loremfile

Sample Arrow IPC file for testing data pipelines, free and CC0

This is a sample Apache Arrow file: the shared synthetic people dataset in Arrow's IPC file format, free to hotlink or download.

https://loremfile.dev/arrow/people-1000.arrow · 260.7 KB (260,674 bytes) · application/vnd.apache.arrow.file · 1,000 rows · 14 columns

The IPC file format stores columnar data the way Arrow holds it in memory, so a reader can map the file and use it without parsing. arrow/people-1000.arrow holds the same records as the CSV, JSON, Parquet and Avro people files, in a single record batch.

Two details matter when testing. First, this is the IPC file format, with a footer, not the streaming format; the two share most of their bytes but are opened by different APIs, and a reader pointed at the wrong one fails. Second, because the rows match the other people fixtures, you can load the same data from several formats and compare the results column by column.

Use it for data pipeline tests, schema inspection, and checks that a service accepting Arrow data handles a file where it might expect a stream. Related formats: Parquet and Avro.

5 files. The same list as JSON: https://loremfile.dev/arrow/index.json

Arrow files
FileSizePropertiesDescription
all-types.arrow 4.8 KB (4,810 bytes) 20 rows Twenty rows in nine columns of types readers often mishandle: list, struct, map, dictionary-encoded string, decimal, timestamp with time zone, date, time and binary, with four nulls in every column.
Snippets 
<a href="https://loremfile.dev/arrow/all-types.arrow">download</a>
curl -O https://loremfile.dev/arrow/all-types.arrow
people-1000-10-batches.arrow 273 KB (272,970 bytes) 1,000 rows The same 1,000 people rows as an IPC file split into ten record batches, for readers that stop after the first batch.
Snippets 
<a href="https://loremfile.dev/arrow/people-1000-10-batches.arrow">download</a>
curl -O https://loremfile.dev/arrow/people-1000-10-batches.arrow
people-1000-zstd.arrow 96.7 KB (96,722 bytes) 1,000 rows The same 1,000 people rows as an IPC file whose buffers are zstd-compressed, which a reader built without that codec cannot open.
Snippets 
<a href="https://loremfile.dev/arrow/people-1000-zstd.arrow">download</a>
curl -O https://loremfile.dev/arrow/people-1000-zstd.arrow
people-1000.arrow 260.7 KB (260,674 bytes) 1,000 rows 1,000 rows of the shared people dataset in the Arrow IPC file format (not the stream format), one record batch.
Snippets 
<a href="https://loremfile.dev/arrow/people-1000.arrow">download</a>
curl -O https://loremfile.dev/arrow/people-1000.arrow
people-1000.arrows 259.8 KB (259,816 bytes) 1,000 rows The same 1,000 people rows in the Arrow IPC stream format, with no footer and no ARROW1 magic, opened by a different API from the file format.
Snippets 
<a href="https://loremfile.dev/arrow/people-1000.arrows">download</a>
curl -O https://loremfile.dev/arrow/people-1000.arrows

Every Arrow file in one command, each checked against its hash: npx loremfile get --format arrow, or loremfile get --format arrow after pip install loremfile.

File names follow the naming grammar. Verify a download against sha256sums.txt.