AttributeError: ‘NumPy NdArray’ Object Has No Attribute ‘Insert’ | Use np.insert, Not .insert

Guide / By

Fix this NumPy error by calling numpy.insert(arr, obj, values) instead of ndarray.insert, which does not exist on arrays.

Seeing AttributeError: ‘NumPy NdArray’ Object Has No Attribute ‘Insert’ tells you one thing: you tried to call a list-style method on a NumPy array. NumPy arrays don’t ship with an insert method on the object. The insert operation lives as a function in the NumPy namespace: numpy.insert. That function returns a new array with your values placed at the given index or slice.

Why The Error Happens

Python lists have an insert() method. You can write a_list.insert(0, x) and it places the item before index 0. Many developers carry that habit over to arrays and write a_array.insert(…). Lists and arrays look similar, but their APIs differ. Lists support insert as an in-place method; arrays do not.

  • Know the container — Lists use list.insert(i, x) as an in-place change. Arrays use numpy.insert(arr, obj, values, axis=None) and return a copy.
  • Watch the import — Call np.insert(...) after import numpy as np. Don’t attempt some_array.insert(...).
  • Expect a new array — numpy.insert does not modify the input in place; it creates a new array with the inserted values. Assign it back.

Quick Fix — Use numpy.insert Correctly

Basic pattern:

import numpy as np

arr = np.array([10, 20, 30])
arr = np.insert(arr, 1, 15)     # insert 15 before index 1
# arr => [10, 15, 20, 30]

Multiple values:

arr2 = np.array([1, 4, 7])
arr2 = np.insert(arr2, [1, 2], [2, 3])  # place 2 before idx 1, 3 before idx 2
# arr2 => [1, 2, 4, 3, 7]

Along an axis:

m = np.array([[1, 2, 3],
              [4, 5, 6]])
m = np.insert(m, 1, [9, 9], axis=1)     # insert a column before column index 1
# m => [[1, 9, 2, 3],
#       [4, 9, 5, 6]]

The parameters let you target a single index, a slice, or even a boolean mask. Behavior details live in the official docs.

Fix “Numpy Ndarray Has No Insert” — The Right Call

Most projects need a small change: move from a method call on the array to the top-level function. Here’s a map of common missteps and the matching fix:

  • If you wrote arr.insert(0, x) — use arr = np.insert(arr, 0, x).
  • If you need a row — pick axis=0 with a row-shaped sequence.
  • If you need a column — pick axis=1 with a column-length sequence.
  • If you had a boolean index — recent NumPy treats it as a mask of insertion points; read the version notes.

AttributeError: ‘NumPy NdArray’ Object Has No Attribute ‘Insert’ In Context

You’ll see the exact message AttributeError: ‘NumPy NdArray’ Object Has No Attribute ‘Insert’ when calling a non-existent attribute on the array. In NumPy, the array type centers on fast math and fixed-shape data, so structural add/remove often happens through functions that return a new array. That split keeps the core object lean. The reference pages describe ndarray and its behavior.

In many codebases, this error appears during a port from plain lists or from pandas. Lists allow .insert; pandas DataFrame also exposes .insert for columns. Arrays do not. Mixing these APIs triggers the exception.

Pick The Right Tool: List vs NumPy vs Pandas

Use the structure that matches your goal. This compact table shows where insert lives and how it behaves.

Structure How To Insert In-Place Or Copy
Python list a.insert(i, x) In-place on the list.
NumPy array np.insert(arr, obj, values, axis) Returns a new array (assign back).
pandas DataFrame df.insert(loc, column, value) Alters the frame in place.

Practical Patterns That Avoid The Error

When performance and clarity both matter, lean on patterns that keep inserts predictable.

  • Build once, fill later — Pre-allocate the full output shape, then write to slices or ranges. This sidesteps repeated copies from many inserts.
  • Stage in a list — Collect values in a list with append/insert, then convert once with np.array.
  • Use concatenate for joins — For large blocks, np.concatenate often reads cleaner than many small inserts.
  • Prefer vector moves — Slice assignment is fast and clear for shift/expand style tasks.

Common Scenarios And Drop-In Fixes

Insert Into A 1D Array

Goal: place a single value before a position.

arr = np.array([2, 4, 6, 8])
arr = np.insert(arr, 2, 5)    # before index 2
# [2, 4, 5, 6, 8]

Insert Multiple Values

Goal: insert more than one item at set positions.

arr = np.array([0, 3, 6])
arr = np.insert(arr, [1, 2], [1, 2])
# [0, 1, 3, 2, 6]

Insert A Column Into A 2D Array

Goal: add a column before an index.

m = np.array([[1, 2],
              [3, 4],
              [5, 6]])
col = np.array([9, 9, 9])
m = np.insert(m, 1, col, axis=1)
# [[1, 9, 2],
#  [3, 9, 4],
#  [5, 9, 6]]

These calls match the documented numpy.insert contract. Check the object and axis rules in the manual for corner cases such as slices and boolean masks.

Performance Notes

Every numpy.insert call allocates a new array. When you loop and insert over and over, the copies pile up. That often slows code more than a single build step. If you must add values repeatedly, collect them in a list, then convert once, or pre-size the array and assign to slices. The array docs outline how arrays store items contiguously, which explains the copy cost on structural changes.

  • Switch the data shape — For frequent front inserts, build in reverse and reverse once at the end.
  • Batch the work — Use one np.insert with an array of positions rather than many single inserts.

When You’re In Pandas, Stay In Pandas

If your object is a DataFrame, use df.insert to add a column at a specific location. That method updates the frame in place and raises a clear error if the name already exists unless you allow duplicates. Trying to funnel a DataFrame column insert through numpy.insert adds extra conversions for no gain.

import pandas as pd

df = pd.DataFrame({"a": [1, 2], "c": [3, 4]})
df.insert(1, "b", [9, 9])   # column "b" between "a" and "c"
# df:
#    a  b  c
# 0  1  9  3
# 1  2  9  4

List Habits That Trigger The Error

Shifting from lists to arrays often keeps the same variable names, which hides the type switch. A quick scan helps catch it:

  • Check type — Print type(obj) or rely on your IDE hint. If it’s numpy.ndarray, don’t call .insert.
  • Match the API — Use list.insert only on lists per the Python docs.
  • Keep namespaces clear — Avoid naming a variable numpy or np. That can shadow imports and break calls.

Troubleshooting Checklist

  • Confirm the object — Is it a list, an array, or a DataFrame?
  • Use the right call — list.insert for lists; np.insert for arrays; df.insert for DataFrames.
  • Assign the result — np.insert returns a new array; rebind the variable.
  • Shape your values — For axis inserts, make sure the inserted vector matches the axis length.
  • Target version quirks — Recent releases changed boolean index handling; skim the note when masks drive the insert points.

Final Word: Turn The Error Into A One-Line Fix

Change the call site and move on. Replace the method-style call with the function version and reassign the result:

# Bad
arr.insert(0, x)        # triggers: AttributeError: 'NumPy NdArray' Object Has No Attribute 'Insert'

# Good
arr = np.insert(arr, 0, x)

This one switch removes the AttributeError: ‘NumPy NdArray’ Object Has No Attribute ‘Insert’ message, keeps your code tidy, and matches the library’s documented API.