v0.7.0

What's Changed

• Agg fix by @kthyng in #55
  • The sorting in filedates2df was slightly wrong and would not consistently return the desired nowcast file over the forecast file. Seems to be correct now.
  • WCOFS has a different update frequency which leads to a different conversion from filename to datetime. This is now included in file2dt.

Full Changelog: v0.6.1...v0.7.0

v0.6.1

What's Changed

• small changes by @kthyng in #54

Full Changelog: v0.6.0...v0.6.1

v0.6.0

What's Changed

• Updates due to tests failing by @kthyng in #49
• small bug fix by @kthyng in #50
• series of fixes by @kthyng in #52
• Split by @kthyng in #53
  • Updated docs.
  • The "freshness" parameter, which determines how much time can pass before different actions must be rerun, now has a default for each of the five actions that have freshness parameters associated with them. These parameters are set in the __init__ file but can be overridden for any model source with the "freshness" parameter in the source metadata. More details are available: https://model-catalogs.readthedocs.io/en/latest/add_model.html.
  • The "known" GOODS model catalog yaml files are no longer distributed with model_catalogs itself in order to enforce more separation between the catalog files themselves and this code. However, the package of catalogs is currently a requirement of model_catalogs and can be found at mc-goods: https://github.com/axiom-data-science/mc-goods. Note that catalog names that had names like CBOFS-RGRID are now called CBOFS_RGRID with underscores instead of hyphens. This was a necessary change for setting up the models in their own packages with entry points.
  • Enforcing single threading in model_catalogs to avoid issue when using xr.open_mfdataset (which is used with noagg sources) in which the first time you read something in you hit an error but the second time it works. For more information check this xarray issue or this netcdf issue.
  • User can work with a local catalog file now! See for details: https://model-catalogs.readthedocs.io/en/latest/catalog_modes.html
    • boundaries are optionally calculated when using mc.open_catalog().
    • boundaries are calculated the first time a catalog file is worked with through mc.setup()
  • Removed requirement for filetype to be in catalog if sources in catalog do not need to be aggregated.
  • LSOFS and LOOFS have new FVCOM versions. So, there are new versions of the model files:
    • lsofs.yaml and loofs.yaml are still the legacy POM version of the models but no longer have source coops-forecast-noagg, and their metadata have been updated to reflect the end dates of the model sources.
    • new catalog files lsofs-fvcom.yaml and loofs-fvcom.yaml have source coops-forecast-noagg that points to the new FVCOM version of the models.
  • If user requests time range that is not available for a source, it will now error instead of warn.
  • Bug fixed in find_availability so that when a source that does not have a catloc entry is checked, the Dataset is read in without extra processing and checks (including limiting the time range which otherwise would impact checking the time availability).

Full Changelog: v0.5.0...v0.6.0

v0.5.0

What's New

• Added long_name and html_desc to the catalog files
• Revised overall_start/end_datetime attributes to be more consistent
• Renamed Intake Source names from "timing"s to "model_source"s. Also some of the specific model names changed. All source names changed to reflect where the model output is coming from and whether the files are unaggregated or not.

v0.4.1

What's New

• Updates to select_date_range():
  • can input catalog or source
  • if find_availability needs to have been run, it will be run for you
• added note about choosing chunks to How to Add a New Model

v0.4.0

What's New

• Files found for unaggregated NOAA OFS models are now sorted by correct time order.
• Repeated files (between nowcast and forecast files, and different nowcast timing cycles) are not included anymore.
• Improved behavior in aggregated NOAA OFS file selection for seamless time series between days with no forecast included and the final day of the selection that includes a forecast.
• Updated many details in Model Aggregations page of docs.
• Package now available on conda-forge!
• source.follow_target() was renamed source.target. It exposes the original catalog information, which is the target of the source transform. It is a property now instead of a method.
• Fixed bug when applying standard_name attributes to coordinate variables.
• model_catalogs can now understand the datetimes associated with NOAA OFS models. See Model Aggregations for more details.
• Improved date and datetime behavior for select_date_range().
• Updates chunking parameter for several models for performance reasons.
• Updates CF coordinate standard names for several catalogs.
• Input to mc.find_availability() can now be an Intake Catalog or Intake Source.
• More robust calls to mc.find_availability() in case time range cannot be found.
• Better server checks, using requests package to see if server is working. A source's server status can be checked with source.status.

v0.3.1

mandatory post release fix: fixed package reqs

v0.3.0

What's New!

• Can now query the transform catalogs for the target urlpath (see docs <https://model-catalogs.readthedocs.io/en/latest/demo.html#urlpath:-model-output-source>_)
• Can explicitly follow the transform catalog to the target catalog with cat.follow_target().
• select_date_range() now explicitly replaces the urlpath of the target when the function is run rather than later when to_dask() is run.

v0.2.0

Expanded documentation!

v0.1.1

Updates to fix PyPI release.