How to Import a Parquet File into Google Sheets

Why you can't just upload it

Google Sheets imports CSV, TSV, XLSX and ODS, but not Parquet. If you upload a .parquet file to Drive and try to open it, Drive offers no Sheets option. So the route is the same as for Excel: Parquet → CSV → Sheets. The interesting part is getting a CSV that Sheets can actually hold and won't mangle.

Step 1: check the size against the cell limit

A Google spreadsheet holds at most 10 million cells, counted across every tab. That ceiling is rows × columns, not rows alone: a 40-column file runs out at 250,000 rows.

Drop the file into the SQL Workbench and check:

SELECT count(*) AS rows FROM 'orders.parquet';
DESCRIBE SELECT * FROM 'orders.parquet';

Multiply rows by columns. If the result is anywhere near 10 million, cut the data down before exporting. Even well under the limit, Sheets slows down noticeably once you get into the millions of cells.

Step 2: shape the export with SQL

Pick only the columns the sheet needs, filter the rows, and turn risky types into safe text in the same query:

SELECT
  CAST(order_id AS VARCHAR)            AS order_id,   -- 19-digit IDs survive
  customer_code,                                      -- already text: '00123'
  strftime(created_at, '%Y-%m-%d %H:%M:%S') AS created_at,
  region,
  amount
FROM 'orders.parquet'
WHERE created_at >= TIMESTAMP '2026-09-01';

Each cast does a specific job:

• Long integer IDs. Sheets keeps about 15 significant digits, so an INT64 ID like 1234567890123456789 comes back rounded. Exporting it as text keeps it exact.
• Timestamps. An unambiguous YYYY-MM-DD HH:MM:SS string parses as a date in every locale. Convert to the time zone your audience expects before formatting; Parquet often stores UTC.
• Nested columns. Structs and lists come out as long text blobs. Flatten them first; see flattening nested columns.

If a summary is what the sheet is really for, aggregate instead. A GROUP BY region, month gives you a few hundred rows, and that is a far better spreadsheet than two million raw ones.

Run the query and download the result as CSV. For a whole file with no reshaping, the Parquet to CSV converter does the same thing in one drop.

Step 3: import with the right options

In Google Sheets, open File → Import → Upload and pick the CSV. The dialog asks two things:

1. Import location. Create a new spreadsheet, insert a new sheet, replace the current sheet, or append to it. Appending is useful for adding a monthly extract to an existing tab, but it counts toward the same 10-million-cell ceiling.
2. Convert text to numbers, dates, and formulas. With this ticked, Sheets strips leading zeros from codes like 00123, rounds long IDs and treats any value starting with = as a formula. Untick it when your data has codes or untrusted text, and then format the date and amount columns yourself.

Leave the separator on Detect automatically unless you exported with a non-comma delimiter.

Check what arrived

Compare the row count in the sheet (the last row number minus the header) with the count(*) from step 1. Then spot-check three things: an ID column for rounding, a code column for lost leading zeros, and a date column for day/month swaps. Fixing these now is much cheaper than after someone has built a pivot table on top.

When Sheets is the wrong tool

If the filtered data still won't fit, Sheets isn't the right destination. Keep the analysis in SQL and export only the results, or use Excel, which fits 1,048,576 rows per sheet (see converting Parquet to Excel). Teams already on BigQuery can query large tables from Sheets through Connected Sheets without importing the raw rows at all.