Migrating from FinDiff to Diff

Starting with version 0.11 the recommended API uses the Diff class. The old FinDiff class still works but emits a DeprecationWarning and will be removed in a future release.

This page maps old patterns to their new equivalents.

Basic Derivatives 

Old (`FinDiff`)	New (`Diff`)
`FinDiff(0, dx)`	`Diff(0, dx)`
`FinDiff(0, dx, 2)`	`Diff(0, dx) ** 2`
`FinDiff(1, dy, 3)`	`Diff(1, dy) ** 3`

Mixed Partial Derivatives 

Old (`FinDiff`)	New (`Diff`)
`FinDiff((0, dx), (1, dy))`	`Diff(0, dx) * Diff(1, dy)`
`FinDiff((0, dx, 2), (1, dy))`	`Diff(0, dx)*2 Diff(1, dy)`

Accuracy Control 

Old (`FinDiff`)	New (`Diff`)
`FinDiff(0, dx, acc=4)`	`Diff(0, dx, acc=4)`
`FinDiff(0, dx, 2, acc=6)`	`Diff(0, dx, acc=6) ** 2`

Coefficient and Identity 

Old	New
`Coef(x) * FinDiff(0, dx)`	`x * Diff(0, dx)`
`Id()`	`Identity()`

Operator Arithmetic 

Both old and new APIs support the same arithmetic. The main difference is how you express higher-order and mixed derivatives:

Old style:

from findiff import FinDiff, Coef, Id

L = Coef(2) * FinDiff(0, dx, 2) + Coef(3) * FinDiff(1, dy, 2)

New style:

from findiff import Diff

L = 2 * Diff(0, dx)**2 + 3 * Diff(1, dy)**2

Lazy Grid Setting 

With Diff you can define operators without specifying the grid and set it later:

L = Diff(0)**2 + Diff(1)**2
L.set_grid({0: dx, 1: dy})
result = L(f)

This was not possible with FinDiff.

Matrix Representation 

Both APIs use the same matrix() method:

# Old
mat = FinDiff(0, dx, 2).matrix(shape)

# New
mat = (Diff(0, dx) ** 2).matrix(shape)

Suppressing the Deprecation Warning 

If you need to keep using FinDiff temporarily, suppress the warning:

import warnings
warnings.filterwarnings('ignore', message='FinDiff is deprecated')