Below are instructions for loading all supported datasets. All examples use the freely available Sample Data.
Amiga Halo Finder¶
There are a couple data formats associated with the Amiga Halo Finder. Both formats save a series of files associated with each snapshot. Parameters are stored in “.parameters” and “.log” files and halo properties in “.AHF_halos” files. In the older format, descendent/ancestor links are stored in several “.AHF_mtree” files, one per snapshot. In the newer format, all halo linking information is stored in a single file beginning with “MergerTree_” and ending with “-CRMratio2”. Make sure to keep all these files together in the same directory. To load, provide the name of the first “.parameter” file.
>>> import ytree >>> a = ytree.load("ahf_halos/snap_N64L16_000.parameter", ... hubble_constant=0.7)
Alternatively, the “MergerTree_” file can also be provided for the newer format.
>>> import ytree >>> a = ytree.load("AHF_100_tiny/MergerTree_GIZMO-NewMDCLUSTER_0047.txt-CRMratio2")
Four important notes about loading AHF data:
- The dimensionless Hubble parameter is not provided in AHF
outputs. This should be supplied by hand using the
hubble_constantkeyword. The default value is 1.0.
- If the “.log” file is named in a unconventional way or cannot
be found for some reason, its path can be specified with the
log_filenamekeyword argument. If no log file exists, values for
box_size(in units of Mpc/h) can be provided with keyword arguments named thusly.
- There will be no “.AHF_mtree” file for index 0 as the “.AHF_mtree” files store links between files N-1 and N.
ytreeis able to load data where the graph has been calculated instead of the tree. However, even in this case, only the tree is preserved in
ytree. See the Amiga Halo Finder Documentation for a discussion of the difference between graphs and trees.
The consistent-trees format consists of a set of files called “locations.dat”, “forests.list”, at least one file named something like “tree_0_0_0.dat”. For large simulations, there may be a number of these “tree_*.dat” files. After running Rockstar and consistent-trees, these will most likely be located in the “rockstar_halos/trees” directory. The full data set can be loaded by providing the path to the locations.dat file.
>>> import ytree >>> a = ytree.load("tiny_ctrees/locations.dat")
Alternatively, data from a single tree file can be loaded by providing the path to that file.
>>> import ytree >>> a = ytree.load("consistent_trees/tree_0_0_0.dat")
Consistent-Trees hlist Files¶
While running consistent-trees, a series of files will be created in the “rockstar_halos/hlists” directory with the naming convention, “hlist_<scale-factor>.list”. These are the catalogs that will be combined to make the final output files. However, these files contain roughly 30 additional fields that are not included in the final output. Merger trees can be loaded by providing the path to the first of these files.
>>> import ytree >>> a = ytree.load("ctrees_hlists/hlists/hlist_0.12521.list")
Note, loading trees with this method will be slower than using
the standard consistent-trees output file as
ytree will have to
assemble each tree across multiple files. This method is not
recommended unless the additional fields are necessary.
is a variant of the consistent-trees format built on HDF5. It is used by
the Skies & Universe project.
This format allows for access by either forests or trees as per the
definitions above. The data can be stored as either a struct of arrays
or an array of structs. Both layouts are supported, but
currently optimized for the struct of arrays layout. Field access with
struct of arrays will be 1 to 2 orders of magnitude faster than with
array of structs.
Datasets from this format consist of a series of HDF5 files with the naming convention, forest.h5, forest_0.5, …, forest_N.h5. The numbered files contain the actual data while the forest.h5 file contains virtual datasets that point to the data files. To load all the data, provide the path to the virtual dataset file:
>>> import ytree >>> a = ytree.load("consistent_trees_hdf5/soa/forest.h5")
To load a subset of the full dataset, provide a single data file or a list/tuple of files.
>>> import ytree >>> # single file >>> a = ytree.load("consistent_trees_hdf5/soa/forest_0.h5") >>> # multiple data files (sample data only has one) >>> a = ytree.load(["forest_0.h5", "forest_1.h5"])
Access by Forest¶
ytree will load consistent-trees-hdf5 datasets to
provide access to each tree, such that
a[N] will return the Nth
tree in the dataset and
a[N]["tree"] will return all halos in
that tree. However, by providing the
access="forest" keyword to
load, data will be loaded
according to the forest it belongs to.
>>> import ytree >>> a = ytree.load("consistent_trees_hdf5/soa/forest.h5", ... access="forest")
In this mode,
a[N] will return the Nth forest and
a[N]["forest"] will return all halos in that forest. In
forest access mode, the “root” of the forest, i.e., the
TreeNode object returned
a[N] will be the root of one of the trees in that
forest. See Accessing All Nodes in a Forest for how to locate all individual
trees in a forest.
The Gadget4 format consists of one or more HDF5 files. Each file contains information on the trees contained within as well as some or all of the associated field data for those trees. Field data for large trees can span multiple data files and the start of any file does not necessarily correspond to the start of field data for the trees it holds. This format supports Accessing All Nodes in a Forest.
To load single-file data, load with the path to that file.
>>> import ytree >>> a = ytree.load("gadget4/trees/trees.hdf5")
To load a dataset consisting of multiple files, provide the path to the zeroth file.
>>> import ytree >>> a = ytree.load("gadget4/treedata/trees.0.hdf5")
For multi-file datasets, all data files must be present for the
dataset to be loaded. It is not possible to load a subseta
multi-file dataset. Because data for any given tree is only loaded
when needed, there is little benefit to trying to load a subset of
the full data. However, if you really want to limit your dataset to
a selection of the full data, your best bet is to save just the
trees you want to a new dataset using the
See Saving Arbors and Trees for more information.
The LHaloTree format is typically one or more files with a naming convention like “trees_063.0” that contain the trees themselves and a single file with a suffix “.a_list” that contains a list of the scale factors at the time of each simulation snapshot.
The LHaloTree format loads halos by forest. There is no need
to provide the
access="forest" keyword here.
In addition to the LHaloTree files,
ytree also requires additional
information about the simulation from a parameter file (in
Gadget format). At
minimum, the parameter file should contain the cosmological parameters
HubbleParam, Omega0, OmegaLambda, BoxSize, PeriodicBoundariesOn,
ComovingIntegrationOn, and the unit parameters
UnitVelocity_in_cm_per_s, UnitLength_in_cm, and
If not specified explicitly (see below), a file with the extension
“.param” will be searched for in the directory containing the
If all of the required files are in the same directory, an LHaloTree catalog can be loaded from the path to one of the tree files.
>>> import ytree >>> a = ytree.load("lhalotree/trees_063.0")
Both the scale factor and parameter files can be specified explicitly through keyword arguments if they do not match the expected pattern or are located in a different directory than the tree files.
>>> a = ytree.load("lhalotree/trees_063.0", ... parameter_file="lhalotree/param.txt", ... scale_factor_file="lhalotree/a_list.txt")
The scale factors and/or parameters themselves can also be passed explicitly from python.
>>> import numpy as np >>> parameters = dict(HubbleParam=0.7, Omega0=0.3, OmegaLambda=0.7, ... BoxSize=62500, PeriodicBoundariesOn=1, ComovingIntegrationOn=1, ... UnitVelocity_in_cm_per_s=100000, UnitLength_in_cm=3.08568e21, ... UnitMass_in_g=1.989e+43) >>> scale_factors = [ 0.0078125, 0.012346 , 0.019608 , 0.032258 , 0.047811 , ... 0.051965 , 0.056419 , 0.061188 , 0.066287 , 0.071732 , ... 0.07754 , 0.083725 , 0.090306 , 0.097296 , 0.104713 , ... 0.112572 , 0.120887 , 0.129675 , 0.13895 , 0.148724 , ... 0.159012 , 0.169824 , 0.181174 , 0.19307 , 0.205521 , ... 0.218536 , 0.232121 , 0.24628 , 0.261016 , 0.27633 , ... 0.292223 , 0.308691 , 0.32573 , 0.343332 , 0.361489 , ... 0.380189 , 0.399419 , 0.419161 , 0.439397 , 0.460105 , ... 0.481261 , 0.502839 , 0.524807 , 0.547136 , 0.569789 , ... 0.59273 , 0.615919 , 0.639314 , 0.66287 , 0.686541 , ... 0.710278 , 0.734031 , 0.757746 , 0.781371 , 0.804849 , ... 0.828124 , 0.851138 , 0.873833 , 0.896151 , 0.918031 , ... 0.939414 , 0.960243 , 0.980457 , 1. ] >>> a = ytree.load("lhalotree/trees_063.0", ... parameters=parameters, ... scale_factors=scale_factors)
This is the same algorithm as LHaloTree, except with data saved in HDF5 files instead of unformatted binary. LHaloTree-HDF5 is one of the formats used by the Illustris-TNG project and is described in detail here. Like LHaloTree, this format supports accessing trees by forest. The LHaloTree-HDF5 format stores trees in multiple HDF5 files contained within a single directory. Each tree is fully contained within a single file, so loading is possible even when only a subset of all files is present. To load, provide the path to one file.
>>> import ytree >>> a = ytree.load("TNG50-4-Dark/trees_sf1_099.0.hdf5")
The files do not contain information on the box size and cosmological parameters of the simulation, but they can be provided by hand, with the box size assumed to be in units of comoving Mpc/h.
>>> import ytree >>> a = ytree.load("TNG50-4-Dark/trees_sf1_099.0.hdf5", ... box_size=35, hubble_constant=0.6774, ... omega_matter=0.3089, omega_lambda=0.6911)
The LHaloTree-HDF5 format contains multiple definitions of halo mass (see here), and as such, the field alias “mass” is not defined by default. However, the alias can be created if one is preferable. This is also necessary to facilitate Accessing the Progenitor Lineage of a Tree.
>>> a.add_alias_field("mass", "Group_M_TopHat200", units="Msun")
MORIA is a merger tree extension of the SPARTA code (Diemer 2017; Diemer 2020a). An output from MORIA is a single HDF5 file, whose path should be provided for loading.
>>> import ytree >>> a = ytree.load("moria/moria_tree_testsim050.hdf5")
Merger trees in MORIA are organized by forest, so
a.size (following the example above) will give the number of
forests, not the number of trees. MORIA outputs contain multiple definitions
of halo mass (see here),
and as such, the field alias “mass” is not defined by default. However,
the alias can be created if one is preferable. This
is also necessary to facilitate Accessing the Progenitor Lineage of a Tree.
>>> a.add_alias_field("mass", "Mpeak", units="Msun")
On rare occasions, a halo will be missing from the output even though
another halo claims it as its descendent. This is usually because the
halo has dropped below the minimum mass to be included. In these cases,
MORIA will reassign the halo’s descendent using the
field (see discussion in here).
ytree encounters such a situation, a message like the one below
will be printed.
>>> t = a >>> print (t["tree", "Mpeak"]) ytree: [INFO ] 2021-05-04 15:29:19,723 Reassigning descendent of halo 374749 from 398837 to 398836. [1.458e+13 1.422e+13 1.363e+13 1.325e+13 1.295e+13 1.258e+13 1.212e+13 ... 1.309e+11 1.178e+11 1.178e+11 1.080e+11 9.596e+10 8.397e+10] Msun/h
Rockstar catalogs with the naming convention “out_*.list” will contain information on the descendent ID of each halo and can be loaded independently of consistent-trees. This can be useful when your simulation has very few halos, such as in a zoom-in simulation. To load in this format, simply provide the path to one of these files.
>>> import ytree >>> a = ytree.load("rockstar/rockstar_halos/out_0.list")
Merger trees created with treefarm can be loaded in by providing the path to one of the catalogs created during the calculation.
>>> import ytree >>> a = ytree.load("tree_farm/tree_farm_descendents/fof_subhalo_tab_000.0.h5")
TreeFrog generates merger trees primarily for VELOCIraptor halo catalogs. The TreeFrog format consists of a series of HDF5 files. One file contains meta-data for the entire dataset. The other files contain the tree data, split into HDF5 groups corresponding to the original halo catalogs. To load, provide the path to the “foreststats” file, i.e., the one ending in “.hdf5”.
>>> import ytree >>> a = ytree.load("treefrog/VELOCIraptor.tree.t4.0-131.walkabletree.sage.forestID.foreststats.hdf5")
Merger trees in TreeFrog are organized by forest, so
a.size (following the example above) will give the number of
forests. Note, however, the id of the root halo for any given forest is not
the same as the forest id.
>>> my_tree = a >>> print (my_tree["uid"]) 131000000000001 >>> print (my_tree["ForestID"]) 104000000011727
TreeFrog outputs contain multiple definitions of halo mass, and as such, the field alias “mass” is not defined by default. However, the alias can be created if one is preferable. This is also necessary to facilitate Accessing the Progenitor Lineage of a Tree.
>>> a.add_alias_field("mass", "Mass_200crit", units="Msun")
Saved Arbors (ytree format)¶
Once merger tree data has been loaded, it can be saved to a
universal format using
save_tree. These can be loaded
by providing the path to the primary HDF5 file.
>>> import ytree >>> a = ytree.load("arbor/arbor.h5")
See Saving Arbors and Trees for more information on saving arbors and trees.