Add endpoints for extracting the data needed by CSV_EXPORT.

src/ert_storage/endpoints/__init__.py

@@ -6,6 +6,7 @@
 from .updates import router as updates_router
 from .compute.misfits import router as misfits_router
 from .responses import router as response_router
+from .exports import router as exports_router
 
 router = APIRouter()
 router.include_router(experiments_router)
@@ -15,3 +16,4 @@
 router.include_router(updates_router)
 router.include_router(misfits_router)
 router.include_router(response_router)
+router.include_router(exports_router)

src/ert_storage/endpoints/exports.py

@@ -0,0 +1,91 @@
+from uuid import uuid4, UUID
+from typing import Dict, List, Optional
+
+from fastapi import APIRouter, Depends, Query, status
+from fastapi.responses import Response
+import pandas as pd
+from pandas.core.frame import DataFrame
+from sqlalchemy import and_, or_
+from ert_storage.database import Session, get_db
+from ert_storage import database_schema as ds
+
+
+router = APIRouter(tags=["exports"])
+
+
+@router.get(
+    "/ensembles/{ensemble_id}/export/csv",
+    responses={
+        status.HTTP_200_OK: {
+            "content": {"text/csv": {}},
+            "description": "Exports emsemble responses as csv",
+        }
+    },
+)
+async def get_eclipse_summary_vectors(
+    *,
+    db: Session = Depends(get_db),
+    ensemble_id: UUID,
+    column_list: Optional[List[str]] = Query(None)
+) -> Response:
+    """
+    Export responses for an ensemble.
+    """
+
+    ensemble = db.query(ds.Ensemble).filter_by(id=ensemble_id).one()
+
+    # Base record info filter
+    filter = [
+        ds.RecordInfo.ensemble_pk == ensemble.pk,
+        ds.RecordInfo.record_class == ds.RecordClass.response,
+    ]
+
+    # Add column name filter
+    if column_list is not None:
+        # Maybe only use LIKE when wildcards are actually present
+        # And use = for explicit names to make resulting query more efficent
+        filter_group = [
+            ds.RecordInfo.name.like(x.replace("*", "%").replace("?", "_"))
+            for x in column_list
+        ]
+
+        filter.append(or_(*filter_group))
+
+    records = (
+        db.query(ds.Record)
+        .filter(ds.Record.realization_index != None)
+        .join(ds.RecordInfo)
+        .filter(and_(*filter))
+    ).all()
+
+    if len(records) == 0:
+        return Response(content="No data found", status_code=status.HTTP_404_NOT_FOUND)
+
+    # Flatten data into required shape
+    # May be more efficent to do this as part of query
+
+    # Keep track of columns so they can be appended
+    column_map: Dict[str, DataFrame] = {}
+
+    for record in records:
+        labels = record.f64_matrix.labels
+        data = {
+            "REAL": record.realization_index,
+            "DATE": labels[0],
+            record.record_info.name: record.f64_matrix.content[0],
+        }
+        data_frame = pd.DataFrame(data)
+        column_name = record.record_info.name
+        if column_name in column_map:
+            column_map[column_name] = column_map[column_name].append(
+                data_frame, ignore_index=True
+            )
+        else:
+            column_map[column_name] = data_frame
+
+    # Index
+    data_frame_list = [df.set_index(["REAL", "DATE"]) for df in column_map.values()]
+    return Response(
+        content=pd.concat(data_frame_list, axis=1, join="outer").to_csv(index=True),
+        media_type="text/csv",
+    )