Data Convert (Visual Studio Code)

中文文档

No network connection, Convert data formats and encodings inside the editor. The result opens in a scratch editor beside the source. Command Palette only, no UI.

Usage

1. Open a file or an untitled buffer. Select a fragment if you only want to convert part of it.
2. Press Cmd/Ctrl+Shift+P and run Data Convert: Convert Format... (data formats) or Data Convert: Encode / Decode... (encodings).
3. Pick an entry from the list, such as YAML → JSON. The list is fuzzy-searchable — type yamljson to jump straight to it.

With no selection the whole document is converted; with a selection only the selection is.

Data format conversion

Data Convert: Convert Format...

Code generation

Type definitions are inferred from your sample data, targeting TypeScript, JSON Schema, Zod, Go, Rust, Python, Java and C#. The output contains type definitions only, without the serialization helpers each language would otherwise ship with.

Inference is driven entirely by the sample: whether a field is optional, how wide a numeric type is — all of it comes from the data you give it. The closer your sample is to real data, the better the result.

Encoding conversion

Data Convert: Encode / Decode...

These are plain text-to-text transforms and do not go through an intermediate representation. A few conventions:

• Query String → JSON accepts a whole URL — everything after ? is used. Repeated keys collapse into an array, and JSON → Query String expands arrays back into repeated keys. Bracket syntax is supported both ways, so a[b]=1 becomes a nested object.
• Text → Unicode Escape escapes non-ASCII characters and backslashes, with emoji escaped one code unit at a time as surrogate pairs. Backslashes have to be escaped for the transform to be reversible — otherwise a literal \u4e2d already in your text would decode back as a character. Decoding accepts \\, \uXXXX and \u{XXXXX}.
• Text → HTML Entities also turns non-ASCII into numeric references (中 → 中); (Minimal) escapes only &<>"' and leaves CJK and emoji intact. HTML Entities → Text reverses both.
• JWT → JSON decodes the header and payload only. It does not verify the signature — the signature segment is passed through untouched.
• Hashes are one-way, so they run in a single direction and output lowercase hex. The input is not preprocessed, so newlines inside your selection count toward the digest (matching md5sum). MD5 and SHA-1 are no longer collision-resistant; use them for checksums and legacy compatibility, not for security.

Settings

How it behaves

Conversions run exactly the rule you picked. There is no source-format detection, and no validation or fallback on the input: if you choose TOML → JSON, the input is expected to be TOML. When it isn't, the parser's own error is reported verbatim.

Formats differ in what they can express, and the extension does not compensate for the gaps:

• Comments are not preserved — the intermediate representation carries data only.
• .env and .properties express strings only, so numbers and booleans come back as strings, and nested values in .env are serialized to JSON strings.
• XML output is always wrapped in a single <root> element, with a top-level array wrapped once more in <item>.
• TOML requires a table at the top level, so smol-toml reports an error when the source is an array.
• CSV / TSV need an array of rows ([{...}, {...}]); a flat object will not convert. Cells hold scalars only, nested values are serialized to JSON strings, and parsed values are always strings — no type inference.

Privacy

The extension makes no network requests and stores nothing. Conversions run entirely in the extension host.

License 📃

MIT License