The Data workspace opens .csv, .tsv and .tab. The file you open stays the document throughout the session: the grid displays its fields without copying them elsewhere.
Open a data file
Balise detects how the file is written and shows its findings in the inspector: delimiter, quoting character, line endings, encoding and any BOM.
If detection is wrong, correct it yourself in the inspector. This happens most often with files exported from a spreadsheet in French, which often use semicolons.
A cell starts as text. A postal code 01234 keeps its zero, and a reference 3-14 does not become a date.
Edit in the grid
The grid behaves as you expect from a table: cell or range selection using A1 coordinates, keyboard navigation, selection extension, page-by-page movement, and column and row resizing.
A value bar shows and edits the active cell’s exact content without reformatting it.
You can copy and paste whole ranges, insert or delete rows and columns, and zoom from 25% to 400%.
Edit > Undo (⌘Z) reverses each transformation.
Understand a column
The left navigation profiles each column: header, letter and probable type — empty, text, integer, decimal, Boolean, ISO date or mixed.
The profile also counts empty and distinct values.
These types are an interpretation rather than a decision. Balise converts no field merely because it guessed a type.
Clean up
The Data menu applies explicit source transformations:
- trim spaces at the start and end of a column;
- convert a column to uppercase or lowercase;
- normalize a column to Unicode NFC;
- deduplicate rows by the active column;
- split a column at the first comma;
- merge selected columns with a space;
- insert or delete a row or column.
Each can be undone.
Diagnostics
The inspector separates errors, warnings and observations. When a diagnostic has a position, a button takes you directly to the affected row.
Balise reports ambiguous delimiters, mixed line endings, unclosed quoted fields, unexpected quotes, content after a closing quote, inconsistent field counts between rows, control characters, empty or duplicate headers, spaces around headers, uncertain encoding, unnormalized text, empty rows, mixed-type columns, numbers with leading zeros, constant or duplicate columns, and duplicate rows.
Repair
Depending on its findings, Balise may propose closing the final unclosed field, unifying line endings or padding short rows with empty fields.
A repair targets a copy by default.
Before writing, a preview shows the before and after formats, content differences, their reasons, severity and total change count. Closing the preview changes neither the session, the original file nor the disk.
Convert
Convert a copy produces a file for another tool using the profile you choose. Available profiles and uses are described in Formats and conversions.
As with repairs, conversion starts with a preview, and the final write checks that the source has not changed meanwhile.
Export a report
Export a complete inspection report as Markdown to read and version with the project, or as JSON for another tool.
It contains file dimensions, dialect, encoding, diagnostics, proposed repairs and column profiles.
Very large files
Above a certain size, Balise opens a partial preview instead of the whole file. The interface says so and switches to read-only mode.
You can read and explore but cannot save over the original: export a segment or copy. This prevents implying that a change affects an entire file when only part is loaded.
Display preferences
Column widths and frozen columns do not belong to your data. Balise can store them in a small companion file beside the CSV or elsewhere, as chosen in Settings.
Your CSV stays complete and readable without it.
What this surface does not do
- Balise does not recalculate chains of formulas: it is not a calculation spreadsheet.
- Balise never converts a field type without your request.
- Balise applies no repair without showing a preview first.
- Balise does not write to the original file when you repair or convert.
For problems on opening, see When a data file causes problems.
Follow this walkthrough from start to finish: Clean up a data file you received.