Exporting and Converting

← Contents

Outside edit mode, TableSkim never changes the files it opens, but it can write what you see to a new file. The view is exported with any unsaved edits.

Feature	Source	Output	What is written
Export View	Any sheet, including SQL results	CSV, TSV, Excel, Parquet	The sheet as shown: sorted, filtered, without hidden columns
Convert to Parquet / DuckDB	CSV, TSV, VCF and other text files	Parquet, DuckDB	The whole file, in file order

Export View

GUI: File → Export View... (Ctrl+Shift+E).
Terminal UI: w.
Web app: Export; choose the format in the dialog, and the browser saves the file into its downloads.

Choose a file name; its extension selects the format:

Extension	Format
`.csv`	Comma-separated text, UTF-8
`.tsv`, `.tab`	Tab-separated text, UTF-8
`.xlsx`	Excel workbook
`.parquet`	Parquet

If the name has none of these extensions, the GUI adds the extension of the type chosen in the file dialog, and the terminal UI refuses it with the message "Choose a name ending in .csv, .tsv, .xlsx or .parquet".

The suggested name is the source file name with -export added, for example sales-export.csv; for a file with several sheets the sheet name is included, as in book-Sheet1-export.xlsx.

What is written:

Only the rows that pass the filter, in the current sort order.
All columns except hidden ones.
CSV / TSV: each cell as it is shown, quoted only where needed, with a header row if the sheet has column names.
Excel / Parquet: numbers, booleans, dates and times keep their types. A column of text (for example from a CSV file) gets the narrowest type that fits every value, so a column of numeric text becomes a numeric column and a column of ISO dates becomes a date column. Parquet columns are named after the header, or A, B, C, ... without one.

If a column mixes integers and decimals, it stays text when converting to floating point would change an integer's value. This also applies to Convert to Parquet / DuckDB.

The export runs in the background with a progress display; Cancel (GUI) or Esc (TUI) stops it. The target file is replaced only when it has been written completely, so a cancelled export leaves an existing file untouched. If the sheet's sort or filter changes during the export, the export fails and can be started again.

Convert to Parquet / DuckDB

Text files have no index, so every sort or filter of a large CSV file reads all its rows into memory. Converting it once to Parquet or DuckDB makes later work fast: those formats are sorted and filtered inside DuckDB, and they can be queried with SQL.

GUI: File → Convert to Parquet / DuckDB.... Choose the target; the extension (.parquet, or .duckdb / .ddb) selects the format. The converted file is opened when the conversion is done.
Terminal UI: P (Parquet) or D (DuckDB) writes the file next to the original, for example data.csv.gz → data.parquet or data.duckdb, and opens it.
Web app: Convert; choose Parquet or DuckDB, and the browser saves the file, named after the source, into its downloads.

What is written:

The whole file in file order: the sort and filter of the view are not applied (use Export View for that).
The current header, ## line and text encoding choices are used, so the result has the same rows and columns as shown.
Each column gets the narrowest type every value fits: integer, decimal, boolean, date, timestamp, or text. Empty cells become NULL in typed columns.
A DuckDB file holds one table named after the source file.

As for exports, the conversion can be cancelled, and the target file is replaced only when it is complete. The source file itself cannot be chosen as the target.