Built-in Datasets

lcdata contains scripts that can be used to download commonly-used light curve datasets and save them in an efficient and fast-to-read HDF5 format. The currently-supported datasets are:

Dataset	Script	Reference
PLAsTiCC	lcdata_download_plasticc	PLAsTiCC
PanSTARRS-1	lcdata_download_ps1	Villar et al. 2020

After installing lcdata, these scripts can be run from any directory on the command line to download the corresponding dataset(s). By default, these will be placed in ./data/, although the location can be changed with the --directory flag.

Using downloaded datasets

A full description of how datasets are used in lcdata can be found on the Usage page.

To open a dataset that can fit in memory and read metadata/light curves from it:

>>> dataset = lcdata.read_hdf5('./data/ps1.h5')

>>> print(dataset.meta)
<Table length=5243>
object_id    ra      dec      type  redshift ...
   str9   float64  float64   str13  float64  ...
--------- -------- -------- ------- -------- ...
PS0909006 333.9503   1.1848    SNIa    0.284 ...
PS0909010  37.1182  -4.0789    SNIa     0.27 ...
PS0910012  52.4718 -28.0867   SNIax    0.079 ...
PS0910016  35.3073    -3.91    SNIa     0.23 ...
      ...      ...      ...     ...      ... ...

>>> print(dataset.light_curves[0])
<Table length=376>
  time    flux  fluxerr  band
float64 float32 float32 bytes6
------- ------- ------- ------
55029.6  -9.919  11.995 ps1::g
55074.4 211.372   33.03 ps1::g
55086.4 112.477  10.463 ps1::g
55089.4 102.993   10.58 ps1::g
    ...     ...     ...    ...

For large datasets that can’t fit into memory, a common workflow is to process the dataset in smaller chunks that can fit in memory. To load a large dataset in chunks of 1000 light curves:

>>> dataset = lcdata.read_hdf5('./data/plasticc_test.h5', in_memory=False)

>>> for chunk in dataset.iterate_chunks(chunk_size=1000):
...    # At each iteration, chunk is an lcdata Dataset with the next 1000
...    # light curves.