Data Grid Table Editor and Data Generator

A simple Data Table Editor that generates Markdown, CSV or JSON. It can also be used to interchange data between the formats, supporting editing in the grid.

• Grid interface
• drag and drop rows to re-order
• drag and drop columns to re-order
• import csv from file
• import csv, markdown, and json from editor
• generate csv, markdown, json and javascript from grid
• export csv, markdown and json files
• add and delete columns
• add and delete rows
• randomly fill data with Test Data
• configure export options

---

Use The Editor

The application is live at AnyWayData.com

• Or clone and run locally with Vite:
• pnpm install
• pnpm run dev:web
• open http://127.0.0.1:4173/app.html, http://127.0.0.1:4173/generator.html, or http://127.0.0.1:4173/webmcp.html

Use pnpm run build:web to create a production build and pnpm run preview:web to preview the built output. The old static-server flow (e.g. python3 -m http.server) is no longer the recommended local runtime path.

Use Storybook

Storybook is available for isolated frontend development and UI review:

• pnpm run storybook
• open http://127.0.0.1:6006

Current story groups:

• Test Data / Embedded Panel for the in-app schema editor inside the main table editor
• Test Data / Generator for the standalone generator schema editor states
• Export Formats / Previews for export-preview flows such as Markdown, CSV, DSV, JSON, JSONL, XML, SQL, HTML, Gherkin, ASCII table, and code-oriented outputs

Useful workflows:

• use Storybook when iterating on schema-editor rendering, validation states, and text/schema mode transitions
• use Storybook when adjusting export-preview UI without needing the full app boot flow
• use pnpm run test:storybook to run Storybook smoke and interaction tests in CI-style mode
• use pnpm run build-storybook to create a static Storybook build in storybook-static

Select Grid Engine

The editor can run with AG Grid or Tabulator using the same import/export and toolbar processing layer.

• Query string: ?grid=ag-grid or ?grid=tabulator
• Global override before app load: window.ANYWAYDATA_GRID_ENGINE = "ag-grid"
• Persisted setting key: localStorage["anywaydata:grid-engine"]

Default is Tabulator.

To Generate Test Data

Use the Generator to create test data rules using a nicer UI.

You can preview the data prior to generating.

You can copy the schema into the editor UI and use it to populate existing tables.

To Generate Test Data in an Grid

You can add data to an existing data table e.g. you import a CSV file and want to a new column of dates, or you want to amend an entire column (perhaps to obfuscate PII Data)

Expand the "> Test Data" section.

Choose the Data fields to add or amend in the table or enter a spec.

You can then create a new table, or amend the existing table or selected rows.

Test Data Text Spec

The spec is a paragraph of text where each line is either a 'name' or a 'rule':

Schema Formatting

• Comments: lines starting with # (optionally prefixed by whitespace) are treated as comments.
• Blank lines: blank lines are allowed and ignored, so you can separate column groups for readability.
• Column definitions: each column is defined as name followed by rule on the next logical content line.
• Constraints: optional IF ... THEN ... statements may appear in text mode after the field definitions, terminated by either ; or ENDIF.

# optional comment

name
rule

# another comment
name
rule

IF [name] = "Bob" THEN [status] = "active" ENDIF

• name will be used as a column name
• rule will be used to generate the data
• constraint parameter references must use bracketed column names, e.g. [Status]
• constraints are currently text-mode only in the editor UI
• pairwise / n-wise generation with constraints supports only constraints that reference enum columns

A rule can be a regex string e.g.

• (connie|bob) which would generate 'connie' or 'bob'
• [1-9][0-9]{0,4} which would generate number between 1 and 99999

A rule can also be a domain mapped faker API call.

Faker API can be found here: https://fakerjs.dev/api/

e.g.

First Name
awd.domain.person.firstName

Noun
domain.hacker.noun

The awd and domain prefixes are optional:

e.g.

First Name
person.firstName

Noun
hacker.noun

The fake method on faker is also supported, which takes a mustache template style string combining api methods e.g.

• helpers.fake("{{name.lastName}}, {{name.firstName}}")

So a sample test data spec might look like:

# person details
name
helpers.fake("{{name.lastName}}, {{name.firstName}}")

# profile text
desc
lorem.paragraph

# preference data
collects
hacker.noun

# regex example
prefers
(Connie|Bob)

Pairwise Combinatorial Test Data

When you have 2 or more enum fields (comma-separated values), you can generate pairwise combinatorial test data using a greedy approximation approach. This provides 100% pairwise coverage and typically reduces test cases substantially (often around 90-99% fewer) compared with full factorial testing.

For enum data, use comma-separated values in your spec:

# pairwise parameters
browser
chrome,firefox,safari,edge

# viewport class
device
desktop,tablet,mobile

# style variant
theme
light,dark

The "Generate Pairwise" button will appear automatically, creating a compact near-minimal set of combinations that tests every pair of values across all parameters.

Similar Apps

Looking for similar apps to compare features sets and functionality?

There is a maintained list in the documentation:

• Related Tools

References

• Markdown Tables Extended Syntax

Libraries Used

• Faker.js for backing the domain api
  • https://fakerjs.dev/
• RandExp.js for regular expression based random data generation
  • http://fent.github.io/randexp.js/
• Tabulator for light weight tables
• PapaParse for csv processing

Tool Categories

This tool falls in to the categories:

• online markdown editor
• markdown table generator
• markdown table editor
• Online Test Data Generation
• Online CSV Editor

Building

Test - pnpm test

Coverage - pnpm run testcoverage

Preview Docs - pnpm run previewdocs

Build for release - pnpm run anywaydata:win

Monorepo Workspaces

This repository now uses pnpm workspaces:

• packages/core -> shared generation engine (@anywaydata/core)
• packages/core-ui -> browser UI modules (@anywaydata/core-ui)
• apps/cli -> npm CLI package (@anywaydata/cli)
• apps/api -> REST API (@anywaydata/api)
• apps/mcp -> MCP server (@anywaydata/mcp)

Imports now resolve directly from package sources. Use packages/core/js/* and packages/core-ui/js/* paths in local runtime/test code.

Install all dependencies from the repo root:

pnpm install

Run workspace build/test orchestration from root:

• pnpm run build:workspaces
• pnpm run test:workspaces
• pnpm run test:workspaces:all

Notes:

• pnpm run test:workspaces runs only the non-duplicating workspace tests that are not already covered by the root Jest suite (@anywaydata/api, @anywaydata/cli, and @anywaydata/mcp).
• pnpm run test:workspaces:all preserves the old “run every workspace test script” behavior.

Run an individual workspace command:

• pnpm --filter @anywaydata/core run test
• pnpm --filter @anywaydata/core-ui run test
• pnpm --filter @anywaydata/api run start
• pnpm --filter @anywaydata/mcp run start

npm CLI Usage (@anywaydata/cli)

The npm CLI package is the workspace app apps/cli (@anywaydata/cli), and Bun executable builds also use apps/cli as the source.

Install globally from npm:

npm install -g @anywaydata/cli

Or run without installing:

npx @anywaydata/cli --help

Run a one-off generation command with npx:

npx @anywaydata/cli generate -i input.txt -n 10 -f csv

Show CLI help:

anywaydata --help

Generate output to console:

anywaydata generate -i input.txt -n 10 -f csv

Generate output to a file:

anywaydata generate -i input.txt -n 10 -f json -o output.json

Run in test mode (forces a single generated row and prints diagnostics):

anywaydata generate -i input.txt -n 10 -f markdown -t

Supported options:

• -i, --inputfile path to the input text spec (required)
• -n, --numberOfLines number of rows to generate (default 1)
• --pairwise generate pairwise combinations for enum fields (requires at least 2 enum rules)
• -f, --format output format (default csv)
• -o, --outputfile write output to file instead of stdout
• -t, --testMode enable diagnostics mode and generate one row
• --unsafe-faker-expressions allow expression-style faker args (disabled by default)
• --trim-input trim whitespace from every imported field value before amend processing
• --trim-input-fields comma-separated imported field names to trim before amend processing

Amend existing data with a schema:

anywaydata amend --schema-file schema.txt --data-file input.csv --input-format csv -f json -o amended.json

Trim imported input values during amend:

anywaydata amend --schema-file schema.txt --data-file input.csv --input-format csv --trim-input-fields Name,Email -f json

Pairwise note:

• when --pairwise is enabled, --numberOfLines is accepted but ignored (the pairwise engine determines row count)

REST API Quick Start

Start the API:

pnpm --filter @anywaydata/api run start

Run the published API package directly with npx:

npx -y @anywaydata/api

Note: If npx fails on Windows, use direct execution instead:

npm install -g @anywaydata/api && anywaydata-api

For interactive local development (foreground process with Ctrl+C stop), prefer:

• pnpm --filter @anywaydata/api run start -- --port 3001
• node apps/api/src/index.js --port 3001

Use npx primarily for consumer-style package execution/testing.

Choose a specific port:

• pnpm --filter @anywaydata/api run start -- --port 3001
• PowerShell: $env:PORT=3001; pnpm --filter @anywaydata/api run start

Port behavior:

• --port overrides PORT
• if a port is explicitly provided (--port or PORT) and is in use, startup fails with a short message
• if no explicit port is provided, API starts at 3000 and auto-tries 3001..3020 when needed

Health check:

GET http://localhost:3000/v1/health

Generate data:

POST http://localhost:3000/v1/generate

/v1/generate request supports optional responseFormat mode:

• responseFormat: "rows" (default) returns headers + rows
• responseFormat: "rendered" returns only rendered text in JSON payload
• responseFormat: "all" returns headers + rows + rendered
• responseFormat: "raw" returns rendered payload directly with content-type matching output format
• outputFormat is optional and defaults to csv
• outputFormat supports: csv, dsv, markdown, json, jsonl, javascript, python, java, typescript, xml, sql, gherkin, html, asciitable

Raw multiline schema/spec endpoint:

POST http://localhost:3000/v1/generate/fromschema?rowCount=10&outputFormat=csv&responseFormat=rows

• Content-Type must be text/plain
• request body is the full multiline spec
• query params: rowCount (required), outputFormat (optional), seed (optional), pairwise (optional), responseFormat (optional: rows|rendered|all|raw)
• pairwise note: when pairwise=true, rowCount is accepted for compatibility but ignored (pairwise output size is computed automatically)

Amend imported data endpoint:

POST http://localhost:3000/v1/generate/amend

• JSON body fields: textSpec (required), inputData (required raw text), inputFormat (required), rowCount (optional, defaults to input row count), outputFormat (optional), responseFormat (optional), stream (optional and ignored)
• amend trim controls: trimInput trims all imported field values; trimInputFieldsCsv trims only the listed imported field names
• rowCount must be <= imported row count
• response returns the full resulting dataset after amendment

Set format default options:

POST http://localhost:3000/v1/generate/options/<format>

Get current format default options:

GET http://localhost:3000/v1/generate/options/<format>

Reset format default options:

POST http://localhost:3000/v1/generate/options/<format>/default

Notes:

• option defaults set for a format are used when options are not provided to /v1/generate or /v1/generate/fromschema
• payload for /v1/generate/options/<format> is the same options object normally sent as options in generation requests

Notes:

• for JSON requests, newline characters inside string literals must be escaped as \\n
• use /v1/generate/fromschema when you want to paste raw multiline text directly

Streaming notes for CLI/core generation:

• stream mode supports: csv, jsonl, dsv, json, xml
• json stream mode emits a valid JSON array payload
• incompatible format options in stream mode are warned and ignored where needed
• REST API generation endpoints currently run in buffered mode

OpenAPI spec:

GET http://localhost:3000/v1/openapi.json

Swagger UI:

GET http://localhost:3000/v1/docs

Example request body:

{
  "textSpec": "# literal example\\n\\nName\\nBob",
  "rowCount": 3,
  "outputFormat": "json"
}

Security: Unsafe Faker Expressions

For security, complex faker expressions that use eval-like functionality are disabled by default. The API uses a hybrid approach that tries safe method calls first, then falls back to secure execution when needed.

Enable Unsafe Expressions Globally

Start the API with unsafe expressions enabled for all requests:

# Via command line flag
node apps/api/src/index.js --unsafe-faker

# Or via npm workspace
pnpm --filter @anywaydata/api run start -- --unsafe-faker

# Or via npx
npx -y @anywaydata/api --unsafe-faker

Enable Unsafe Expressions Per Request

Add unsafeFakerExpressions: true to individual requests:

{
  "textSpec": "# faker + numeric range\\n\\nName\\nperson.firstName\\n\\nScore\\nnumber.int({\"min\": 18, \"max\": 65})",
  "rowCount": 5,
  "outputFormat": "json",
  "unsafeFakerExpressions": true
}

For /fromschema endpoint, use the query parameter:

POST /v1/generate/fromschema?rowCount=10&unsafeFakerExpressions=true

Security Details

• Safe by default: Most faker expressions use direct method calls without eval
• Hybrid approach: Tries safe .apply() method first, falls back to Function() constructor with security filtering
• Expression patterns blocked: process, require, import, eval, __, constructor, prototype
• Granular control: Enable globally via CLI or per-request via parameter

Examples of expressions that work safely without the flag:

• person.firstName or person.firstName()
• number.int({"min": 10, "max": 100})
• helpers.arrayElement(["red", "green", "blue"])

Examples that require the unsafe flag:

• () => this.person.firstName() (arrow functions with this)
• {count: () => \${this.number.int()}`}` (template literals)

MCP Quick Start

Start the MCP server:

pnpm --filter @anywaydata/mcp run start

Note: MCP runs over stdio in this version and does not bind to a TCP port. OpenAPI/Swagger routes are available on the REST API only.

The server exposes tools:

• generate_data_from_spec
• amend_data_from_spec
• get_output_format_options_schema

Inputs:

• textSpec (required string)
• rowCount (required integer, >= 0)
• pairwise (optional boolean, default false; when true, rowCount is ignored)
• outputFormat (required string e.g. csv, json, jsonl, xml, sql)
• options (optional object)
• seed (optional number)
• trimInput (optional boolean for amend_data_from_spec)
• trimInputFieldsCsv (optional string for amend_data_from_spec)

Discoverability support:

• tools/list publishes typed inputSchema and outputSchema
• get_output_format_options_schema returns per-format defaults and typed option schema
• resources/list and resources/read expose:
  • anywaydata://schemas/output-format-options
  • anywaydata://install/config-examples

Security note:

• MCP input validation restricts faker rule arguments to literals only (string/number/boolean/null) or no args.
• Expression-style faker arguments (e.g. arrow functions, this, template expressions) are rejected in MCP mode.
• Node and Bun CLIs are also safe-by-default with the same rule; use --unsafe-faker-expressions to opt in.

Error conventions:

• Tool-level validation/safety failures return result.isError=true
• Tool payload shape for failures is:
  • { "ok": false, "error": { "code": "...", "message": "...", "details": ... } }
• Common error.code values:
  • invalid_output_format
  • unsafe_faker_rule
  • text_spec_too_large
  • row_count_too_large
  • generation_failed

Using MCP With Codex

Codex connects to this server as an MCP tool provider over stdio (not HTTP).

1. Ensure dependencies are installed from repo root:
   • pnpm install
2. Start MCP server (or let your MCP host launch it):
   • pnpm --filter @anywaydata/mcp run start
3. Configure your MCP host/Codex to launch the server command.

Example MCP server config (published package via npx):

{
  "mcpServers": {
    "anywaydata": {
      "command": "npx",
      "args": ["-y", "@anywaydata/mcp"]
    }
  }
}

Example MCP server config (Claude Desktop style):

{
  "mcpServers": {
    "anywaydata": {
      "command": "npx",
      "args": ["-y", "@anywaydata/mcp"]
    }
  }
}

Example MCP server config (Cursor/Cline-style local repo command):

{
  "mcpServers": {
    "anywaydata-local": {
      "command": "node",
      "args": ["apps/mcp/src/index.js"],
      "cwd": "/path/to/grid-table-editor"
    }
  }
}

Windows-safe alternative (avoids npx bin resolution issues in some shells):

{
  "mcpServers": {
    "anywaydata": {
      "command": "node",
      "args": ["-e", "import('@anywaydata/mcp')"]
    }
  }
}

Example MCP server config (local repo):

{
  "mcpServers": {
    "anywaydata": {
      "command": "node",
      "args": ["apps/mcp/src/index.js"],
      "cwd": "/path/to/grid-table-editor"
    }
  }
}

For MCP hosts, prefer direct node execution or npx as shown above. pnpm run can emit non-JSON stdout lines, which may interfere with stdio MCP message parsing in some clients.

Notes:

• Transport is stdio, so PORT / --port do not apply to MCP.
• Use REST API (@anywaydata/api) for HTTP/OpenAPI/Swagger use cases.
• MCP tool name exposed is generate_data_from_spec.

Publishing to npm

Packages are configured with publishConfig.access=public.

• npm login
• pnpm install
• pnpm run verify:local

Versioning and release notes are managed with Changesets:

• Create new changeset
• pnpm changeset
• Commit the generated changeset file
• Bump version
• pnpm changeset version
• publish package
• pnpm changeset publish

Refresh npm token

• Get/create a new token from npm (or your registry UI).
• Update your user .npmrc entry (Windows):

notepad $HOME\.npmrc

• Replace the auth line, e.g.:

//registry.npmjs.org/:_authToken=YOUR_NEW_TOKEN

If you use a project-level .npmrc, update/remove token there too.

Verify:

npm whoami

npm adhoc publish

For manual one-off publish, set the version in that package’s package.json first, then publish.

Bump version:

• pnpm --filter @anywaydata/cli version patch

or minor / major / exact like 1.2.3

Publish a single workspace package manually:

• pnpm --filter @anywaydata/core publish
• pnpm --filter @anywaydata/core-ui publish
• pnpm --filter @anywaydata/cli publish

Docker API

Build images from repo root:

• docker build -f apps/api/Dockerfile -t anywaydata-api .

Run the API container:

docker run --rm -p 8082:3000 anywaydata-api

Then sent requests to:

localhost:8082

e.g.

• GET http://localhost:8082/v1/health
• GET http://localhost:8082/v1/generate/options/csv

Notes:

• docker run: starts a new container from an image.
• --rm: automatically removes the container when it stops.
• -p 8082:3000: maps host port 8082 to container port 3000.
• anywaydata-api: the image name to run.
• Connect from your machine at http://localhost:8082.
• Inside the container, the app still listens on port 3000.

Docker Web App

Build the browser app image from repo root:

• docker build -f apps/web/Dockerfile -t anywaydata-web .

Run the browser app container:

• docker run --rm -p 8080:80 anywaydata-web

Then open:

• http://localhost:8080/
• http://localhost:8080/app.html
• http://localhost:8080/generator.html
• http://localhost:8080/webmcp.html

Notes:

• This image serves the static browser app assets only.
• It does not build or include Docusaurus (docs-src pipeline).

Docker MCP

Build images from repo root:

• docker build -f apps/mcp/Dockerfile -t anywaydata-mcp .

Configure MCP with stdio for Docker

Example MCP config:

{
  "mcpServers": {
    "anywaydata-docker": {
      "command": "docker",
      "args": ["run", "--rm", "-i", "anywaydata-mcp"]
    }
  }
}

Notes:

• Build first: docker build -f apps/mcp/Dockerfile -t anywaydata-mcp .
• -i is required for stdio transport.
• No -p mapping is needed for MCP.

If you need local files available inside the container, add a bind mount, e.g. -v /host/path:/workspace.

Docker Compose

Start both services with Compose:

docker compose up --build

CI Coverage Artifacts

GitHub Actions runs linting and tests for pushes and pull requests to master.

The workflow also publishes coverage output as build artifacts on each run:

• coverage-report-node-20.x contains the full coverage/ directory
• coverage-html-report-node-20.x contains the HTML report from coverage/lcov-report/

Open the workflow run in the Actions tab and download the artifact from the run summary page.

Probable TODO:

• TODO: convert all JS to TypeScript
  • https://blog.logrocket.com/a-simple-guide-for-migrating-from-javascript-to-typescript/
• TODO: improve and expand the test data generation - written in TypeScript, don't use faker directly - create a full semantic abstraction
• TODO: create a DSL for the test data generation... probably just JSON initially
• TODO: convert UI to use React?

Useful Data Sources

Large sources of data:

• https://catalog.data.gov/ca