Open Parquet Files

VisiGrid opens Apache Parquet (.parquet) files in the desktop app, the vgrid CLI, and the web app. Parquet support starts in VisiGrid 0.34.0; this page describes 0.34.2 and later.

Parquet is read-only: VisiGrid reads Parquet files and exports them to other formats, but it does not write Parquet.

Looking for an overview rather than reference? See Parquet Viewer.

Open a Parquet file

Open the file with Ctrl+O, the same as any other spreadsheet. The import runs in the background, so large files do not freeze the window.

The column names become the header row (row 1).
Each record becomes one row below it.
Top-level columns become sheet columns, in file order.

How column types come through

Parquet stores a type for every column, so VisiGrid uses it instead of guessing from text the way CSV import has to.

Parquet type	In VisiGrid
Integers	Numbers. Integers larger than ±2^53 (9,007,199,254,740,992) stay as text so they are not rounded.
Floats and doubles	Numbers. `NaN` and infinity become text.
Booleans	`TRUE` or `FALSE`.
Strings	Text, exactly as stored. `007` keeps its leading zeros, and `=1+1` stays text rather than becoming a formula.
Decimals	Numbers when the value has 15 significant digits or fewer; otherwise text, so the exact value is kept.
Dates	Dates. Dates before 1 March 1900 stay as text.
Times	Times.
Timestamps (milli, micro, nano)	Date-times, as stored in the file, with no time-zone conversion. Timestamps before 1 March 1900 stay as text.
UUIDs	Text.
Binary	Text when the bytes are valid UTF-8; otherwise hexadecimal (`0x…`).
Structs, lists, and maps	JSON text, for example `["vip","eu"]`.
Nulls	Empty cells.

Size limits

A sheet holds 1,048,576 rows (including the header row) and 16,384 columns, the same as Excel. Before 0.35.0 it held 65,536 rows and 256 columns, so files that were refused then may open now.

Desktop app: a larger file loads what fits, and a note tells you how many rows and columns were left out.
CLI: vgrid convert, vgrid sheet import, and vgrid sheet inspect --calc refuse a file that does not fit:

error: big.parquet has 2000000 rows and 1 columns; a sheet holds 1048576 rows (including the header row) and 16384 columns
hint:  split or filter the file into smaller Parquet files first

Other vgrid sheet inspect views show the rows that fit and print the truncation note on stderr, so JSON on stdout stays parseable.

A file that fits still has to fit in memory: a million rows of five columns needs roughly a gigabyte while it is open. For files past the row limit, or wide files on a small machine, filter or split them with a query engine first, then open the result in VisiGrid.

Export

Once a Parquet file is open, export it from the desktop app to CSV, TSV, JSON, or Excel (.xlsx), or save it as a VisiGrid workbook.

CSV and TSV write dates, times, and timestamps as ISO 8601 (2026-09-01 14:02:00, with milliseconds when non-zero). Before 0.34.2 they were written as spreadsheet serial numbers.

From the command line

vgrid uses the same Parquet reader as the desktop app. It needs a file path; Parquet cannot be read from stdin.

# Parquet to CSV
vgrid convert events.parquet -t csv -o events.csv

# Parquet to Excel
vgrid convert events.parquet -t xlsx -o events.xlsx

# Filter rows and choose columns on the way out
vgrid convert events.parquet -t csv --headers \
  --where 'status=paid' --select order_id,amount

# Answer a question without opening the file
vgrid sheet inspect events.parquet --headers --calc "SUM(amount)"

# Import into a VisiGrid workbook
vgrid sheet import events.parquet events.sheet --headers

See vgrid convert and vgrid sheet for every option.

In the web app

The web app at app.visigrid.app imports .parquet files up to 30 MB, with the same row and column limits.