DATA & CONVERSION · VERSION 1.0.0

Data Cleaning & Conversion Workbench

Turn messy files into useful tables. Clean columns, combine datasets, summarize results and export with a reusable recipe—all in your browser.

Start with the example or import your first file.

Quick start

Ready workflows still show a preview before changing your table.

Clean data with a preview

Your working table

No data loaded

Your table will appear here. Start with the example to explore the full workflow.

Page 1 of 1

Activity & import notes

  • Changes are applied only after you review them.

Using Data Cleaning & Conversion Workbench

Clean and combine CSV, Excel, JSON and XML tables locally. Preview transformations, join datasets, pivot, validate, edit cells, save projects and reusable recipes, and export ten formats.

  1. Import files, choose spreadsheet sheets or try the messy sales example. Check the import notes and column profiles.
  2. Choose a cleaning, column, join or summary action. Preview its result, compare it with the original and apply when it is correct.
  3. Validate the table, save your project or recipe, and export the complete working table or rows matching your search.

Example

The example contains padded customer IDs, duplicated sales records and a companion customer list.
Trim the customer IDs, remove exact duplicates, then left-join Customers to add regions before grouping sales totals.

Questions & answers

What can I import and export?

Import CSV, TSV, JSON record arrays, JSON Lines, flat XML records, XLSX, XLS and ODS. Export CSV, TSV, XLSX, ODS, JSON, JSON Lines, XML, HTML, Markdown and SQLite-compatible SQL. Workbooks are imported and exported as text values only.

How can I undo changes?

Every operation has a preview before Apply. Undo and redo retain up to eight recent states per dataset, subject to a shared text budget. Original imports are retained. Save a project to keep data and steps across visits.

Can I join two spreadsheets?

Import both tables and choose Combine tables. Match a column from each dataset and choose left, inner, full or unmatched-only joins. Blank keys never match; duplicated companion keys require an explicit policy. Append can stack matching columns or form a union.

Does the tool preserve formulas and numeric precision?

It does not execute or recalculate formulas. Spreadsheet cached values may be stale or missing. JSON numeric digit text is preserved on import, and all exported JSON/workbook values are text. Explicit numeric transformations and summaries use JavaScript floating point and are unsuitable for exact financial calculations.

How do recipes and projects differ?

Recipes contain transformation steps and column/dataset names, without the source tables. Projects contain originals, working tables, steps, validation rules and import notes. Replay starts from the active dataset’s original; companion datasets must be loaded by the recorded name.

Are files uploaded and what are the limits?

Processing runs locally with existing browser libraries. Public pages include third-party ads described in the privacy policy. Use up to eight datasets, 10,000 rows and 100 columns per table, within workspace cell/text bounds. Workers stop after 30 seconds. Large files may hit browser memory limits sooner.

KEEP YOUR WORKFLOW MOVING

A useful next step for your task.

Browse Developer & Data

A complete workflow, without spreadsheet guesswork

1. Bring your files together

Import delimited text, JSON records or spreadsheet sheets. Inspect the original alongside the working table. Missing JSON values become empty cells; nested values remain JSON text.

2. Preview, then apply

Trim text, fix casing, remove duplicate records, split columns or match a second dataset. Review row counts and warnings before applying. Your originals stay available.

3. Reuse your work

Download a recipe to repeat cleaning steps on another import, or save a project with your datasets, originals and steps. Export the complete table or just matching search rows.

Practical limits: eight datasets; 10,000 rows, 100 columns and 300,000 cells per table. The combined originals and working copies support up to 800,000 cells and 12 million text characters. Worker operations stop after 30 seconds and can be cancelled. Large files may exceed browser memory before these limits.

Know what changes: this is a text-based data workspace. It does not recalculate spreadsheet formulas, preserve workbook design, infer financial decimal types or edit arbitrary nested XML. XML imports support flat record elements with scalar child elements, or this tool’s records/record/field format. DTDs, namespaces and nested fields are rejected. Numeric transformations and summaries use floating point, so keep identifiers and high-precision amounts as text.

Processing uses locally hosted libraries in your browser. The public website includes third-party advertising; see our privacy information. Projects contain your actual data—store and share them carefully.