findiff

0.13.1 verified Mon Apr 27 auth: no python

A Python package for finite difference derivatives in any number of dimensions. Current version 0.13.1, released on PyPI. Provides tools to compute partial derivatives of discrete functions on uniform and non-uniform grids with automatic stencil generation. Supports periodic boundary conditions and symbolic capabilities. Active development with periodic releases.

pip install findiff

Common errors

error ModuleNotFoundError: No module named 'findiff.fin_diff' ↓

cause The module path changed in v0.11.0. Old code using `from findiff.fin_diff import FinDiff` will fail.

fix

Change import to from findiff import FinDiff.

error ValueError: The dimension parameter must be an integer. ↓

cause When initializing `FinDiff` with a tuple for grid coords (like `FinDiff(0, x, 1)`), the grid coords were passed incorrectly if x is a 1D array but you intended to use a multi-dimensional grid.

fix

Ensure the second argument is the list of grid coordinates for each dimension. For 1D, FinDiff(0, x, 1) is correct. For higher dimensions, pass a list of arrays, e.g., FinDiff(0, [x, y], 1).

error AttributeError: 'Coefficient' object has no attribute 'eval' ↓

cause In older versions (<0.10.0), the Coefficient class did not have an eval method. Symbolic capabilities were introduced in v0.10.0.

fix

Upgrade findiff to 0.10.0 or later: pip install --upgrade findiff.

error TypeError: __init__() got an unexpected keyword argument 'acc' ↓

cause In v0.11.0, the `acc` parameter was renamed to `accuracy` in some contexts, or the Diff class expects different arguments.

fix

Use accuracy instead of acc when initializing FinDiff or Diff. For example: FinDiff(0, x, 1, accuracy=2).

Warnings

breaking In v0.11.0, the API was significantly restructured. The old `FinDiff` initialization with positional arguments may break if you relied on old defaults. The `Diff` class was introduced; stencils are now lazy. ↓

fix Update import to `from findiff import FinDiff, Diff`. If using old positional arguments, ensure accuracy is passed correctly. Review migration guide in changelog.

gotcha The `to_sparse()` method returns a sparse matrix only if SciPy is installed. Without it, the method raises an ImportError. ↓

fix Install scipy: `pip install scipy` before calling to_sparse().

gotcha When using non-uniform grids, the accuracy parameter may not behave as expected for high-order derivatives. The stencil generation assumes uniform spacing by default. ↓

fix For non-uniform grids, consider using the `Diff` class with explicit grid coordinates. Check the documentation for `FinDiff` on non-uniform grids.

deprecated The old import path `from findiff.fin_diff import FinDiff` is deprecated since v0.11.0 and may be removed in future versions. ↓

fix Use `from findiff import FinDiff`.

Imports

FinDiff
wrong
```
from findiff.fin_diff import FinDiff
```
correct
```
from findiff import FinDiff
```
The old submodule path was deprecated in v0.11.0. Use top-level import.
Diff
```
from findiff import Diff
```
Introduced in v0.11.0 for easier definition of differential operators.
Coefficient
```
from findiff import Coefficient
```
Used to define variable coefficients on a grid.
Id
```
from findiff import Id
```
Identity operator for finite differences.
FinEq
```
from findiff import FinEq
```
Finite difference equation class.

Quickstart

Compute the first derivative of sin(x) on a uniform grid with second-order accuracy.

import numpy as np
from findiff import FinDiff

# Define a grid and a function
x = np.linspace(0, 1, 5)
f = np.sin(x)

# First derivative (second order accuracy)
d_dx = FinDiff(0, x, 1, acc=2)
df_dx = d_dx(f)
print(df_dx)