Dump a NumPy array into a csv file

python

dataframe

pandas

csv

byNikita Barsukov·Jan 17, 2025

To swiftly export a NumPy array to CSV, make use of the numpy.savetxt function:

import numpy as np

# Let's pretend 'array' is your NumPy array
np.savetxt("output.csv", array, delimiter=",")

This will take care of the whole conversion process. By default, it applies , as the CSV delimiter. Need more control over your output format? Tweak the fmt parameter to tailor the data formatting.

Saving NumPy arrays: A deep dive

Numeric data formatting 101

Formatting numeric data is a key step you should not overlook, especially for later data analysis. To do this, use the fmt parameter in numpy.savetxt to specify the precision. Let's say %10.5f for floating numbers limited to up to 5 decimals.

Header management and strings: Avoid the pitfalls

If your CSV file requires headers, you simply need to include the header parameter in numpy.savetxt. What if your data includes string data with commas? Avoid breaking your CSV structure by wrapping them in quotes.

Dealing with large arrays? Compression is your best friend

Save your disk space when working with large NumPy arrays by using .gz compression. Just add .gz to your filename in numpy.savetxt. Spaceship not included! 🚀

Common mistakes: Your guide to staying out of trouble

Before hitting that 'run' button, beware of potential data loss or formatting issues. An example is using numpy.ndarray.tofile, which might result in a botched CSV format, as well as losing out crucial endianness and precision info.

Ascending to CSV mastery: Advanced use cases

Giving pandas a shot

Thanks to its flexibility, DataFrame.to_csv from pandas could be your strong ally, especially if you need more sophisticated export capabilities like omitting headers or indexes.

Loading CSV data back into NumPy arrays

When you need to do the reverse, np.recfromcsv and numpy.genfromtxt are great for loading CSV data back into NumPy arrays. These guys can handle various data types and missing values, like a boss.

Working with record arrays

If you're dealing with record arrays, np.dtype.names can help you extract headers from your NumPy array, perfectly suited for your CSV headers for better readability.

Handling unicorns and encodings

Okay, maybe not unicorns, but dealing with unicode or special characters in CSVs is equally said to be of mythical complexity. Always specify the right encoding when loading and saving data. Unicorns not guaranteed. 🦄

explain-codes / Python / Dump a NumPy array into a csv file

Linked

How do you read a file into a list in Python?



Pretty-print a NumPy array without scientific notation and with given precision



_csv.error: field larger than field limit (131072)



How do I convert all of the items in a list to floats?



Initialising an array of fixed size in Python

