statements; indentation follows VS Code. All dependencies are standard-library modules.
• C++: insert outside other functions, use C++17 or later, and add the headers named in the generated comment. Standard library names are qualified with std::.
• Java: insert inside a class, outside other methods. Generated methods are static where shown; collection classes are fully qualified, so no import needs to be inserted inside a class. Data-structure demos are self-contained example methods.
• C is not supported. Java inputs must be non-null (including grid rows). C++ collection dimensions must fit in int; sums use long long / Java long.
• Read each variant's picker details for input preconditions. DP requires inputs small enough for its stated memory complexity; graph vertices/weights and tree structure must follow the displayed contract. Code is a reusable starting point, not an automatic conversion of surrounding data structures.

Settings

Both settings may be overridden per workspace/language. Keyboard shortcuts are customizable in VS Code's Keyboard Shortcuts editor. If another extension uses the same shortcut, use the Command Palette or remap AlgoSnap.

Context and privacy

The scanner masks comments and literals, then uses lightweight declaration patterns to suggest names from preceding code. It is not a compiler, AST, or scope/type checker. Review suggestions; a nearby variable name does not prove that its type or scope matches the generated function. Grid DFS uses grid as an editable default. Unsupported declarations fall back to nums, target, k. Restored multi-parameter algorithms use stable descriptive defaults (weights, capacity, graph, etc.) with explicit linked tab stops; the flat scanner cannot reliably infer these roles. Standalone examples/classes may have no editable tab stops.

To keep editor work bounded, automatic triggers only scan prefixes up to 200,000 characters. Past that position, the manual command still works with default names. Unusual syntax (for example C++ line splicing or macros) may require disabling auto-trigger. VS Code does not identify the origin of every edit; single-character programmatic inserts/pastes can resemble typing.

The extension does not execute generated code, inspect other files, transmit source, or collect telemetry. See PRIVACY.md.

Local installation

In VS Code, choose Extensions > ... > Install from VSIX... and select the provided algosnap-0.3.0.vsix. Or run:

code --install-extension algosnap-0.3.0.vsix

The configured publisher is SaumyaJoshi2005. This package has not been published on Marketplace. If you installed the earlier algosnap-local.algosnap build, uninstall it before installing this build to avoid duplicate commands.

Development

Requires Node.js 22.13 or later. In Windows PowerShell, use npm.cmd if script execution policy blocks npm.ps1; changing execution policy is unnecessary.

npm ci
npm test
code .

Press F5 to launch the configured Extension Development Host. The build runs automatically. npm run watch recompiles edits during development.

npm run test:algorithms    # Python 3, g++ with C++17, javac/java on PATH
npm run test:integration   # Downloads an isolated VS Code test instance
npm run package           # Builds/tests, then creates a VSIX

For integration tests with an existing VS Code executable, set VSCODE_EXECUTABLE_PATH to its full path. The runner uses isolated test user data and extensions. CI tests the minimum supported VS Code (1.85.2) and stable.

The algorithm runner accepts ALGOSNAP_PYTHON, ALGOSNAP_CXX, ALGOSNAP_JAVAC, and ALGOSNAP_JAVA as full executable paths if your preferred tools are not on PATH.

See CONTRIBUTING.md, architecture, and publishing instructions. Report bugs with reproduction steps at GitHub Issues.

Release notes

Version 0.3.0 restores the 44 native implementations found in GitHub but absent from the original ZIP, with fixes and per-variant runtime checks. The core still omits incomplete is_valid/feasible skeletons and the mixed import-and-statement Java wrapper, as documented in 0.2.0. See CHANGELOG.md.