Convert NPZ to PARQUET

Turn NPZ files into PARQUET files. One Parquet table for each array in the archive, named after it, with a typed column for each column the CSV conversion makes.

Converted in an isolated sandbox on our serversDeleted after 24 hours

Drop NPZ files on the dialor turn it to pick what to do

One Parquet table for each array.Each array in the archive becomes its own table, named after the array.

  1. What you get

    One Parquet table for each array in the NPZ archive, named after that array. Each table has a typed column for every column the CSV conversion makes.

    A 1-D array becomes one column. A 2-D array becomes a table with a column for each index of its second axis, named 0, 1, 2 and so on.

  2. Records and larger arrays

    A record array gets a column for each field. A field that holds an array or nested records is written as JSON text.

    An array of three or more dimensions becomes one row for each index of all but its last axis. Those rows are led by columns named axis0, axis1 and so on.

  3. What changes or fails

    NaN and NaT become NULL in the table.

    Arrays of Python objects are not loaded, because unpickling them could run code. The conversion stops with unsupported_elements when the archive holds one.

NPZ and PARQUET.What each format is.

The formats

From · NPZ

NumPy Archive

Several NumPy arrays zipped together by numpy.savez, each under its own name.

Extensions
.npz
MIME type
application/x-npz
Kind
Text and data
To · PARQUET

Parquet

The columnar data format of analytics. Compressed, typed and fast to scan.

Extensions
.parquet .pq
MIME type
application/vnd.apache.parquet
Kind
Text and data

In your code.The same NPZ to PARQUET conversion through the API, for your app or your agent.

terminal
$ cv convert report.npz report/ --to parquet
$ cv convert report.npz report/ --to parquet

Questions.What people ask about NPZ to PARQUET.

01OF 04

One or more. Each array in the archive gets its own Parquet table, named after it.

Each index of the second axis becomes a column, named 0, 1, 2 and so on.

NaN and NaT become NULL in the Parquet table.

The archive holds an array of Python objects. These are not loaded, since unpickling them could run code.

More conversions.Other ways in and out of NPZ and PARQUET.