Uploading a dataset

A dataset always lives inside a project, so you will need a project to upload into — create one first if you have not already. You will also need a dataset saved as a .csv file whose first four rows describe each column.

If you are preparing the file from scratch, read the dataset file format first. It explains the four reserved header rows, the units a column may declare, and the rules each data value has to satisfy. This page assumes you already have a file in hand and focuses on getting it uploaded.

Datasets are uploaded into a project. To add one:

1. Open the project you want to add the dataset to.
2. Click the + Upload Dataset button.
3. Choose a .csv file from your computer.
4. Press Submit to start the upload.

The file is read and checked in the browser as soon as you submit it, so you usually get a result within a moment or two — there is no long server round-trip to wait on for the validation itself.

Before the contents are inspected, the file itself must:

• use the .csv format,
• not be empty, and
• be no larger than 10 MB.

These are the cheapest checks to fail, so they run first. A file that is the wrong type, has no content, or is over the size limit is rejected before the toolkit looks at any rows or columns.

Once the basic requirements pass, the toolkit reads the four reserved header rows and then every data row beneath them. Each column is validated against the type and unit it declares, and each value is checked against the column it belongs to.

If anything is wrong, the upload status panel describes the problem and points to the exact row and column so you can find it quickly. The original file on your computer is never changed; fix the issue there, then upload again.

When validation passes, the dataset is added to the project and appears in its list of datasets. From there you can open it to filter, bookmark, and compare candidates in the parallel-coordinates view.

A dataset is a snapshot of the file you uploaded. To revise the data, edit your .csv and upload it again as a new dataset rather than editing values in place.

Ready to try it? Head to your projects and upload a dataset, or explore the tool with the sample data first.