PySplashSurf

0.14.1.0 verified Fri May 01 auth: no python

Python bindings for splashsurf, a high-performance surface reconstruction library for SPH (Smoothed Particle Hydrodynamics) simulations. Current version 0.14.1.0, release cadence irregular; follows the underlying C++ library.

pip install pysplashsurf

Common errors

error ImportError: cannot import name 'reconstruct_surface' from 'splashsurf' ↓

cause Old version of the library (pre-0.12) where function was named differently or not exported at top level.

fix

Upgrade to latest version: pip install --upgrade pysplashsurf. If still fails, check installed version with pip show pysplashsurf.

error ValueError: Buffer dtype mismatch, expected 'float64' but got 'float32' ↓

cause Particle positions array has dtype float32 instead of float64.

fix

Convert positions to float64 before calling: positions = positions.astype(np.float64).

error Segmentation fault (core dumped) ↓

cause Non-contiguous array (e.g., transpose or slice) passed to the C++ backend.

fix

Make array contiguous: positions = np.ascontiguousarray(positions).

Warnings

breaking The Python API changed between version 0.11.x and 0.12.x: the function `reconstruct_surface` now returns a tuple `(vertices, faces)` instead of a namedtuple-like object. Old code using attribute access will break. ↓

fix Update code to unpack the tuple: `vertices, faces = reconstruct_surface(positions)`.

gotcha Input particle positions must be a contiguous numpy array of dtype float64 (double precision). Using float32 or non-contiguous arrays (e.g., slices) may cause segmentation faults or incorrect results. ↓

fix Ensure positions are `np.ascontiguousarray(positions, dtype=np.float64)`.

gotcha On macOS with Apple Silicon, the pip-installed wheel may fall back to a slow pure-Python fallback if the native binary is not available. Install from conda-forge or build from source for optimal performance. ↓

fix Use `conda install -c conda-forge pysplashsurf` or build from source following the GitHub instructions.

Imports

reconstruct_surface
wrong
```
import splashsurf; splashsurf.reconstruct_surface
```
correct
```
from splashsurf import reconstruct_surface
```
Function is a module-level export; direct attribute access may fail depending on version.
compute_normals
```
from splashsurf import compute_normals
```
mesh_to_obj
```
from splashsurf import mesh_to_obj
```

Quickstart

Basic surface reconstruction from particle positions.

import numpy as np
from splashsurf import reconstruct_surface

# Example: random particle positions
positions = np.random.rand(1000, 3).astype(np.float64)

# Reconstruct surface (returns vertices, faces as numpy arrays)
vertices, faces = reconstruct_surface(positions)
print(f"Surface: {len(vertices)} vertices, {len(faces)} faces")