Skip to content

numpy/file-roundtrip

Rule file-roundtrip
Pack NumPy
Severity warning
Fix none, edit by hand
Library versions every version

Writing an array as text with np.savetxt and parsing it back is slow and loses the dtype and the shape. Use np.save and np.load, which were hundreds of times faster.

np.savetxt("cache.txt", data)
data = np.loadtxt("cache.txt")
np.save("cache.npy", data)
data = np.load("cache.npy")

The rule reports np.loadtxt(f) and np.genfromtxt(f) in a file that also calls np.savetxt(f, ...), and np.fromfile(f) in a file that also calls x.tofile(f), with the same file argument. np.save and np.load keep the dtype and the shape.

xpd has no automatic fix, so make the change by hand.

xpd prints this message with the issue. A name that starts with $ stands for code from your file.

This program writes $FILE as text with np.savetxt and parses it back with $NP.$LOAD. Write it with np.save and read it with np.load, which ran 27 to 880 times faster reading and 420 to 510 times faster writing 1,000,000 values, and keeps the dtype and shape.

Browse the other rules of the NumPy pack.