Feature/multi dates match

.pre-commit-config.yaml

@@ -44,7 +44,7 @@ repos:
   hooks:
   - id: ruff
     # Next line if for documenation cod snippets
-    exclude: '^[^_].*_\.py$'
+    exclude: '.*/[^_].*_\.py$'
     args:
     - --line-length=120
     - --fix

CHANGELOG.md

@@ -13,6 +13,7 @@ Keep it human-readable, your future self will thank you!
 ### Added
 
 - Adding the user recipe in the dataset PR #59.
+- Add `multi_dates_match` action in create.
 
 ### Changed
 

docs/building/sources.rst

@@ -28,6 +28,7 @@ The following `sources` are currently available:
    sources/netcdf
    sources/opendap
    sources/recentre
+   sources/repeated_dates
    sources/xarray-kerchunk
    sources/xarray-zarr
    sources/zenodo

docs/building/sources/repeated_dates.rst

@@ -0,0 +1,25 @@
+################
+ repeated_dates
+################
+
+The generale format of the `repeated_dates` source is:
+
+.. literalinclude:: yaml/repeated_dates1.yaml
+
+**********
+ constant
+**********
+
+.. literalinclude:: yaml/repeated_dates2.yaml
+
+*************
+ climatology
+*************
+
+.. literalinclude:: yaml/repeated_dates3.yaml
+
+*********
+ closest
+*********
+
+.. literalinclude:: yaml/repeated_dates4.yaml

docs/building/sources/yaml/repeated_dates1.yaml

@@ -0,0 +1,6 @@
+
+repeated_dates:
+    mode: mode
+    # ... parameters related to the mode ...
+    source:
+      # ... a source definition ...

docs/building/sources/yaml/repeated_dates2.yaml

@@ -0,0 +1,6 @@
+repeated_dates:
+  mode: constant
+  source:
+      xarray-zarr:
+        url: dem.zarr
+        variable: dem

docs/building/sources/yaml/repeated_dates3.yaml

@@ -0,0 +1,8 @@
+repeated_dates:
+    mode: climatology
+    year: 2019
+    day: 15
+    source:
+        grib:
+          path: some/path/to/data.grib
+          param: [some_param]

docs/building/sources/yaml/repeated_dates4.yaml

@@ -0,0 +1,9 @@
+repeated_dates:
+    mode: closest
+    frequency: 24h
+    maximum: 30d
+    skip_all_nans: true
+    source:
+      grib:
+          path: path/to/data.grib
+          param: [some_param]

docs/index.rst

@@ -1,8 +1,8 @@
 .. _index-page:
 
-####################################
- Welcome to Anemoi's documentation!
-####################################
+#############################################
+ Welcome to `anemoi-datasets` documentation!
+#############################################
 
 .. warning::
 

src/anemoi/datasets/create/__init__.py

@@ -25,9 +25,11 @@
 from anemoi.utils.dates import frequency_to_timedelta
 from anemoi.utils.humanize import compress_dates
 from anemoi.utils.humanize import seconds_to_human
+from earthkit.data.core.order import build_remapping
 
 from anemoi.datasets import MissingDateError
 from anemoi.datasets import open_dataset
+from anemoi.datasets.create.input.trace import enable_trace
 from anemoi.datasets.create.persistent import build_storage
 from anemoi.datasets.data.misc import as_first_date
 from anemoi.datasets.data.misc import as_last_date
@@ -309,7 +311,6 @@ def create_elements(self, config):
 
 
 def build_input_(main_config, output_config):
-    from earthkit.data.core.order import build_remapping
 
     builder = build_input(
         main_config.input,
@@ -563,7 +564,7 @@ def _run(self):
             # assert isinstance(group[0], datetime.datetime), type(group[0])
             LOG.debug(f"Building data for group {igroup}/{self.n_groups}")
 
-            result = self.input.select(dates=group)
+            result = self.input.select(group_of_dates=group)
             assert result.group_of_dates == group, (len(result.group_of_dates), len(group), group)
 
             # There are several groups.
@@ -1031,7 +1032,6 @@ def run(self):
 
 def creator_factory(name, trace=None, **kwargs):
     if trace:
-        from anemoi.datasets.create.trace import enable_trace
 
         enable_trace(trace)
 

src/anemoi/datasets/create/check.py

@@ -140,9 +140,15 @@ class StatisticsValueError(ValueError):
 
 def check_data_values(arr, *, name: str, log=[], allow_nans=False):
 
+    shape = arr.shape
+
     if (isinstance(allow_nans, (set, list, tuple, dict)) and name in allow_nans) or allow_nans:
         arr = arr[~np.isnan(arr)]
 
+    if arr.size == 0:
+        warnings.warn(f"Empty array for {name} ({shape})")
+        return
+
     assert arr.size > 0, (name, *log)
 
     min, max = arr.min(), arr.max()

src/anemoi/datasets/create/functions/sources/__init__.py

@@ -16,6 +16,10 @@
 
 
 def _expand(paths):
+
+    if not isinstance(paths, list):
+        paths = [paths]
+
     for path in paths:
         if path.startswith("file://"):
             path = path[7:]
@@ -40,8 +44,10 @@ def iterate_patterns(path, dates, **kwargs):
     given_paths = path if isinstance(path, list) else [path]
 
     dates = [d.isoformat() for d in dates]
+    if len(dates) > 0:
+        kwargs["date"] = dates
 
     for path in given_paths:
-        paths = Pattern(path, ignore_missing_keys=True).substitute(date=dates, **kwargs)
+        paths = Pattern(path, ignore_missing_keys=True).substitute(**kwargs)
         for path in _expand(paths):
             yield path, dates

src/anemoi/datasets/create/functions/sources/accumulations.py

@@ -376,6 +376,7 @@ def accumulations(context, dates, **request):
         ("ea", "oper"): dict(data_accumulation_period=1, base_times=(6, 18)),
         ("ea", "enda"): dict(data_accumulation_period=3, base_times=(6, 18)),
         ("rr", "oper"): dict(data_accumulation_period=3, base_times=(0, 3, 6, 9, 12, 15, 18, 21)),
+        ("l5", "oper"): dict(data_accumulation_period=1, base_times=(0,)),
     }
 
     kwargs = KWARGS.get((class_, stream), {})

src/anemoi/datasets/create/functions/sources/grib.py

@@ -135,7 +135,7 @@ def execute(context, dates, path, latitudes=None, longitudes=None, *args, **kwar
             s = s.sel(valid_datetime=dates, **kwargs)
             ds = ds + s
 
-    if kwargs:
+    if kwargs and not context.partial_ok:
         check(ds, given_paths, valid_datetime=dates, **kwargs)
 
     if geography is not None:

src/anemoi/datasets/create/functions/sources/xarray/__init__.py

@@ -10,10 +10,9 @@
 import logging
 
 from earthkit.data.core.fieldlist import MultiFieldList
-from earthkit.data.indexing.fieldlist import FieldArray
 
 from anemoi.datasets.data.stores import name_to_zarr_store
-from anemoi.datasets.utils.fields import NewMetadataField
+from anemoi.datasets.utils.fields import NewMetadataField as NewMetadataField
 
 from .. import iterate_patterns
 from .fieldlist import XarrayFieldList
@@ -31,7 +30,7 @@ def check(what, ds, paths, **kwargs):
         raise ValueError(f"Expected {count} fields, got {len(ds)} (kwargs={kwargs}, {what}s={paths})")
 
 
-def load_one(emoji, context, dates, dataset, options={}, match_all_dates=False, flavour=None, **kwargs):
+def load_one(emoji, context, dates, dataset, options={}, flavour=None, **kwargs):
     import xarray as xr
 
     """
@@ -52,12 +51,8 @@ def load_one(emoji, context, dates, dataset, options={}, match_all_dates=False,
 
     fs = XarrayFieldList.from_xarray(data, flavour)
 
-    if match_all_dates:
-        match = fs.sel(**kwargs)
-        result = []
-        for date in dates:
-            result.append(FieldArray([NewMetadataField(f, valid_datetime=date) for f in match]))
-        result = MultiFieldList(result)
+    if len(dates) == 0:
+        return fs.sel(**kwargs)
     else:
         result = MultiFieldList([fs.sel(valid_datetime=date, **kwargs) for date in dates])