galform_analysis

A modular Python framework for reading and analysing outputs of the GALFORM semi-analytic model of galaxy formation, stored as galaxies.hdf5 files. It covers everything from low-level HDF5 I/O to mass functions, clustering statistics and redshift-space distortions.

Features

• I/O: robust readers for GALFORM galaxies.hdf5 files, tolerant of different output versions and incomplete subvolumes.
• Aggregation: scan a snapshot directory and stack galaxies across subvolumes into polars DataFrames.
• Mass functions: stellar mass functions (SMF), halo mass functions (HMF) and halo occupation distributions (HOD), per subvolume or averaged over subvolumes and snapshots.
• Theoretical predictions (optional): halo mass functions from hmf with GALFORM's Mvir mass definition, the GPS+ model, and the linear matter correlation function from CAMB.
• Clustering (optional): real-space 2-point correlation functions, satellite–central cross-correlations and galaxy bias, built on Corrfunc.
• Subvolume-weighted corrections (optional): ξ(r) and projected w_p(r_p) from an auto/cross pair-count decomposition that removes the sub-sampling bias when only some subvolumes are used.
• Redshift-space distortions (optional): ξ(s, μ) and the multipoles ξ₀, ξ₂, ξ₄.
• Simulation metadata: box sizes, cosmologies and snapshot redshifts for L800, Millennium I/II, EAGLE, COLIBRE, FLAMINGO, Dove and nIFTy.

Installation

pip install galform_analysis

The base install covers I/O, aggregation, simulation metadata and the SMF/HMF/HOD. Optional features are available as extras:

pip install "galform_analysis[all]"

Note on clustering: Corrfunc is distributed only as source and compiles during installation. It needs a C compiler with OpenMP and the GSL headers, for example sudo apt-get install libgsl-dev (Debian/Ubuntu), brew install gsl (macOS) or module load gsl on an HPC cluster. If an optional dependency is missing, the functions that need it raise an ImportError naming the extra to install. The rest of the package still works.

galform_analysis supports Python 3.10–3.14 and is tested on Linux.

Quick start

GALFORM writes one directory per snapshot, each holding one or more subvolumes:

<base_dir>/iz<NNN>/ivol<M>/galaxies.hdf5

Each ivol is an independent realisation of the full simulation box, not a spatial tile of it.

Simulation metadata

from galform_analysis import (
    SimulationConfig,
    find_snapshot_at_redshift,
    get_snapshot_redshift,
)

sim = SimulationConfig("L800")
print(sim.box_size, sim.omega_m, sim.h0, sim.n_subvolumes)

# Snapshot closest to z = 1 for this simulation: "iz176"
snapshot = find_snapshot_at_redshift(1.0, "L800")
z = get_snapshot_redshift(snapshot, "L800")

Reading a subvolume

from galform_analysis import close_snapshot, read_snapshot_data

data = read_snapshot_data("/path/to/Galform_Out/L800/model/iz271", ivol=0)
mstar = data["mstar"]  # stellar mass (disk + bulge), M_sun/h
mhalo = data["mhalo"]  # halo mass, M_sun/h
print(data["z"], data["V_ivol"])
close_snapshot(data)  # releases the open HDF5 file handle

Stellar and halo mass functions

from galform_analysis import set_base_dir, smf_given_redshift_and_subvolume
from galform_analysis.analysis import avg_smf_given_redshift_and_subvolumes

set_base_dir("/path/to/Galform_Out/L800/model")

smf = smf_given_redshift_and_subvolume("/path/to/Galform_Out/L800/model/iz271", ivol=0)
smf["centers"], smf["phi"]  # log10(M*) bin centres, phi in (Mpc/h)^-3 dex^-1

# Mean and scatter over subvolumes 0-7 of snapshot iz271, under the base dir
avg = avg_smf_given_redshift_and_subvolumes(271, ivols=list(range(8)))
avg["phi"], avg["phi_std"]

Correlation functions (requires clustering)

import numpy as np
from galform_analysis import compute_xi_corrfunc

rng = np.random.default_rng(1)
positions = rng.uniform(0, 100.0, size=(20_000, 3))  # (N, 3) array, Mpc/h
rbins = np.logspace(0, 1.5, 11)  # 1-32 Mpc/h
xi = compute_xi_corrfunc(positions, boxsize=100.0, rbins=rbins, nthreads=4)
xi["r"], xi["xi"]  # consistent with zero for a uniform random field

Functions that read data take nthreads arguments for Corrfunc. On shared compute nodes, set these (and OMP_NUM_THREADS) to your allocation.

Configuration

• Data location: call galform_analysis.set_base_dir(path) or set the GALFORM_BASE_DIR environment variable. Functions that take a snapshot number (such as 271 for iz271) rather than a path resolve it against this directory, unless you pass base_dir=. There is no default: if none is configured these functions raise an error explaining how to set one.
• Simulation configs: these come from the galform_execution package, so analysis and execution use identical parameters. A copy bundled with this package is used as a fallback.
• Snapshot redshifts: load_redshift_mapping(sim_name) maps iz indices to redshifts for simulations with a bundled redshift list.

Examples

Jupyter notebooks covering each module are in the examples/ directory of the repository, including reading snapshots, stellar mass functions, correlation functions and RSD multipoles.

Development

git clone https://github.com/OscarHickman/galform_analysis.git
cd galform_analysis
pip install -e ".[all,dev]"

pytest tests --cov       # test suite with coverage (CI requires >= 80%)
ruff check galform_analysis tests
ruff format galform_analysis tests

The tests build synthetic galaxies.hdf5 files that follow the real GALFORM schema, so no simulation data is needed. Tests that need an optional dependency are skipped when it is not installed.

Citation

If you use galform_analysis in your research, please cite it using the metadata in CITATION.cff:

@software{galform_analysis,
  author  = {Hickman, Oscar},
  title   = {galform_analysis: A modular Python framework for analysing GALFORM simulation outputs},
  version = {0.2.0},
  url     = {https://github.com/OscarHickman/galform_analysis}
}

License

MIT. See LICENSE.