simplest_dask_gcsfs.py

import dask.dataframe as dd
import gcsfs  # registers with dask filesystems

# Initial authentication
# gcs = gcsfs.core.GCSFileSystem(project='continuum-compute',
# token='/Users/mdurant/.config/gcloud/application_default_credentials.json')

# Tiny block-size to ensure more than one block
df = da.read_csv('s3://blaze-data/iris/*', blocksize=2500,
                 names=['petal_length', 'sepal_length', 'petal_width',
                        'sepal_width', 'species'])
df.to_parquet('gs://dask_example_data/iris',
              storage_options={'project': 'continuum-compute'})
df2 = dd.read_parquet('gs://dask_example_data/iris',
                      storage_options={'project': 'continuum-compute'})
# suggest project as part of URL: gs://continuum-compute@dask_example_data/iris , 
# does not yet work

df2.compute()