365Lab CSV 在线工具箱365Lab CSV Online Toolkit

What it does

Convert between CSV and JSON, NDJSON, XML, Markdown tables, SQL INSERT, HTML tables and Excel, with customisable delimiters and quotes.

How to use

1. Pick the source format, then paste content or upload a file
2. Choose the target format
3. Click "Convert" — copy the result or download it

FAQ

Why does Excel conversion download immediately?

xlsx is a binary format that cannot be previewed in a text box, so the file is downloaded as soon as it is generated.

Where did the line breaks in my Markdown table go?

Markdown table syntax cannot hold line breaks inside a cell, so they become spaces — a toast warns you when this happens.

What it does

Remove duplicate rows, trim whitespace, normalise letter case and handle empty values — plus masking and format validation for phones, emails and more.

How to use

1. Choose the scope (all columns or selected ones)
2. Tick the operations you need, and fill in mask or validation rules if required
3. Run it and export the result — every step can be undone with global undo

FAQ

Is de-duplication row-based or column-based?

Whole rows by default; in "selected columns" mode you can choose which columns define a duplicate.

What happens to rows that fail validation?

They are listed after validation and can be removed with one click — undo is available.

What it does

Rename, delete, copy and reorder columns, add calculated columns with formulas, or split one column into several by a delimiter.

How to use

1. Select the target column in the list
2. Run rename, move, delete or copy
3. For a calculated column, type a formula such as [price] * [qty] and give it a name

FAQ

What can the formula engine do?

Arithmetic plus common functions like SUM, AVG, ROUND and CONCAT. Expressions are evaluated in a sandbox — no arbitrary code runs.

Does splitting a column lose the original data?

No. The original column is kept and the split parts are appended as new columns.

What it does

Column profiling (type, missing values, distinct values, min/max, median), frequency distribution, group summaries, and one-click bar / line / pie charts.

How to use

1. Pick a column for profiling or frequency counts
2. For group summaries, choose the grouping column and the numeric column to aggregate
3. Select a chart type and the X / Y axes, then generate — charts can be exported as images

FAQ

Can it compute medians and percentiles?

The column profile for numeric columns includes median, minimum and maximum.

Can charts be exported?

Yes — generated charts can be downloaded as image files.

What it does

Combine several CSV / Excel files into one (append, side-by-side, or key-based join), or split one file by column value or row count into a ZIP.

How to use

1. Drop in two or more files
2. Choose the merge mode — columns are matched by name even if the order differs
3. Click "Merge" to preview; for splitting, switch to the "Split" tab, set the rule and download a ZIP

FAQ

What if the files have different columns?

Columns are matched by name and missing ones are filled with empty values — nothing gets misaligned.

How do I get the split files?

All parts are packaged into a single ZIP for download.

What it does

Filter rows by one or more conditions — equals, contains, greater/less than, range, empty, regex — combined with AND / OR logic.

How to use

1. Add one or more conditions, picking the column and operator
2. Enter the comparison value (range conditions need both bounds)
3. Choose the AND / OR logic and apply — the result can be exported

FAQ

Does it support regular expressions?

Yes — pick the "matches regex" operator.

Does filtering destroy the original data?

The filter produces a result from your current data and can be undone at any time.

What it does

Detect GBK, Big5, Shift_JIS, UTF-8 and more, restore garbled text, and normalise line endings.

How to use

1. Upload a file and click "Auto-detect"
2. Check the preview — if it still looks wrong, pick the source encoding manually
3. Choose the target encoding (e.g. UTF-8), convert and download

FAQ

What if auto-detection is wrong?

Pick the source encoding manually in the dropdown — the preview updates instantly.

Why does the downloaded file have a suffix?

The filename includes the target encoding (e.g. xxx_windows1252.csv) so different encoding versions do not overwrite each other.

What it does

Apply case conversion, Base64, URL encode/decode, ROT13 and JSON formatting to text or an entire column.

How to use

1. Paste text, or select the column to process
2. Choose the transformation
3. Copy the result, or write it back into the column

FAQ

Can it process a whole column at once?

Yes — pick the column and the transformation applies to every row.

Does case conversion affect Chinese characters?

No, case conversion only applies to Latin letters.

What it does

Generate random test data — names, emails, phones, addresses, order numbers — with built-in templates, configurable row counts and value/date ranges.

How to use

1. Pick a built-in template, or add columns and choose their data types
2. Set the row count plus any value or date range
3. Preview the result, then load it into the workspace to keep working

FAQ

Is the generated data safe to use in tests?

Yes — it is randomly generated on your device and contains no real personal information.

Can I export the generated data?

Load it into the workspace and export it as CSV or Excel.

What it does

Compare your current data against another CSV to find added, removed and modified rows, then export the result or load it back for editing.

"Compare options" let you fine-tune the rules: numeric tolerance (values within the tolerance count as equal — 0.01 absorbs floating-point noise), ignore case, ignore leading/trailing whitespace, and ignore columns (selected columns are excluded entirely — handy for "updated_at" or notes fields that always change).

How to use

1. Load the data that will act as the baseline
2. Upload or paste the file to compare, and choose the key column used for matching
3. Optionally set tolerance, ignore rules and ignored columns under "Compare options"
4. Run the comparison and switch between the added / removed / changed views

FAQ

Can it compare files whose columns are in a different order?

Yes — the key column is matched by name, so differences are still detected correctly.

Can I see which fields changed?

Yes — the "changed" view lists the before/after value for every differing field.

1.0 and 1.00 are reported as different — how do I fix that?

Use numeric tolerance: enter a tiny number such as 0.0001 and any pair whose difference is smaller counts as equal.

One column always changes and I want to skip it.

Tick it under "Ignore columns" — it is then excluded from both the comparison and the diff details.

I mixed up baseline and target.

Click "Swap direction"; added and removed results swap accordingly.

What it does

Cross-tabulate detail data by choosing row dimensions, column dimensions and an aggregation — count, sum, average, min or max.

"Value display" can switch the numbers to percentages: raw values, % of row total, % of column total, or % of grand total — very handy for structural analysis such as category share or regional contribution.

How to use

1. Choose the row dimensions (multiple columns allowed)
2. Choose the column dimension and the numeric column to aggregate
3. Pick the aggregation, then switch "Value display" if you want percentages
4. Generate the matrix, then export or load it back

FAQ

Can I combine several columns as a row dimension?

Yes — multi-select to combine columns into one dimension.

How are the percentages calculated?

"% of row" uses the row total as denominator, so each row sums to 100%; "% of column" uses the column total; "% of grand total" uses the overall total, so the whole table sums to 100%.

If I load a percentage view back, do I get percentages?

No — loading back always uses the raw values so percentages never pollute downstream calculations. Only export follows the current display mode.

Can I change the number of decimal places?

Currently fixed at one decimal. Export and adjust in your spreadsheet if you need another precision.

Can I keep editing the pivot result?

Yes — load it back into the workspace and treat it as normal data.

What it does

Sort by one or more columns. Each sort key has its own direction (ascending / descending), sort mode and empty-value placement, with a preview of the first rows before you apply it.

Four sort modes: Smart (numeric for numbers, locale-aware for the rest — the default choice), Natural (understands embedded numbers, so item2 comes before item10), Pinyin (Chinese ordered by pinyin), Numeric (by value, with non-numbers pushed to the end).

How to use

1. Click "Add sort key", then set column, direction, sort mode and empty-value placement for each
2. Reorder the keys with the arrow buttons — their order defines the priority
3. Preview the first rows, then apply and export

FAQ

How is priority determined with multiple keys?

Keys apply in order — the next key only breaks ties left by the previous one.

Where do empty values go?

Configured per key: first or last. The default is last.

Why does item10 come before item2?

That is plain character-by-character string comparison. Switch the key's sort mode to "Natural".

Can Chinese be sorted by pinyin?

Yes — choose the "Pinyin" sort mode. It relies on the browser's Chinese collation, which is consistent across browsers.

Can I undo a sort?

Yes, global undo (Ctrl+Z) restores the previous order.

What it does

Record a repeatable chain of operations (de-duplicate, clean, sort, drop columns, add formula columns, find & replace…) as a "recipe", save it, and replay it with one click. Ideal when you run the same transformations on similar files every week or month.

How to use

1. Work normally in Sort / Data Cleaning / Column Operations — each action is recorded automatically
2. Or add steps manually: pick an operation and fill in its parameters
3. Reorder steps with the arrow buttons, then click "Run" to replay the whole flow
4. Click "Save" to store the recipe locally — it will still be there next time you open the page

FAQ

Are steps stored by column name or index?

By column name. As long as the new file uses the same headers, the recipe still works even if column order changed.

Why isn't my filtering recorded?

Search & Filter is preview-only and does not modify data, so it is not recorded automatically. Add a filter step manually if you need one.

Are saved recipes uploaded?

No — they live in your browser's localStorage only. Clearing site data removes them, so export anything important.

Replayed the wrong recipe?

A snapshot is taken before replay, so global undo (Ctrl+Z) steps back through it.

What it does

Treat your current data as a database table and query it with SQL. Powered by SQLite (sql.js compiled to WebAssembly), so SELECT, WHERE, GROUP BY, ORDER BY, JOIN, subqueries, CASE WHEN, aggregate and window functions all work.

How to use

1. Load data first (upload a file in Parse & Preview, or use the sample data)
2. Switch to "SQL Query" — the data is registered automatically as a table named data
3. Write SQL in the editor and click "Run"; export the result as CSV or load it back as the current data

FAQ

What is the table name? How do I reference columns?

The table is always data. Column names are sanitised into valid identifiers (spaces and punctuation become underscores) — check the schema panel on the right for the exact names.

My column names are Chinese — does that work?

Yes. Wrap them in backticks or double quotes, e.g. SELECT \`姓名\` FROM data.

Is every value text?

Columns that are purely numeric become INTEGER / REAL; everything else is TEXT. Use CAST(col AS REAL) when you need numeric semantics.

Does it need an internet connection?

No. The SQLite engine is bundled with the site and loaded locally, so it works offline and your data never leaves the browser.

Can I modify data (INSERT / UPDATE)?

The syntax is supported, but results only appear in the result area and are not written back to the workspace — click "Load as current data" to keep them.

My query failed — why?

The error is shown in the result area; it is usually a column-name mismatch or a syntax issue. Compare against the schema panel.