Misc:

Basic registry, with survey stores
Move to standard src/ dir
versions: sqlmodel official, pydantic v2
etc...

src/gisaf/models/models_base.py

@@ -0,0 +1,118 @@
+from typing import Any
+import logging
+
+from sqlmodel import Field, SQLModel, MetaData, JSON, TEXT, Relationship, Column
+from pydantic import computed_field
+import numpy as np
+import pandas as pd
+import geopandas as gpd
+import shapely
+from sqlalchemy.sql import sqltypes
+from geoalchemy2.types import Geometry
+
+pandas_cast_map = {
+    sqltypes.Integer: 'Int64',
+    sqltypes.Float: 'float64',
+}
+
+logger = logging.getLogger('model_base_base')
+
+class Model(SQLModel):
+    """
+    Base mixin class for models that can be converted to a Pandas dataframe with get_df
+    """
+
+    class Meta:
+        filtered_columns_on_map: list[str] = []
+
+    @classmethod
+    def get_store_name(cls):
+        return "{}.{}".format(cls.metadata.schema, cls.__tablename__)
+
+    @classmethod
+    def get_table_name_prefix(cls):
+        return "{}_{}".format(cls.metadata.schema, cls.__tablename__)
+
+    @classmethod
+    async def get_df(cls, where=None,
+                     with_related=None, recursive=True,
+                     cast=True,
+                     with_only_columns=None,
+                     geom_as_ewkt=False,
+                     **kwargs):
+        """
+        Return a Pandas dataframe of all records
+        Optional arguments:
+        * an SQLAlchemy where clause
+        * with_related: automatically get data from related columns, following the foreign keys in the model definitions
+        * cast: automatically transform various data in their best python types (eg. with date, time...)
+        * with_only_columns: fetch only these columns (list of column names)
+        * geom_as_ewkt: convert geometry columns to EWKB (handy for eg. using upsert_df)
+        :return:
+        """
+        query = cls.query
+
+        if with_related is not False:
+            if with_related or getattr(cls, 'get_gdf_with_related', False):
+                joins = get_join_with(cls, recursive)
+                model_loader = cls.load(**joins)
+                query = _get_query_with_table_names(model_loader)
+
+        if where is not None:
+            query.append_whereclause(where)
+
+        if with_only_columns:
+            query = query.with_only_columns([getattr(cls, colname) for colname in with_only_columns])
+
+        ## Got idea from https://github.com/MagicStack/asyncpg/issues/173.
+        async with query.bind.raw_pool.acquire() as conn:
+            ## Convert hstore fields to dict
+            await conn.set_builtin_type_codec('hstore', codec_name='pg_contrib.hstore')
+
+            compiled = query.compile()
+            stmt = await conn.prepare(compiled.string)
+            columns = [a.name for a in stmt.get_attributes()]
+            data = await stmt.fetch(*[compiled.params.get(param) for param in compiled.positiontup])
+            df = pd.DataFrame(data, columns=columns)
+
+            ## Convert primary key columns to Int64:
+            ## allows NaN, fixing type convertion to float with merge
+            for pk in [c.name for c in cls.__table__.primary_key.columns]:
+                if pk in df.columns and df[pk].dtype=='int64':
+                    df[pk] = df[pk].astype('Int64')
+
+            if cast:
+                ## Cast the type for known types (datetime, ...)
+                for column_name in df.columns:
+                    col = getattr(query.columns, column_name, None)
+                    if col is None:
+                        logger.debug(f'Cannot get column {column_name} in query for model {cls.__name__}')
+                        continue
+                    column_type = getattr(query.columns, column_name).type
+                    ## XXX: Needs refinment, eg. nullable -> Int64 ...
+                    if column_type.__class__ in pandas_cast_map:
+                        df[column_name] = df[column_name].astype(pandas_cast_map[column_type.__class__])
+                    elif isinstance(column_type, (sqltypes.Date, sqltypes.DateTime)):
+                        ## Dates, times
+                        df[column_name] = pd.to_datetime(df[column_name])
+                    #elif isinstance(column_type, (sqltypes.Integer, sqltypes.Float)):
+                    #    ## Numeric
+                    #    df[column_name] = pd.to_numeric(df[column_name], errors='coerce')
+                    ## XXX: keeping this note about that is about "char" SQL type, but the fix of #9694 makes it unnessary
+                    #elif isinstance(column_type, sqltypes.CHAR) or (isinstance(column_type, sqltypes.String) and column_type.length == 1):
+                    #    ## Workaround for bytes being used for string of length 1 (not sure - why???)
+                    #    df[column_name] = df[column_name].str.decode('utf-8')
+
+            ## Rename the columns, removing the schema_table prefix for the columns in that model
+            prefix = cls.get_table_name_prefix()
+            prefix_length = len(prefix) + 1
+            rename_map = {colname: colname[prefix_length:] for colname in df.columns if colname.startswith(prefix)}
+            df.rename(columns=rename_map, inplace=True)
+
+            ## Eventually convert geometry columns to EWKB
+            if geom_as_ewkt:
+                geometry_columns = [col.name for col in cls.__table__.columns if isinstance(col.type, Geometry)]
+                for column in geometry_columns:
+                    df[column] = shapely.to_wkb(shapely.from_wkb(df.geom), hex=True, include_srid=True)
+
+            return df