Tool Launcher

Search for a tool...

1tt.dev1tt.dev
All guides

CSV Viewer with Dataset Splitting

Open, filter, sort, and edit CSV files in the browser — then split them into train/test sets for machine learning with one click.

A CSV viewer built for data work

The 1tt.dev CSV Viewer is more than a spreadsheet preview. It's a lightweight data tool that lets you open CSV files of any size, filter and sort rows, toggle column visibility, edit cells inline, and export subsets of your data — all without leaving the browser. Nothing is uploaded to a server; everything runs client-side.

If you work with datasets for machine learning, analytics, or data engineering, the built-in dataset splitting feature is especially useful. You can split a CSV into train/test sets with a single click using common ratios like 80/20, 70/30, or any custom split you need.

Opening and exploring CSV files

Drag and drop a .csv file onto the viewer, or click to browse. The parser auto-detects the delimiter — commas, tabs, semicolons, and pipes are all supported.

Once loaded, you get a fast, paginated table (100 rows per page) with:

  • Column sorting — click any header to sort ascending or descending.
  • Search / filter — type in the search bar to filter rows across all columns in real time.
  • Column visibility — toggle individual columns on/off to focus on the fields that matter. Hidden columns are excluded from exports.
  • Row count and stats — the toolbar shows total rows, filtered rows, and column count at a glance.

Editing data in place

Click the pencil icon on any row to open the row editor. You can modify cell values directly and save changes back to the in-memory dataset. You can also add new rows — useful for appending labels, annotations, or test cases before exporting.

All edits stay in the browser. The original file is never modified. When you're done, export the edited data as a new CSV.

Exporting subsets

The export menu gives you precise control over what gets downloaded:

  • All rows — export the full dataset (respecting column visibility).
  • Current view — export only the rows matching your current filter/search.
  • First N / Last N rows — grab the top or bottom slice of your data.
  • Random sample — export a random percentage of rows (e.g., 10% sample for quick analysis).
  • Row range — export rows between specific indices (e.g., rows 500–1000).

Dataset splitting for machine learning

This is the feature that sets the CSV Viewer apart from generic spreadsheet tools. When preparing data for machine learning, you typically need to split your dataset into training and testing subsets. The viewer does this with built-in presets:

  • 80 / 20 split — the most common ratio. 80% for training, 20% for testing.
  • 70 / 30 split — a more conservative split with a larger test set.
  • 90 / 10 split — for large datasets where you want maximum training data.
  • 60 / 40 split — when you need a substantial validation set.

When you select a split preset, the viewer shuffles the data randomly before splitting to avoid any ordering bias. It then downloads two separate CSV files — for example, mydata_train80.csv and mydata_test20.csv.

The shuffle uses a Fisher-Yates algorithm on the entire dataset before splitting. This means every row has an equal chance of ending up in either subset, regardless of the original row order.

Column visibility is respected during splits — if you've hidden columns (like an ID column or irrelevant metadata), they won't appear in the exported train/test files. This lets you do feature selection visually before splitting.

Typical workflow

Here's how a data scientist or ML engineer might use the CSV Viewer in practice:

  1. Load the dataset — drag and drop your CSV file.
  2. Inspect the data — sort columns, search for anomalies, check row counts.
  3. Clean up — hide irrelevant columns (IDs, timestamps, debug fields), edit incorrect values.
  4. Filter if needed — use search to focus on a subset (e.g., only rows where status=active), then export the filtered view.
  5. Split for ML — pick an 80/20 or 70/30 preset, download the train and test files.
  6. Feed into your pipeline — use the exported CSVs directly in pandas, scikit-learn, PyTorch, or any ML framework.

No upload, no limits

Everything happens in the browser using client-side JavaScript. Your data never leaves your machine — there's no upload, no server processing, no file size limits beyond your browser's memory. This makes it safe for sensitive or proprietary datasets.

The CSV Viewer is free and requires no account. Open it at /tools/csv and start working with your data immediately.