- Add
dask_kwargs
to therhg_compute_tools.xarray
functions
- Add
retry_with_timeout
torhg_compute_tools.utils.py
- Drop
matplotlib.font_manager._rebuild()
call indesign.__init__
- no longer supported (:issue:`96`)
- Refactor
datasets_from_delayed
to speed up
- Add gcs.ls function
- Fix tag kwarg in get_cluster
- Make the gsutil API consistent, so that we have cp, sync and rm, each of which accept the same args and kwargs (:issue:`69`)
- Swap
bumpversion
forsetuptools_scm
to handle versioning (:issue:`78`) - Cast coordinates to dict before gathering in
rhg_compute_tools.xarray.dataarrays_from_delayed
andrhg_compute_tools.xarray.datasets_from_delayed
. This avoids a mysterious memory explosion on the local machine. Also addname
in the metadata used by those functions so that the name of each dataarray or Variable is preserved. (:issue:`83`) - Use
dask-gateway
when available when creating a cluster inrhg_compute_tools.kubernetes
. Add some tests using a local gateway cluster. TODO: More tests. - Add
tag
kwarg torhg_compute_tools.kuberentes.get_cluster
function (PR #87)
- ?
- Add remote scheduler deployment (part of dask_kubernetes 0.10)
- Remove extraneous GCSFUSE_TOKENS env var no longer used in new worker images
- Set library thread limits based on how many cpus are available for a single dask thread
- Change formatting of the extra env_items passed to get_cluster to be a list rather than a list of dict-like name/value pairs
- Add CLI tools (:issue:`37`). See
rctools gcs repdirstruc --help
to start - Add new function
rhg_compute_tools.gcs.replicate_directory_structure_on_gcs
to copy directory trees into GCS. Users can authenticate with cred_file or with default google credentials (:issue:`51`) - Fixes to docstrings and metadata (:issue:`43`) (:issue:`45`)
- Add new function
rhg_compute_tools.gcs.rm
to remove files/directories on GCS using thegoogle.cloud.storage
API - Store one additional environment variable when passing
cred_path
torhg_compute_tools.kubernetes.get_cluster
so that thegoogle.cloud.storage
API will be authenticated in addition togsutil
- Deployment fixes
- Design tools: use RHG & CIL colors & styles
- Plotting helpers: generate cmaps with consistent colors & norms, and apply a colorbar to geopandas plots with nonlinear norms
- Autoscaling fix for kubecluster: switch to dask_kubernetes.KubeCluster to allow use of recent bug fixes
- Add
rhg_compute_tools.gcs.cp_gcs
andrhg_compute_tools.gcs.sync_gcs
utilities
- need to figure out how to use this rever thing
- Bug fix again in
rhg_compute_tools.kubernetes.get_worker
- Bug fix in
rhg_compute_tools.kubernetes.get_worker
- Add xarray from delayed methods in
rhg_compute_tools.xarray
(:issue:`12`) rhg_compute_tools.gcs.cp_to_gcs
now callsgsutil
in a subprocess instead ofgoogle.storage
operations. This dramatically improves performance when transferring large numbers of small files (:issue:`11`)- Additional cluster creation helpers (:issue:`3`)
- New google compute helpers (see
rhg_compute_tools.gcs.cp_to_gcs
,rhg_compute_tools.gcs.get_bucket
) - New cluster creation helper (see
rhg_compute_tools.kubernetes.get_worker
) - Dask client.map helpers (see
rhg_compute_tools.utils submodule
)
- First release on PyPI.