fix: Ensure SG/CS get passed to the .cif file and fix structure conversion

Author: Linked-Liszt

compose.yaml

@@ -6,7 +6,7 @@ services:
     env_file:
       - .env
     volumes:
-      - ./data/raw_cif:/data/raw_cif
+      - ./data/:/data/
       - ./configs:/configs:ro
       - ./src/downloader:/app:ro
     user: "${UID}:${GID}"

src/downloader/downloader.py

@@ -31,32 +31,66 @@ def get_api_key() -> str:
     return api_key
 
 
-def _write_cif_from_struct(structure: Structure, out_path: Path) -> None:
-    """Write structure to CIF using pymatgen's CifWriter."""
+def _write_cif_from_struct(
+    structure: Structure,
+    out_path: Path,
+    sg_num: Optional[int] = None,
+    sg_symbol: Optional[str] = None,
+    symprec: float = 0.01,
+    angle_tolerance: float = 5.0,
+) -> None:
+    """
+    Write structure to CIF using pymatgen's CifWriter, ensuring symmetry metadata (_symmetry_* tags)
+    are present and reflect the detected space group from SpacegroupAnalyzer.
+    """
     try:
-        CifWriter(structure).write_file(out_path)
+        # Write CIF with symmetry detection enabled
+        CifWriter(structure, symprec=symprec, angle_tolerance=angle_tolerance).write_file(out_path)
     except Exception as e:
         raise RuntimeError(f"Failed to write CIF to {out_path}: {e}")
 
+    # Post-process CIF to force writing of symmetry tags
+    try:
+        with open(out_path, "r", errors="ignore") as f:
+            lines = f.readlines()
+
+        def set_or_insert(tag: str, value: Any, quote: bool = False) -> None:
+            val_str = f"'{value}'" if quote else str(value)
+            for i, line in enumerate(lines):
+                if line.strip().startswith(tag):
+                    lines[i] = f"{tag}   {val_str}\n"
+                    break
+            else:
+                # Insert after the first line (typically the data_ line)
+                insert_idx = 1 if len(lines) >= 1 else 0
+                lines.insert(insert_idx, f"{tag}   {val_str}\n")
+
+        if sg_num is not None:
+            set_or_insert("_symmetry_Int_Tables_number", sg_num, quote=False)
+        if sg_symbol:
+            set_or_insert("_symmetry_space_group_name_H-M", sg_symbol, quote=True)
+
+        with open(out_path, "w") as f:
+            f.writelines(lines)
+    except Exception:
+        # Non-fatal if symmetry tag insertion fails; CIF already written
+        pass
+
 
 def niggli_reduce(
     structure: Structure,
     symprec: float = 0.01,
     angle_tolerance: float = 5.0,
 ) -> Structure:
     """
-    Build Niggli-reduced lattice from the conventional standard structure.
+    Build the Niggli-reduced structure using pymatgen's canonical workflow:
+      1) Convert to conventional standard cell via Structure.to_conventional(...)
+      2) Apply Niggli reduction via Structure.get_reduced_structure(reduction_algo='niggli')
+    This ensures a proper change-of-basis for atomic coordinates, avoiding lattice-only swaps.
     """
-    sga = SpacegroupAnalyzer(structure, symprec=symprec, angle_tolerance=angle_tolerance)
-    conv = sga.get_conventional_standard_structure()
-    red_lat = conv.lattice.get_niggli_reduced_lattice()
-    reduced = Structure(
-        lattice=red_lat,
-        species=conv.species,
-        coords=conv.frac_coords,
-        to_unit_cell=True,
-    )
-    return reduced
+    conventional = structure.to_conventional(symprec=symprec, angle_tolerance=angle_tolerance)
+    reduced_structure = conventional.get_reduced_structure(reduction_algo="niggli")
+    return reduced_structure
 
 
 def passes_filters(
@@ -310,7 +344,14 @@ def download_cifs_entire_db(
                         summary["skipped_existing"] += 1
                         continue
 
-                    _write_cif_from_struct(std_struct, out_path)
+                    _write_cif_from_struct(
+                        std_struct,
+                        out_path,
+                        sg_num=sg_num,
+                        sg_symbol=sga_conv.get_space_group_symbol(),
+                        symprec=symprec,
+                        angle_tolerance=angle_tolerance,
+                    )
                     summary["downloaded"] += 1
 
                 except Exception as e: