VSCode Transmute

Transmute means to change form: text transformations, encoding and encryption, data format conversion, and cURL-to-code generation in one extension. The Marketplace extension ID is gaopal.vscode-transmute, and the source repository is gaopal/vscode-transmute

Version 1.13.1 groups transformations under one menu and applies each operation to selected text or the entire document

Open the command palette and run Transform: Open Menu.... This is the only Transform entry in the palette. Choose a group, then a function; related modes and output choices appear inside that function

Group	Functions
Text / Lines	Sort, reverse, randomize, unique/count, filter, trim, join/split, JSON text, remove accents, Chinese-English spacing
CSV / Alignment	Align or compact CSV, align to cursor
Encoding / Crypto	Encodings, hashes, HMAC, AES encryption/decryption
Data Formats	Convert between JSON, YAML, XML, Properties, TOML and INI
cURL to Code	Choose a language and HTTP client
Selections / Helpers	Reorder selected fragments, copy input, selection tools, list VS Code commands

Text transformations use one scope rule: nonempty selections are independent inputs; without any selected text the input is the entire document, once. Empty cursors are ignored when real selections exist. Only selected characters are replaced, including partial lines. Reordering selected fragments requires at least two nonempty selections; selection tools keep their selection-specific behavior

Replacements are applied in one undo step after every input succeeds. Cancelling a menu or prompt leaves the source unchanged. Functions that open new documents produce a separate result for each input. Filter context is restricted to the input selections

All extension command IDs use the transmute.* namespace, for example transmute.openMenu and transmute.sortLines. Update custom keybindings and scripts to the new namespace; previous command IDs are no longer registered. Individual commands are hidden from the command palette and use the same automatic scope rule. The transmute.addChineseEnglishSpacingInDocument command is an alias for the unified spacing function, and transmute.escapes opens the unified UTF-8 encoding tools

flowchart TD
    A[Transform menu] --> B[Choose group and function]
    B --> C{Any selected text?}
    C -->|Yes| D[Each nonempty selection]
    C -->|No| E[Entire document once]
    D --> F[Transform and validate all inputs]
    E --> F
    F --> G[One edit or new result documents]

Publishing to Visual Studio Marketplace

The Publish VS Code extension workflow in .github/workflows/publish.yml supports two triggers

Manual: open the repository's Actions tab, choose the workflow and select Run workflow after the workflow exists on the default branch
Release: publish a stable GitHub Release whose tag is exactly v followed by the version in package.json, for example v1.13.1

Draft releases and pre-releases do not publish a stable Marketplace version. A regular code push does not trigger publication

Configure the gaopal publisher at Manage Publishers & Extensions before the first run. Under Trusted Publishing, create a policy with these exact values

Field	Value
Repository owner	`gaopal`
Repository	`vscode-transmute`
Workflow	`publish.yml`
Environment	`production`, matching the publish job and Marketplace policy

The publish job uses OIDC with id-token: write, so no VSCE_PAT secret is required. After renaming the repository, verify that the Marketplace Trusted Publishing policy matches the repository name and ID before publishing. See the official Trusted Publishing instructions

flowchart LR
    A[Manual run or stable Release] --> B[Build on Node 24]
    B --> C[Compile and run regression tests]
    C --> D[Package and verify VSIX]
    D --> E[Upload build artifact]
    E --> F[Download in Node 22 publish job]
    F --> G[Publish same VSIX with OIDC]

The publishing job only runs after the build job succeeds and downloads that job's VSIX. It does not rebuild or increment the version. A failed test, compilation or package validation stops publication

To run the build locally, use Node.js 24 or newer with npm and Git, then run

node scripts/build-release.cjs

The script downloads the shared library at commit e8dec053a8830a5efc2ada3cd8cd8735f6a3e9b3, installs isolated build tools and production dependencies from the official npm registry without lifecycle scripts or lock files, compiles both projects and runs npm run test:all. It then creates out/release/vscode-transmute-<version>.vsix and checks both packaged identity manifests. Temporary source, tools and dependencies stay under out/release/build-*; the project manifest and installed development environment are left intact

This project uses npm install --package-lock=false in its isolated build directories because it has no lock file. The shared library commit, direct runtime dependencies and build-tool versions are pinned; transitive dependency ranges can still resolve newer compatible versions on later runs

TypeScript checks and formatting

EditorConfig and Prettier use two spaces for indentation. Prettier preserves the file's existing line-ending convention and Markdown paragraph wrapping. TypeScript uses double quotes and semicolons; CommonJS .cjs scripts retain single quotes and omit semicolons

Command	Purpose
`npm run lint`	Check TypeScript source with ESLint
`npm run lint:fix`	Apply supported ESLint fixes to TypeScript source
`npm run format:check`	Report files that differ from the formatting rules
`npm run format`	Format supported project files, excluding dependencies and generated output
`npm run typecheck`	Check TypeScript types without writing compiled files

Run these commands after installing the project's development dependencies and making the shared library available. The isolated release build remains available through node scripts/build-release.cjs when the sibling library is not installed locally

ESLint 8.57.1, typescript-eslint 5.62.0 and eslint-config-prettier 8.10.0 are pinned to retain compatibility with TypeScript 3.9. Prettier is pinned to 3.6.2. These are legacy lint-tool versions; update them together with the compiler when migrating to a newer TypeScript version

ESLint's recommended rules check correctness and TypeScript usage. Unused names and explicit any are warnings during migration, and CommonJS require calls remain allowed for the existing Node.js and lazy-loading code. Prettier handles formatting; ESLint's overlapping formatting rules are disabled

The compiler includes src/**/*.ts, writes to out, checks filename casing and does not emit when compilation fails. The release script reads this same tsconfig.json, overriding only the isolated dependency type location. Strict mode is not enabled by this configuration change

Check commands report lint findings and formatting differences without modifying files. Run lint:fix or format explicitly to apply changes; automatic fixes on save are not enabled. Lint and formatting remain separate checks, while the release workflow runs compilation and regression tests

Features

Chinese-English Spacing
Convert cURL to Code
Encoding and Crypto Tools
Data Format Conversion
Align CSV
Align To Cursor
Compact CSV
Copy To New Document
Count Duplicate Lines As New Document
Encode / Decode
Filter Lines As New Document
Filter Lines
Join Lines
JSON String As Text
Lines As JSON String Array
Normalize Diacritical Marks
Randomize Lines
Randomize Selections
Reverse Lines
Reverse Selections
Rotate Backward Selections
Rotate Forward Selections
Select Highlights
Select Lines
Selection As JSON String
Sort Lines By Length
Sort Lines
Sort Selections
Split Lines After
Split Lines Before
Split Lines
Trim Lines
Trim Selections
Unique Lines As New Document
Unique Lines

Encoding and Crypto Tools

Choose Encoding / Crypto... in the Transform menu to select a conversion. Nonempty selections are processed independently, or the entire document when nothing is selected. Ordinary conversions validate every input before applying replacements in one undo step. If any input is invalid, nothing is replaced. Duplicate ranges are processed once, and overlapping selections are rejected

Category	Operations
Text encodings	UTF-8 Base64, Base64URL, Hex, URL, URL Component, form URL encoding
Escapes	HTML entities, decimal HTML entities, XML entities, Unicode escapes, JSON strings
Bytes	Text to/from JSON byte arrays, Base64 to/from byte arrays, Hex to/from byte arrays
Integers	Decimal integers to/from Crockford Base32
Hashes	MD5, SHA-1, SHA-256, SHA-512, each as Hex or Base64
Keyed hashes	HMAC-SHA256 and HMAC-SHA512, each as Hex or Base64
Encryption	Password-based AES-256-GCM encryption and decryption

Inspired by mitchdenny/ecdc. The legacy Encode / Decode command ID opens these tools. Encodings use UTF-8, including Chinese and emoji. Byte arrays accept integers from 0 through 255; text decoders reject invalid UTF-8, while byte-to-byte conversions can preserve arbitrary binary bytes. Crockford conversion accepts nonnegative integers through 9007199254740991, supports case-insensitive input, hyphens and the O/I/L aliases, and does not use the optional checksum

Hash and HMAC results are one-way values. HMAC keys are literal UTF-8 text entered in a masked prompt. These operations do not decrypt hashes

AES asks for a password in a masked prompt and opens its result in a new unsaved document, leaving the source intact. Multiple selections produce separate result documents. Cancelling the password prompt produces no output. Passwords are not persisted, and processing stays on the local machine

Encrypted text is a JSON envelope with version, algorithm, kdf, salt, iv, tag and ciphertext. Version 1 uses AES-256-GCM, a 16-byte random salt, a 12-byte random IV and a 16-byte authentication tag. The key is derived with scrypt (N=16384, r=8, p=1, 32-byte key), and the authenticated metadata is the UTF-8 string transformer:aes256gcm:v1. Binary fields use Base64. Fresh randomness produces different envelopes for the same text and password; the password and envelope are both needed for decryption. Incorrect passwords, unsupported envelope versions and authentication failures produce an error without replacing source text

Data Format Conversion

Choose Data Formats... in the Transform menu, then choose the source and target formats. The current document's language or filename puts the likely source format first in the menu. The command converts every nonempty selection independently, or the whole document when nothing is selected. All inputs must succeed before it opens separate unsaved result documents with available syntax highlighting

All 30 distinct source/target pairs between JSON, YAML, XML, Properties, TOML and INI are available. Conversion is for data rather than original formatting: comments, indentation, quoting style and object-key order are not preserved

flowchart LR
    A[Selection or document] --> B[Parse source]
    B --> C[Generate target]
    C --> D[Parse target again]
    D --> E{Data and types preserved?}
    E -->|Yes| F[Open result document]
    E -->|No| G[Show error and keep source]

The second parse checks that the target retains the supported structure and value types. A format pair being available does not mean every value can be represented in both formats. Unsupported mappings fail rather than silently dropping data or coercing types

Format	Mapping and limits
JSON and YAML	Objects, arrays, strings, booleans, finite numbers and null are supported; unsafe integer values, YAML sets, binary tags, cyclic aliases and non-string keys are rejected
TOML	Requires an object at the root; null and types that cannot fit the common data model are rejected, including native timestamp values rather than quoted date strings
XML	Preserves the root element, maps attributes to `@_name`, text with attributes to `#text`, and repeated elements to arrays; mixed content and interleaved repeated element groups are rejected because their order cannot fit this mapping
Properties	Dot-separated keys become nested objects and values stay strings; output requires nested objects with string leaves, with no empty or literal dotted keys, duplicate paths, empty nested objects or arrays
INI	Sections form nested objects, values stay strings, and `key[]` entries form arrays of strings; entries must use `key=value`, and duplicate scalar keys or conflicting paths are rejected

For example, {"app":{"port":"8080"}} converts to an INI [app] section with port=8080. The numeric JSON value 8080 instead of the string "8080" is rejected for INI because re-importing it would change its type to a string

XML output uses an existing single root key when possible. Otherwise it adds <root>; a top-level array is represented under <root><item>…</item></root>. XML has no native numeric, boolean or null types, and singleton arrays cannot be inferred from a lone element, so those mappings fail when type or array shape would be lost. XML declarations and layout-only whitespace between elements are omitted; DTDs, custom entities, invalid XML characters and reserved object keys are rejected. The keys __proto__, constructor and prototype are reserved throughout the common data model

Chinese-English Spacing

Choose Text / Lines → Add Chinese-English spacing in the Transform menu to insert spaces between Chinese characters and adjacent English letters or digits. It processes all nonempty selections, or the whole document when nothing is selected

使用VSCode处理100行文本
使用 VSCode 处理 100 行文本

The commands preserve punctuation, existing whitespace and line endings. English letters and numbers stay together, so English123 is unchanged. Repeating the command does not add more spaces, and each operation uses a single undo step. These commands format text, including selected code, without interpreting its programming language

Inspired by vscode-pangu. Formatting runs only when you invoke a command

Convert cURL to Code

Choose cURL to Code... in the Transform menu. Each nonempty selection is a separate curl input; without a selection, the command reads the entire document. All inputs must convert successfully before opening output documents. Choose a language and HTTP client library to open the generated code in a new unsaved document. The original text stays intact, and cancelling the menu opens nothing

The converter uses Postman's open-source curl importer and code generators. The menu includes all 33 non-cURL variants supplied by the pinned generator version, across 19 language and output groups

Language or output	Client libraries
C	libcurl
C#	HttpClient, RestSharp
Dart	dio, http
Go	Native
HTTP	HTTP
Java	OkHttp, Unirest
JavaScript	Fetch, jQuery, XHR
Kotlin	Okhttp
Node.js	Axios, Native, Request, Unirest
Objective-C	NSURLSession
OCaml	Cohttp
PHP	cURL, Guzzle, HTTP_Request2, pecl_http
PowerShell	RestMethod
Python	http.client, Requests
R	httr, RCurl
Ruby	Net::HTTP
Rust	reqwest
Shell	Httpie, wget
Swift	URLSession

For example, select the following request and choose Python - Requests

curl 'https://example.com/api/items' \
  -H 'Content-Type: application/json' \
  --data-raw '{"name":"demo"}'

Parsing and generation run locally without executing curl or sending an HTTP request. Supported request features follow the Postman importer, including common headers, authentication, query strings, JSON, form bodies and shell line continuations. This is request conversion rather than a shell interpreter: shell variables, command substitution and arbitrary curl flags are not guaranteed to translate. File references remain references rather than uploading or reading local files

Generated code keeps request credentials when they are part of the input. The chosen language's libraries must be installed in the destination project. Syntax highlighting uses installed VS Code language support, with plain text as the fallback. This feature does not use Postman accounts, collections or cloud services

Unique Lines

Removes duplicate lines from the document
Operates on selection or entire document if no selection

Unique Lines As New Document

Unique lines are opened in a new document
Operates on selection or entire document if no selection

Filter Lines

Keep matching lines of filter
Operates on selection or entire document if no selection
Filter using regex or literal

Filter Lines As New Document

A new document is created with lines matching filter
Gutter decorators show original line number from original document
Operates on selection or entire document if no selection
Filter using regex or literal
Add lines relative to filtered lines by count and regular expression match
Parent Levels add context by indentation level. Similar to vscode folding levels
- Examples
  - 0 will include all siblings of the matched lines. Where a sibling is determined by being at the same indentation level next to the matched line
  - 1 will include all siblings + the parent siblings ( one less indentation level )