# Query Parquet and Arrow, then save a reviewed correction

Local workflow · Windows and macOS

Use projected fields for large columnar sources and keep bounded editing separate from whole-dataset rewriting.

Updated Oct 27, 2026 · [RowCove](https://rowcove.com/)

**Can I edit a large Parquet or Arrow file in RowCove?** Viewing and querying support larger columnar sources. Current edit-copy limits are 128 MiB per source, 200,000 rows and 512 columns; do not treat a GB-scale query as proof of GB-scale editing. Open an eligible individual file to stage corrections and save a separate copy.

**In this guide** [1. Inspect the source and select useful fields](https://rowcove.com/guides/edit-parquet-arrow#inspect) [2. Query beyond the first page](https://rowcove.com/guides/edit-parquet-arrow#query) [3. Stage an eligible correction and verify the copy](https://rowcove.com/guides/edit-parquet-arrow#edit) [4. Mock measurements by actual operation](https://rowcove.com/guides/edit-parquet-arrow#mock) [5. Put a price on repeated setup and waiting](https://rowcove.com/guides/edit-parquet-arrow#value) [All practical guides ↗](https://rowcove.com/guides)

## 1. Inspect the source and select useful fields

Select a Parquet, Arrow IPC or Feather file through Open data. Review detected columns and exact values. Use Columns to show the fields needed for the question; a projected read can avoid decoding unrelated columnar values. The same record population can occupy different physical sizes in compressed Parquet and Arrow.

For a partitioned folder, use [the multi-shard guide](https://rowcove.com/guides/query-dataset-shards) . Collection browsing is read-only; an individual eligible source file has a separate edit workflow.

## 2. Query beyond the first page

Use full-file Filters to find the target category or interval, then apply a supported sort when ordering is relevant. Review the count and later result windows. A fast metadata/first-page operation is not a complete scan, and a filter-then-sort is not a sort of every source record.

The [Parquet workflow](https://rowcove.com/parquet-viewer) and [Arrow workflow](https://rowcove.com/arrow-viewer) explain format-specific access. Choose a useful scope before starting work that reads wide text or nested payloads.

## 3. Stage an eligible correction and verify the copy

For a source within the edit limits, switch to Edit, select the source cell, review the replacement and choose Stage edit. Save copy writes to a new destination. Reopen that artifact, check changed and unaffected values, and verify schema and units before using it downstream.

Match the source field type: a timestamp’s numeric unit, a decimal’s scale and a large unsigned identifier matter. Unsupported edits to complex values must be rejected rather than silently converted. File compression and batch layout can change when rewriting; save-copy is not byte-for-byte duplication.

The updated writer has regression coverage for exact unsigned integers, decimals, UTC timestamps, schema metadata and unchanged nested arrays. The native-type-preserving writer is included in RowCove 0.1.1.

Beyond the editing budget, retain query access and choose a deliberate downstream transformation workflow. Do not silently split a dataset or imply that editing one shard rewrites the entire collection.

## 4. Mock measurements by actual operation

All examples and measurements below use generated mock data. No customer files or trading records are used.

Same 1,048,576 logical records; three-trial medians

| Operation | Parquet · 640.7 MB Zstd | Arrow IPC · 1,106.8 MB |
| --- | --- | --- |
| Open and read first 100, all fields | 2.985 ms | 3.374 ms |
| Open and read first 100, projected id/group | 1.546 ms | 2.836 ms |
| Open and return all 10,810 matching records | 2.060 s | 1.159 s |
| Open, filter and numerically sort the matches | 2.092 s | 1.189 s |
| Open and read first/middle/final projected windows | 31.287 ms | 6.868 ms |

Measured on Windows with a Release native engine build on 27 October 2026: three fresh-process trials per task, the same 1,048,576 logical mock records, and a 1,024-byte synthetic payload. Median elapsed times include opening, index/schema preparation and the requested work. UI rendering, IPC and the separate correctness pass are excluded. OS caches were not cleared; these are not cold-disk guarantees or a comparison with another application. These engineering measurements cover the updated engine included in 0.1.1; results depend on the device, source and requested operation.

All matching records, exact IDs and order were verified. Native query/sort peak working set was at most 31.4 MiB for Parquet and 28.5 MiB for Arrow; total desktop memory is not included.

A separate 100,000-record mock edit fixture used 256-byte payloads: complete one-label edit-copy medians were 148.649 ms for a 14.5 MB Parquet source and 35.551 ms for a 28.8 MB Arrow source. Every saved record and native schema was checked. These bounded edit results cannot be extrapolated to the larger query fixtures.

## 5. Put a price on repeated setup and waiting

Pro is useful when locating and correcting records is repeated work: fewer one-off conversion scripts, less manual file splitting and a reviewed independent output. Free Preview opens up to 100 records read-only. Try representative files first; occasional small-file inspection may already be covered by free tools. See [introductory prices and renewal terms](https://rowcove.com/pricing) : first month US$5, then US$10/month; first year US$39, then US$100/year, for eligible new subscribers.

Time saved depends on the tasks you repeat and your existing tools. These timings do not establish purchase intent, paid conversion or a universal speed ranking.

## Continue your workflow

- [Parquet viewing and querying](https://rowcove.com/parquet-viewer)

- [Arrow IPC and Feather workflows](https://rowcove.com/arrow-viewer)

- [Browse a dataset collection](https://rowcove.com/guides/query-dataset-shards)

## Try the first page on your own data.

Start on Windows or macOS: source information and up to 100 records, read-only. Review reader capabilities before planning a full-file Pro workflow.

[Download for Windows ↓](https://rowcove.com/download/windows) [Download for macOS ↓](https://rowcove.com/download/macos)

Version 0.1.1 · Windows x64 · macOS Apple Silicon · [Install guide →](https://rowcove.com/download)

[Reader capabilities →](https://rowcove.com/formats)

---

Canonical page: https://rowcove.com/guides/edit-parquet-arrow

Updated: 2026-10-27

[Related Agent reference](https://rowcove.com/agent/workflows.md)