pysdmx 1.5.2__py3-none-any.whl → 1.6.0__py3-none-any.whl

pysdmx/io/csv/sdmx20/writer/__init__.py

@@ -1,17 +1,21 @@
 """SDMX 2.0 CSV writer module."""
 
-from copy import copy
 from pathlib import Path
-from typing import Optional, Sequence, Union
+from typing import Literal, Optional, Sequence, Union
 
 import pandas as pd
 
-from pysdmx.io.csv.sdmx20 import SDMX_CSV_ACTION_MAPPER
+from pysdmx.io.csv.__csv_aux_writer import (
+    _write_csv_2_aux,
+)
 from pysdmx.io.pd import PandasDataset
 
 
 def write(
     datasets: Sequence[PandasDataset],
+    labels: Optional[Literal["name", "id", "both"]] = None,
+    time_format: Optional[Literal["original", "normalized"]] = None,
+    keys: Optional[Literal["obs", "series", "both"]] = None,
     output_path: Optional[Union[str, Path]] = None,
 ) -> Optional[str]:
     """Write data to SDMX-CSV 2.0 format.
@@ -19,6 +23,25 @@ def write(
     Args:
         datasets: List of datasets to write.
           Must have the same components.
+        labels: How to write the name of the columns.
+            If None, only the IDs are written.
+            if "id", the names are written as ID only.
+            if "name", a colum called "STRUCTURE_NAME" is
+            added after struture ID.
+            If "both", the names are witten as id:Name.
+        time_format: How to write the time period.
+            If None, the time period is not modified.
+            If "original", the time period is written as it
+            is in the dataset.
+            "normalized" is not implemented yet.
+        keys: to write or not the keys columns
+            If None, no keys are written.
+            If "obs", the keys are write as a single
+            column called "OBS_KEY".
+            If "series", the keys are write as a single
+            column called "SERIES_KEY".
+            If "both", the keys are write as two columns:
+            "OBS_KEY" and "SERIES_KEY".
         output_path: Path to write the data to.
           If None, the data is returned as a string.
 
@@ -28,28 +51,12 @@ def write(
     # Link to pandas.to_csv documentation on sphinx:
     # https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.to_csv.html
 
-    dataframes = []
-    for dataset in datasets:
-        # Create a copy of the dataset
-        df: pd.DataFrame = copy(dataset.data)
-
-        # Add additional attributes to the dataset
-        for k, v in dataset.attributes.items():
-            df[k] = v
-
-        structure_ref, unique_id = dataset.short_urn.split("=", maxsplit=1)
-        if structure_ref in ["DataStructure", "Dataflow"]:
-            structure_ref = structure_ref.lower()
-        else:
-            structure_ref = "dataprovision"
-
-        # Insert two columns at the beginning of the data set
-        df.insert(0, "STRUCTURE", structure_ref)
-        df.insert(1, "STRUCTURE_ID", unique_id)
-        action_value = SDMX_CSV_ACTION_MAPPER[dataset.action]
-        df.insert(2, "ACTION", action_value)
-
-        dataframes.append(df)
+    dataframes = _write_csv_2_aux(
+        datasets,
+        labels,
+        time_format,
+        keys,
+    )
 
     all_data = pd.concat(dataframes, ignore_index=True, axis=0)
 

pysdmx/io/csv/sdmx21/__init__.py

@@ -0,0 +1 @@
+"""SDMX 2.1 CSV reader and writer."""

pysdmx/io/csv/sdmx21/reader/__init__.py

@@ -0,0 +1,86 @@
+"""SDMX 2.1 CSV reader module."""
+
+from io import StringIO
+from typing import Sequence
+
+import pandas as pd
+
+from pysdmx.errors import Invalid
+from pysdmx.io.csv.__csv_aux_reader import __generate_dataset_from_sdmx_csv
+from pysdmx.io.pd import PandasDataset
+
+
+def read(input_str: str) -> Sequence[PandasDataset]:
+    """Reads csv data and returns a sequence of Datasets.
+
+    Args:
+        input_str: str.
+
+    Returns:
+        A Sequence of Pandas Datasets.
+
+    Raises:
+        Invalid: If it is an invalid CSV file.
+    """
+    # Get Dataframe from CSV file
+    df_csv = pd.read_csv(
+        StringIO(input_str), keep_default_na=False, na_values=[""]
+    )
+    # Drop empty columns
+    df_csv = df_csv.dropna(axis=1, how="all")
+
+    # Determine SDMX-CSV version based on column names
+    if (
+        "STRUCTURE" not in df_csv.columns
+        or "STRUCTURE_ID" not in df_csv.columns
+    ):
+        # Raise an exception if the CSV file is not in SDMX-CSV format
+        raise Invalid(
+            "Only SDMX-CSV 2.1 is allowed",
+            "Invalid SDMX-CSV 2.1 file. "
+            "Check the docs for the proper structure on content.",
+        )
+
+    # Convert all columns to strings
+    df_csv = df_csv.astype(str).replace({"nan": "", "<NA>": ""})
+    # Check if any column headers contain ':', indicating mode, label or text
+    mode_label_text = any(":" in x for x in df_csv.columns)
+    # if values in the columns contain ':', split them
+    for col in df_csv.columns[2:]:
+        df_csv[col] = (
+            df_csv[col]
+            .astype(str)
+            .apply(lambda x: x.split(":")[0] if ":" in x else x)
+        )
+
+    id_column = "STRUCTURE_ID"
+    # If mode, label or text is present, modify the DataFrame
+    if mode_label_text:
+        # Split the ID column to remove mode, label or text
+        df_csv[id_column] = df_csv[id_column].map(lambda x: x.split(": ")[0])
+        # Split the other columns to remove mode, label, or text
+        sequence = 3
+        for x in df_csv.columns[sequence:]:
+            df_csv[x.split(":")[0]] = df_csv[x].map(
+                lambda x: x.split(": ", 2)[0], na_action="ignore"
+            )
+            # Delete the original columns
+            del df_csv[x]
+
+    # Grouping columns to separate datasets
+    grouping_columns = ["STRUCTURE", "STRUCTURE_ID"]
+    # Separate SDMX-CSV in different datasets per Structure ID
+    list_df = [data for _, data in df_csv.groupby(grouping_columns)]
+
+    # Create a payload dictionary to store datasets with the
+    # different unique_ids as keys
+    payload = []
+    for df in list_df:
+        # Generate a dataset from each subset of the DataFrame
+        dataset = __generate_dataset_from_sdmx_csv(data=df, references_21=True)
+
+        # Add the dataset to the payload dictionary
+        payload.append(dataset)
+
+    # Return the payload generated
+    return payload

pysdmx/io/csv/sdmx21/writer/__init__.py

@@ -0,0 +1,70 @@
+"""SDMX 2.1 CSV writer module."""
+
+from pathlib import Path
+from typing import Literal, Optional, Sequence, Union
+
+import pandas as pd
+
+from pysdmx.io.csv.__csv_aux_writer import (
+    _write_csv_2_aux,
+)
+from pysdmx.io.pd import PandasDataset
+
+
+def write(
+    datasets: Sequence[PandasDataset],
+    labels: Optional[Literal["name", "id", "both"]] = None,
+    time_format: Optional[Literal["original", "normalized"]] = None,
+    keys: Optional[Literal["obs", "series", "both"]] = None,
+    output_path: Optional[Union[str, Path]] = None,
+) -> Optional[str]:
+    """Write data to SDMX-CSV 2.1 format.
+
+    Args:
+        datasets: List of datasets to write.
+          Must have the same components.
+        labels: How to write the name of the columns.
+            If None, only the IDs are written.
+            if "id", the names are written as ID only.
+            if "name", a colum called "STRUCTURE_NAME" is
+            added after struture ID.
+            If "both", the names are witten as id:Name.
+        time_format: How to write the time period.
+            If None, the time period is not modified.
+            If "original", the time period is written as it
+            is in the dataset.
+            "normalized" is not implemented yet.
+        keys: to write or not the keys columns
+            If None, no keys are written.
+            If "obs", the keys are write as a single
+            column called "OBS_KEY".
+            If "series", the keys are write as a single
+            column called "SERIES_KEY".
+            If "both", the keys are write as two columns:
+            "OBS_KEY" and "SERIES_KEY".
+        output_path: Path to write the data to.
+          If None, the data is returned as a string.
+
+    Returns:
+        SDMX CSV data as a string, if output_path is None.
+    """
+    # Link to pandas.to_csv documentation on sphinx:
+    # https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.to_csv.html
+
+    dataframes = _write_csv_2_aux(
+        datasets, labels, time_format, keys, references_21=True
+    )
+
+    all_data = pd.concat(dataframes, ignore_index=True, axis=0)
+
+    all_data = all_data.astype(str).replace({"nan": "", "<NA>": ""})
+
+    # If the output path is an empty string we use None
+    output_path = (
+        None
+        if isinstance(output_path, str) and output_path == ""
+        else output_path
+    )
+
+    # Convert the dataset into a csv file
+    return all_data.to_csv(output_path, index=False, header=True)

pysdmx/io/format.py

@@ -10,8 +10,10 @@ class Format(Enum):
 
     DATA_SDMX_CSV_1_0_0 = f"{_BASE}data+csv;version=1.0.0"
     DATA_SDMX_CSV_2_0_0 = f"{_BASE}data+csv;version=2.0.0"
+    DATA_SDMX_CSV_2_1_0 = f"{_BASE}data+csv;version=2.1.0"
     DATA_SDMX_JSON_1_0_0 = f"{_BASE}data+json;version=1.0.0"
     DATA_SDMX_JSON_2_0_0 = f"{_BASE}data+json;version=2.0.0"
+    DATA_SDMX_JSON_2_1_0 = f"{_BASE}data+json;version=2.1.0"
     DATA_SDMX_ML_2_1_GEN = f"{_BASE}genericdata+xml;version=2.1"
     DATA_SDMX_ML_2_1_STR = f"{_BASE}structurespecificdata+xml;version=2.1"
     DATA_SDMX_ML_2_1_GENTS = f"{_BASE}generictimeseriesdata+xml;version=2.1"
@@ -22,7 +24,9 @@ class Format(Enum):
     DATA_SDMX_ML_3_1 = f"{_BASE}data+xml;version=3.1.0"
     GDS_JSON = "application/json"
     REFMETA_SDMX_CSV_2_0_0 = f"{_BASE}metadata+csv;version=2.0.0"
+    REFMETA_SDMX_CSV_2_1_0 = f"{_BASE}metadata+csv;version=2.1.0"
     REFMETA_SDMX_JSON_2_0_0 = f"{_BASE}metadata+json;version=2.0.0"
+    REFMETA_SDMX_JSON_2_1_0 = f"{_BASE}metadata+json;version=2.1.0"
     REFMETA_SDMX_ML_3_0 = f"{_BASE}metadata+xml;version=3.0.0"
     REFMETA_SDMX_ML_3_1 = f"{_BASE}metadata+xml;version=3.1.0"
     REGISTRY_SDMX_ML_2_1 = f"{_BASE}registry+xml;version=2.1"
@@ -30,11 +34,13 @@ class Format(Enum):
     REGISTRY_SDMX_ML_3_1 = f"{_BASE}registry+xml;version=3.1.0"
     SCHEMA_SDMX_JSON_1_0_0 = f"{_BASE}schema+json;version=1.0.0"
     SCHEMA_SDMX_JSON_2_0_0 = f"{_BASE}schema+json;version=2.0.0"
+    SCHEMA_SDMX_JSON_2_1_0 = f"{_BASE}schema+json;version=2.1.0"
     SCHEMA_SDMX_ML_2_1 = f"{_BASE}schema+xml;version=2.1"
     SCHEMA_SDMX_ML_3_0 = f"{_BASE}schema+xml;version=3.0.0"
     SCHEMA_SDMX_ML_3_1 = f"{_BASE}schema+xml;version=3.1.0"
     STRUCTURE_SDMX_JSON_1_0_0 = f"{_BASE}structure+json;version=1.0.0"
     STRUCTURE_SDMX_JSON_2_0_0 = f"{_BASE}structure+json;version=2.0.0"
+    STRUCTURE_SDMX_JSON_2_1_0 = f"{_BASE}structure+json;version=2.1.0"
     STRUCTURE_SDMX_ML_2_1 = f"{_BASE}structure+xml;version=2.1"
     STRUCTURE_SDMX_ML_3_0 = f"{_BASE}structure+xml;version=3.0.0"
     STRUCTURE_SDMX_ML_3_1 = f"{_BASE}structure+xml;version=3.1.0"
@@ -57,6 +63,7 @@ class DataFormat(Enum):
 
     SDMX_CSV_1_0_0 = Format.DATA_SDMX_CSV_1_0_0.value
     SDMX_CSV_2_0_0 = Format.DATA_SDMX_CSV_2_0_0.value
+    SDMX_CSV_2_1_0 = Format.DATA_SDMX_CSV_2_1_0.value
     SDMX_JSON_1_0_0 = Format.DATA_SDMX_JSON_1_0_0.value
     SDMX_JSON_2_0_0 = Format.DATA_SDMX_JSON_2_0_0.value
     SDMX_ML_2_1_GEN = Format.DATA_SDMX_ML_2_1_GEN.value
@@ -71,6 +78,7 @@ class RefMetaFormat(Enum):
     """The SDMX Reference Metadata formats."""
 
     SDMX_CSV_2_0_0 = Format.REFMETA_SDMX_CSV_2_0_0.value
+    SDMX_CSV_2_1_0 = Format.REFMETA_SDMX_CSV_2_1_0.value
     SDMX_JSON_2_0_0 = Format.REFMETA_SDMX_JSON_2_0_0.value
     SDMX_ML_3_0 = Format.REFMETA_SDMX_ML_3_0.value
     SDMX_ML_3_1 = Format.REFMETA_SDMX_ML_3_1.value

pysdmx/io/input_processor.py

@@ -84,10 +84,24 @@ def __get_sdmx_csv_flavour(input_str: str) -> Tuple[str, Format]:
     if "DATAFLOW" in headers:
         return input_str, Format.DATA_SDMX_CSV_1_0_0
     elif "STRUCTURE" in headers and "STRUCTURE_ID" in headers:
-        return input_str, Format.DATA_SDMX_CSV_2_0_0
+        return input_str, Format.DATA_SDMX_CSV_2_1_0
     raise Invalid("Validation Error", "Cannot parse input as SDMX-CSV.")
 
 
+def __get_sdmx_json_flavour(input_str: str) -> Tuple[str, Format]:
+    flavour_check = input_str[:400].lower()
+    if "2.0.0/sdmx-json-structure-schema.json" in flavour_check:
+        return input_str, Format.STRUCTURE_SDMX_JSON_2_0_0
+    elif "2.0.0/sdmx-json-metadata-schema.json" in flavour_check:
+        return input_str, Format.REFMETA_SDMX_JSON_2_0_0
+    elif "sdmx-json" in flavour_check:
+        raise NotImplemented(
+            "Unsupported format", "This flavour of SDMX-JSON is not supported."
+        )
+    else:
+        raise Invalid("Validation Error", "Cannot parse input as SDMX-JSON.")
+
+
 def __check_sdmx_str(input_str: str) -> Tuple[str, Format]:
     """Attempts to infer the SDMX format of the input string."""
     if __check_xml(input_str):
@@ -95,7 +109,7 @@ def __check_sdmx_str(input_str: str) -> Tuple[str, Format]:
     if __check_csv(input_str):
         return __get_sdmx_csv_flavour(input_str)
     if __check_json(input_str):
-        raise NotImplemented("JSON formats reading are not supported yet")
+        return __get_sdmx_json_flavour(input_str)
     raise Invalid("Validation Error", "Cannot parse input as SDMX.")
 
 

pysdmx/io/json/fusion/messages/code.py

@@ -12,6 +12,7 @@ from pysdmx.io.json.fusion.messages.core import (
     FusionString,
 )
 from pysdmx.model import (
+    Annotation,
     Code,
     HierarchicalCode,
 )
@@ -31,6 +32,7 @@ class FusionCode(Struct, frozen=True):
     """Fusion-JSON payload for codes."""
 
     id: str
+    urn: Optional[str] = None
     annotations: Sequence[FusionAnnotation] = ()
     names: Sequence[FusionString] = ()
     descriptions: Sequence[FusionString] = ()
@@ -49,10 +51,11 @@ class FusionCode(Struct, frozen=True):
             valid_to = self.__handle_date(vals[1]) if vals[1] else None
             return (valid_from, valid_to)
 
-    def to_model(self) -> Code:
+    def to_model(self, extract_urn: bool = False) -> Code:
         """Converts a FusionCode to a standard code."""
         vp = [a for a in self.annotations if a.type == "FR_VALIDITY_PERIOD"]
         vf, vt = self.__get_val(vp[0]) if vp else (None, None)
+        urn = self.urn if extract_urn else None
         return Code(
             id=self.id,
             name=self.names[0].value,
@@ -61,6 +64,7 @@ class FusionCode(Struct, frozen=True):
             ),
             valid_from=vf,
             valid_to=vt,
+            urn=urn,
         )
 
 
@@ -75,7 +79,7 @@ class FusionCodelist(Struct, frozen=True, rename={"agency": "agencyId"}):
     version: str = "1.0"
     items: Sequence[FusionCode] = ()
 
-    def to_model(self) -> CL:
+    def to_model(self, extract_urns: bool = False) -> CL:
         """Converts a JsonCodelist to a standard codelist."""
         t = "codelist" if "Codelist" in self.urn else "valuelist"
         return CL(
@@ -86,7 +90,7 @@ class FusionCodelist(Struct, frozen=True, rename={"agency": "agencyId"}):
                 self.descriptions[0].value if self.descriptions else None
             ),
             version=self.version,
-            items=[i.to_model() for i in self.items],
+            items=[i.to_model(extract_urns) for i in self.items],
             sdmx_type=t,  # type: ignore[arg-type]
         )
 
@@ -108,6 +112,7 @@ class FusionCodelistMessage(Struct, frozen=True):
 class FusionHierarchicalCode(Struct, frozen=True):
     """Fusion-JSON payload for hierarchical codes."""
 
+    id: str
     code: str
     validFrom: Optional[int] = None
     validTo: Optional[int] = None
@@ -140,6 +145,16 @@ class FusionHierarchicalCode(Struct, frozen=True):
         rvf = self.__convert_epoch(self.validFrom) if self.validFrom else None
         rvt = self.__convert_epoch(self.validTo) if self.validTo else None
         codes = [c.to_model(codelists) for c in self.codes]
+        if self.id != code.id:
+            # The ID of the hierarchical code differs from the ID of the
+            # code that the hierarchical code references. We therefore need
+            # to store the ID of the hierarchical code, else the information
+            # will be lost and we won't be able to write the hierarchy back
+            # to the Registry, if requested.
+            a = Annotation(id="hcode", type="pysdmx", text=self.id)
+            annotations = [a]
+        else:
+            annotations = []
         return HierarchicalCode(
             code.id,
             code.name,
@@ -149,6 +164,8 @@ class FusionHierarchicalCode(Struct, frozen=True):
             rvf,
             rvt,
             codes,
+            tuple(annotations),
+            code.urn,
         )
 
 
@@ -222,7 +239,7 @@ class FusionHierarchyMessage(Struct, frozen=True):
 
     def to_model(self) -> HCL:
         """Returns the requested hierarchy."""
-        cls = [cl.to_model() for cl in self.Codelist]
+        cls = [cl.to_model(True) for cl in self.Codelist]
         return self.Hierarchy[0].to_model(cls)
 
 

pysdmx/io/json/fusion/messages/concept.py

@@ -2,7 +2,7 @@
 
 from typing import Optional, Sequence
 
-from msgspec import Struct
+import msgspec
 
 from pysdmx.io.json.fusion.messages.code import FusionCodelist
 from pysdmx.io.json.fusion.messages.core import (
@@ -13,7 +13,7 @@ from pysdmx.model.concept import Concept, DataType
 from pysdmx.model.concept import ConceptScheme as CS
 
 
-class FusionConcept(Struct, frozen=True):
+class FusionConcept(msgspec.Struct, frozen=True):
     """Fusion-JSON payload for concepts."""
 
     id: str
@@ -51,7 +51,9 @@ class FusionConcept(Struct, frozen=True):
         )
 
 
-class FusionConceptScheme(Struct, frozen=True, rename={"agency": "agencyId"}):
+class FusionConceptScheme(
+    msgspec.Struct, frozen=True, rename={"agency": "agencyId"}
+):
     """Fusion-JSON payload for a concept scheme."""
 
     id: str
@@ -61,23 +63,29 @@ class FusionConceptScheme(Struct, frozen=True, rename={"agency": "agencyId"}):
     version: str = "1.0"
     items: Sequence[FusionConcept] = ()
 
+    def __set_urn(self, concept: Concept) -> Concept:
+        urn = (
+            "urn:sdmx:org.sdmx.infomodel.conceptscheme.Concept="
+            f"{self.agency}:{self.id}({self.version}).{concept.id}"
+        )
+        return msgspec.structs.replace(concept, urn=urn)
+
     def to_model(self, codelists: Sequence[FusionCodelist]) -> CS:
         """Converts a FusionConceptScheme to a standard concept scheme."""
         d = self.descriptions[0].value if self.descriptions else None
+        concepts = [c.to_model(codelists) for c in self.items]
+        concepts = [self.__set_urn(c) for c in concepts]
         return CS(
             id=self.id,
             name=self.names[0].value,
             agency=self.agency,
             description=d,
             version=self.version,
-            items=[c.to_model(codelists) for c in self.items],
+            items=concepts,
         )
 
 
-class FusionConceptSchemeMessage(
-    Struct,
-    frozen=True,
-):
+class FusionConceptSchemeMessage(msgspec.Struct, frozen=True):
     """Fusion-JSON payload for /conceptscheme queries."""
 
     ConceptScheme: Sequence[FusionConceptScheme]

pysdmx/io/json/fusion/messages/dataflow.py

@@ -15,6 +15,7 @@ from pysdmx.model import (
 from pysdmx.model import (
     Dataflow as DF,
 )
+from pysdmx.model.dataflow import Group
 
 
 class FusionDataflow(Struct, frozen=True, rename={"agency": "agencyId"}):
@@ -62,7 +63,12 @@ class FusionDataflowMessage(Struct, frozen=True):
             return df.agency == agency and df.id == id_
 
     def to_model(
-        self, components: Components, agency: str, id_: str, version: str
+        self,
+        components: Components,
+        grps: Optional[Sequence[Group]],
+        agency: str,
+        id_: str,
+        version: str,
     ) -> DataflowInfo:
         """Returns the requested dataflow details."""
         prvs: List[DataProvider] = []
@@ -83,6 +89,7 @@ class FusionDataflowMessage(Struct, frozen=True):
             version=df.version,
             providers=prvs,
             dsd_ref=df.dataStructureRef,
+            groups=grps,
         )
 
 

pysdmx/io/json/fusion/messages/dsd.py

@@ -110,6 +110,10 @@ class FusionAttribute(Struct, frozen=True):
         )
         lvl = self.__derive_level(groups)
         desc = c.descriptions[0].value if c.descriptions else None
+        if self.representation and self.representation.representation:
+            local_enum_ref = self.representation.representation
+        else:
+            local_enum_ref = None
         return Component(
             id=self.id,
             required=self.mandatory,
@@ -122,6 +126,7 @@ class FusionAttribute(Struct, frozen=True):
             local_codes=codes,
             attachment_level=lvl,
             array_def=ab,
+            local_enum_ref=local_enum_ref,
         )
 
 
@@ -160,6 +165,10 @@ class FusionDimension(Struct, frozen=True):
             self.id, self.representation, cls, cons
         )
         desc = c.descriptions[0].value if c.descriptions else None
+        if self.representation and self.representation.representation:
+            local_enum_ref = self.representation.representation
+        else:
+            local_enum_ref = None
         return Component(
             id=self.id,
             required=True,
@@ -171,6 +180,7 @@ class FusionDimension(Struct, frozen=True):
             description=desc,
             local_codes=codes,
             array_def=ab,
+            local_enum_ref=local_enum_ref,
         )
 
 
@@ -209,6 +219,10 @@ class FusionMeasure(Struct, frozen=True):
             self.id, self.representation, cls, cons
         )
         desc = c.descriptions[0].value if c.descriptions else None
+        if self.representation and self.representation.representation:
+            local_enum_ref = self.representation.representation
+        else:
+            local_enum_ref = None
         return Component(
             id=self.id,
             required=self.mandatory,
@@ -220,6 +234,7 @@ class FusionMeasure(Struct, frozen=True):
             description=desc,
             local_codes=codes,
             array_def=ab,
+            local_enum_ref=local_enum_ref,
         )
 
 

pysdmx/io/json/fusion/messages/schema.py

@@ -10,6 +10,7 @@ from pysdmx.io.json.fusion.messages.constraint import FusionContentConstraint
 from pysdmx.io.json.fusion.messages.core import FusionLink
 from pysdmx.io.json.fusion.messages.dsd import FusionDataStructure
 from pysdmx.model import Components, HierarchyAssociation, Schema
+from pysdmx.model.dataflow import Group
 from pysdmx.util import parse_item_urn
 
 
@@ -44,6 +45,7 @@ class FusionSchemaMessage(msgspec.Struct, frozen=True):
         components = self.DataStructure[0].get_components(
             self.ConceptScheme, cls, self.DataConstraint
         )
+        grps = self.DataStructure[0].groups
         comp_dict = {c.id: c for c in components}
         urns = [a.urn for a in self.meta.links]
         for ha in hierarchies:
@@ -60,4 +62,9 @@ class FusionSchemaMessage(msgspec.Struct, frozen=True):
                 f"{h.agency}:{h.id}({h.version})"  # type: ignore[union-attr]
             )
         comps = Components(comp_dict.values())
-        return Schema(context, agency, id_, comps, version, urns)
+        mapped_grps = [
+            Group(g.id, dimensions=g.dimensionReferences) for g in grps
+        ]
+        return Schema(
+            context, agency, id_, comps, version, urns, groups=mapped_grps
+        )