Mera.jl public test simulations, v2

Small RAMSES simulation outputs you can download and analyse with
Mera.jl. They are the same simulations Mera's own test
suite runs against, so every one has a known answer to check against.

Total 307 MB across 16 archives (15 simulations plus the documentation set).
Each unpacks to a single directory.

v2 adds four simulations and changes none: sedov3d_bisection, dm_gas3d_bisection and
sedov3d_iogroups, RAMSES's own outputs in its bisection ordering and in IO-group folders, and
cosmo3d, a small cosmological run from RAMSES's own initial conditions (one snapshot; the initial
conditions themselves are not included, NOTICE.md says where to get them). All four serve Mera's
repartition and writeramses. The other eleven are the same data as in testdata-v1,
which stays in place unchanged for Mera 1.8. The tag is versioned independently of the package:
it changes only when the set of simulations changes.

Getting them

git clone https://github.com/ManuelBehrendt/Mera.jl && cd Mera.jl
./testdata/fetch_fixtures.sh --small     # 142 MB, everything except the Bondi run
./testdata/fetch_fixtures.sh             # everything, 307 MB
julia --project -e 'using Pkg; Pkg.test("Mera")'

The script downloads only what is missing, and skips the download entirely if the test
simulations are already on disk.

Try one

These are ordinary RAMSES outputs, so there is nothing special about opening them. Point Mera at a
directory and explore. sedov3d_amr is a good first one: 2 MB, a spherical blast wave on an AMR
grid, and every quantity has a value you can check against theory.

using Mera
info = getinfo(7, "sedov3d_amr")        # the last snapshot
gas  = gethydro(info)

println(length(gas.data), " cells over levels ", gas.lmin, " to ", gas.lmax)
println("time ", info.time)

# any derived quantity works the same way as on a research-scale run
rho = getvar(gas, :rho, :nH)
println("density ", minimum(rho), " to ", maximum(rho), " cm^-3")

# and a map
projection(gas, :sd, :Msol_pc2, pxsize=[0.005, :standard])   # ~100 x 100 pixels

Nothing here is test-only: subregion, profile, filterdata, savedata and the rest behave
exactly as they do on a multi-gigabyte simulation, just faster. If you want something with more
physics in it, sedov3d_grav_part adds gravity and particles, mhdtube3d adds magnetic fields,
and stromgren3d adds radiative transfer.

The Mera documentation works through all of
this in detail.

What each one is

Generated from namelists committed in testdata/namelists/, so every one is
reproducible from first principles given RAMSES itself. Each is checked against an analytic
oracle, meaning something theory fixes independently of Mera.

RAMSES's own test configurations

These three are not ours. They are configurations from RAMSES's own test suite, run
unmodified, so the reference solutions the RAMSES developers validate their solver against apply
directly. That is 87 published quantities here, or 100 counting sedov3d_amr's 13, checked 1:1
using their reduction
(tests/visu/visu_ramses.py :: check_solution) at each test's own published tolerance
(3e-13 by default; 8e-11 for rt-dirac).

Source: https://github.com/ramses-organisation/ramses, tag 2026.05.

READMEs.tar.gz holds the README, the third-party NOTICE.md and RAMSES's licence.

Each simulation directory also carries a SOURCE.txt: where the data came from, the namelist
it was produced from, and the RAMSES attribution. A folder copied elsewhere keeps its
provenance without needing this page.

Attribution and licence

Mera.jl is MIT-licensed, but this data is not entirely ours. It is produced by RAMSES
(Copyright CEA and Romain Teyssier), governed by the CeCILL licence
(http://www.cecill.info), and three of these archives redistribute RAMSES namelists unmodified.
READMEs.tar.gz carries NOTICE.md and RAMSES's licence text; please cite
Teyssier (2002), and Rosdahl et al. (2013) for the RT test simulations, if you use these in published
work.