fix(dpdata): use public format APIs

Avoid importing dpdata backend parser modules directly. Use dpdata's registered public format interface for ABACUS STRU, VASP POSCAR, and SIESTA output handling so DP-GEN works with both dpdata 0.2.x and 1.x without depending on moved backend module paths.

Authored by OpenClaw (model: gpt-5.5)

Author: njzjz-bot

dpgen/auto_test/lib/abacus.py

@@ -4,9 +4,6 @@
 
 import dpdata
 import numpy as np
-from dpdata.abacus.stru import make_unlabeled_stru
-from dpdata.utils import uniq_atom_names
-from dpdata.vasp import poscar as dpdata_poscar
 
 import dpgen.generator.lib.abacus_scf as abacus_scf
 
@@ -142,11 +139,7 @@ def poscar2stru(poscar, inter_param, stru="STRU"):
                             - deepks_desc:  a string of deepks descriptor file
     - stru:            output filename, usally is 'STRU'.
     """
-    # if use dpdata.System, the structure will be rotated to make cell to be lower triangular
-    with open(poscar) as fp:
-        lines = [line.rstrip("\n") for line in fp]
-    stru_data = dpdata_poscar.to_system_data(lines)
-    stru_data = uniq_atom_names(stru_data)
+    stru_data = dpdata.System(poscar, fmt="vasp/poscar").data
 
     atom_mass = []
     pseudo = None
@@ -185,16 +178,15 @@ def poscar2stru(poscar, inter_param, stru="STRU"):
     if "deepks_desc" in inter_param:
         deepks_desc = "./pp_orb/{}\n".format(inter_param["deepks_desc"])
 
-    stru_string = make_unlabeled_stru(
-        data=stru_data,
+    dpdata.System(data=stru_data).to(
+        "abacus/stru",
+        stru,
         frame_idx=0,
         pp_file=pseudo,
         numerical_orbital=orb,
         numerical_descriptor=deepks_desc,
         mass=atom_mass,
     )
-    with open(stru, "w") as fp:
-        fp.write(stru_string)
 
 
 def stru_fix_atom(struf, fix_atom=[True, True, True]):

dpgen/generator/lib/abacus_scf.py

@@ -1,9 +1,10 @@
 import copy
 import os
 import re
+import tempfile
 
+import dpdata
 import numpy as np
-from dpdata.abacus.stru import get_frame_from_stru, make_unlabeled_stru
 
 from dpgen.auto_test.lib import vasp
 
@@ -259,13 +260,19 @@ def make_abacus_scf_stru(
     if len(cells.shape) == 2:
         sys_data_copy["cells"] = np.array([cells])
         sys_data_copy["coords"] = np.array([coords])
-    c = make_unlabeled_stru(
-        sys_data_copy,
-        0,
-        pp_file=fp_pp_files,
-        numerical_orbital=fp_orb_files,
-        numerical_descriptor=fp_dpks_descriptor,
-    )
+    with tempfile.NamedTemporaryFile(
+        mode="r+", prefix="dpgen-abacus-", suffix=".stru"
+    ) as fp:
+        dpdata.System(data=sys_data_copy).to(
+            "abacus/stru",
+            fp.name,
+            frame_idx=0,
+            pp_file=fp_pp_files,
+            numerical_orbital=fp_orb_files,
+            numerical_descriptor=fp_dpks_descriptor,
+        )
+        fp.seek(0)
+        c = fp.read()
 
     return c
 
@@ -302,7 +309,7 @@ def get_abacus_STRU(STRU):
         "dpks_descriptor": str,
     }
     """
-    data = get_frame_from_stru(STRU)
+    data = dpdata.System(STRU, fmt="abacus/stru").data
     data["atom_masses"] = data.pop("masses")
     data["cells"] = data.pop("cells")[0]
     data["coords"] = data.pop("coords")[0]

dpgen/generator/run.py

@@ -4585,9 +4585,10 @@ def post_fp_abacus_scf(iter_index, jdata):
 
         all_sys = None
         for ii, oo in zip(sys_input, sys_output):
-            _sys = dpdata.LabeledSystem(
-                oo, fmt="abacus/scf", type_map=jdata["type_map"]
-            )
+            _sys = dpdata.LabeledSystem(oo, fmt="abacus/scf")
+            if len(_sys) > 0:
+                _sys.data["atom_types"] = np.asarray(_sys.data["atom_types"], dtype=int)
+                _sys.apply_type_map(jdata["type_map"])
             if len(_sys) > 0:
                 if all_sys is None:
                     all_sys = _sys
@@ -4626,17 +4627,7 @@ def post_fp_siesta(iter_index, jdata):
         sys_output.sort()
         sys_input.sort()
         for idx, oo in enumerate(sys_output):
-            _sys = dpdata.LabeledSystem()
-            (
-                _sys.data["atom_names"],
-                _sys.data["atom_numbs"],
-                _sys.data["atom_types"],
-                _sys.data["cells"],
-                _sys.data["coords"],
-                _sys.data["energies"],
-                _sys.data["forces"],
-                _sys.data["virials"],
-            ) = dpdata.siesta.output.obtain_frame(oo)
+            _sys = dpdata.LabeledSystem(oo, fmt="siesta/output")
             if idx == 0:
                 all_sys = _sys
             else: