sawnergy 1.0.2__tar.gz → 1.0.4__tar.gz

{sawnergy-1.0.2/sawnergy.egg-info → sawnergy-1.0.4}/PKG-INFO

@@ -1,10 +1,12 @@
 Metadata-Version: 2.4
 Name: sawnergy
-Version: 1.0.2
+Version: 1.0.4
 Summary: Toolkit for transforming molecular dynamics (MD) trajectories into rich graph representations
 Home-page: https://github.com/Yehor-Mishchyriak/SAWNERGY
 Author: Yehor Mishchyriak
 License: Apache-2.0
+Project-URL: Source, https://github.com/Yehor-Mishchyriak/SAWNERGY
+Project-URL: Tracker, https://github.com/Yehor-Mishchyriak/SAWNERGY/issues
 Classifier: Programming Language :: Python :: 3
 Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Operating System :: OS Independent
@@ -25,12 +27,17 @@ Dynamic: description-content-type
 Dynamic: home-page
 Dynamic: license
 Dynamic: license-file
+Dynamic: project-url
 Dynamic: requires-dist
 Dynamic: requires-python
 Dynamic: summary
 
 # SAWNERGY
 
+[![PyPI](https://img.shields.io/pypi/v/sawnergy)](https://pypi.org/project/sawnergy/)
+[![License](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/LICENSE)
+![Python](https://img.shields.io/badge/python-3.11%2B-blue)
+
 A toolkit for transforming molecular dynamics (MD) trajectories into rich graph representations, sampling
 random and self-avoiding walks, learning node embeddings, and visualising residue interaction networks (RINs). SAWNERGY
 keeps the full workflow — from `cpptraj` output to skip-gram embeddings (node2vec approach) — inside Python, backed by efficient Zarr-based archives and optional GPU acceleration.
@@ -72,8 +79,8 @@ node indexing, and RNG seeds stay consistent across the toolchain.
 ---
 
 ## Small visual example (constructed fully from trajectory and topology files)
-![rin](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/assets/rin.png)
-![embedding](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/assets/embedding.png)
+![RIN](https://raw.githubusercontent.com/Yehor-Mishchyriak/SAWNERGY/main/assets/rin.png)
+![Embedding](https://raw.githubusercontent.com/Yehor-Mishchyriak/SAWNERGY/main/assets/embedding.png)
 
 ---
 

{sawnergy-1.0.2 → sawnergy-1.0.4}/README.md

@@ -1,5 +1,9 @@
 # SAWNERGY
 
+[![PyPI](https://img.shields.io/pypi/v/sawnergy)](https://pypi.org/project/sawnergy/)
+[![License](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/LICENSE)
+![Python](https://img.shields.io/badge/python-3.11%2B-blue)
+
 A toolkit for transforming molecular dynamics (MD) trajectories into rich graph representations, sampling
 random and self-avoiding walks, learning node embeddings, and visualising residue interaction networks (RINs). SAWNERGY
 keeps the full workflow — from `cpptraj` output to skip-gram embeddings (node2vec approach) — inside Python, backed by efficient Zarr-based archives and optional GPU acceleration.
@@ -41,8 +45,8 @@ node indexing, and RNG seeds stay consistent across the toolchain.
 ---
 
 ## Small visual example (constructed fully from trajectory and topology files)
-![rin](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/assets/rin.png)
-![embedding](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/assets/embedding.png)
+![RIN](https://raw.githubusercontent.com/Yehor-Mishchyriak/SAWNERGY/main/assets/rin.png)
+![Embedding](https://raw.githubusercontent.com/Yehor-Mishchyriak/SAWNERGY/main/assets/embedding.png)
 
 ---
 

{sawnergy-1.0.2 → sawnergy-1.0.4}/sawnergy/rin/rin_builder.py

@@ -690,9 +690,9 @@ class RINBuilder:
 
         2. For each frame batch:
             
-            a) Run cpptraj `pairwise` on atoms → EMAP + VMAP → sum (atomic matrix).
+            a) Run cpptraj 'pairwise' on atoms → EMAP + VMAP → sum (atomic matrix).
             
-            b) Project atomic → residue with ``R = Pᵀ @ A @ P``.
+            b) Project atomic → residue with 'R = Pᵀ @ A @ P'.
             
             c) Post-process residue matrix:
                 split into (attractive, repulsive) channels,
@@ -700,7 +700,7 @@ class RINBuilder:
                 remove self-interactions,
                 symmetrize.
 
-            d. Optionally store **pre-normalized energies** (attractive or repulsive or both, depending on `include_<kind>`).
+            d. Optionally store **pre-normalized energies** (attractive or repulsive or both, depending on 'include_<kind>').
             
             e. Row-wise L1 normalize (directed transition probabilities) and store.
 

{sawnergy-1.0.2 → sawnergy-1.0.4}/sawnergy/walks/walker.py

@@ -63,7 +63,7 @@ class Walker:
 
         # Load numpy arrays from read-only storage
         with sawnergy_util.ArrayStorage(RIN_path, mode="r") as storage:
-            attr_name   = storage.get_attr("attractive_transitions_name")
+            attr_name = storage.get_attr("attractive_transitions_name")
             repuls_name = storage.get_attr("repulsive_transitions_name")
             attr_matrices  : np.ndarray | None = (
                 storage.read(attr_name, slice(None)) if attr_name is not None else None
@@ -133,6 +133,9 @@ class Walker:
             ) if repuls_matrices is not None else None
         )
 
+        self._attr_owner_pid   = os.getpid() if self.attr_matrices  is not None else None
+        self._repuls_owner_pid = os.getpid() if self.repuls_matrices is not None else None
+
         _logger.debug(
             "SharedNDArray created | attr name=%r; repuls name=%r",
             getattr(self.attr_matrices, "name", None),
@@ -159,38 +162,56 @@ class Walker:
 
     # explicit resource cleanup
     def close(self) -> None:
-        """Close shared-memory handles and (in main process) unlink segments.
+        """Release shared-memory resources used by this Walker.
+
+        This method:
+        - Closes local handles to the shared-memory backed arrays
+        (`self.attr_matrices`, `self.repuls_matrices`) in **the current process**.
+        - If the current process is the **creator** of a segment (its PID matches
+        `_attr_owner_pid` / `_repuls_owner_pid`), it also **unlinks** that segment
+        so the OS can reclaim it once all handles are closed.
+
+        Behavior & guarantees
+        ---------------------
+        - **Idempotent:** safe to call multiple times; subsequent calls are no-ops.
+        - **Multi-process aware:** non-creator processes only close their handles;
+        creators close **and** unlink. This prevents `resource_tracker` “leaked
+        shared_memory” warnings when using `ProcessPoolExecutor`/spawn.
+        - **Best-effort unlink:** `FileNotFoundError` during unlink (already unlinked
+        elsewhere) is swallowed.
+        - Invoked automatically by the context manager (`__exit__`) and destructor
+        (`__del__`), but it's fine to call explicitly.
+
+        After calling `close()`, any operation that relies on the shared arrays may
+        fail; treat the instance as finalized.
 
-        Idempotent: if cleanup already occurred, returns immediately. Always
-        closes local handles in the current process. If the caller is the main
-        process (per ``sawnergy_util.is_main_process()``), also attempts to
-        unlink the underlying shared-memory segments (best-effort; suppresses
-        ``FileNotFoundError`` if already unlinked elsewhere).
+        Returns:
+            None
         """
         if self._memory_cleaned_up:
             _logger.debug("close(): already cleaned up; returning")
             return
-        _logger.debug("Closing Walker resources (is_main=%s)", sawnergy_util.is_main_process())
+        _logger.debug("Closing Walker resources (pid=%s)", os.getpid())
         try:
             if self.attr_matrices is not None:
                 self.attr_matrices.close()
             if self.repuls_matrices is not None:
                 self.repuls_matrices.close()
             _logger.debug("SharedNDArray handles closed")
-            if sawnergy_util.is_main_process():
-                _logger.debug("Attempting to unlink shared memory segments (main process)")
-                try:
-                    if self.attr_matrices is not None:
-                        self.attr_matrices.unlink()
-                except FileNotFoundError:
-                    _logger.warning("attr SharedMemory already unlinked")
-                try:
-                    if self.repuls_matrices is not None:
-                        self.repuls_matrices.unlink()
-                except FileNotFoundError:
-                    _logger.warning("repuls SharedMemory already unlinked")
-            else:
-                _logger.debug("Not main process; skipping unlink")
+
+            # Unlink in whichever process actually CREATED the segment(s)
+            try:
+                if self.attr_matrices is not None and getattr(self, "_attr_owner_pid", None) == os.getpid():
+                    self.attr_matrices.unlink()
+            except FileNotFoundError:
+                _logger.debug("attr SharedMemory already unlinked elsewhere")
+
+            try:
+                if self.repuls_matrices is not None and getattr(self, "_repuls_owner_pid", None) == os.getpid():
+                    self.repuls_matrices.unlink()
+            except FileNotFoundError:
+                _logger.debug("repuls SharedMemory already unlinked elsewhere")
+
         finally:
             self._memory_cleaned_up = True
             _logger.debug("Cleanup complete")

{sawnergy-1.0.2 → sawnergy-1.0.4/sawnergy.egg-info}/PKG-INFO

@@ -1,10 +1,12 @@
 Metadata-Version: 2.4
 Name: sawnergy
-Version: 1.0.2
+Version: 1.0.4
 Summary: Toolkit for transforming molecular dynamics (MD) trajectories into rich graph representations
 Home-page: https://github.com/Yehor-Mishchyriak/SAWNERGY
 Author: Yehor Mishchyriak
 License: Apache-2.0
+Project-URL: Source, https://github.com/Yehor-Mishchyriak/SAWNERGY
+Project-URL: Tracker, https://github.com/Yehor-Mishchyriak/SAWNERGY/issues
 Classifier: Programming Language :: Python :: 3
 Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Operating System :: OS Independent
@@ -25,12 +27,17 @@ Dynamic: description-content-type
 Dynamic: home-page
 Dynamic: license
 Dynamic: license-file
+Dynamic: project-url
 Dynamic: requires-dist
 Dynamic: requires-python
 Dynamic: summary
 
 # SAWNERGY
 
+[![PyPI](https://img.shields.io/pypi/v/sawnergy)](https://pypi.org/project/sawnergy/)
+[![License](https://img.shields.io/badge/license-Apache--2.0-blue)](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/LICENSE)
+![Python](https://img.shields.io/badge/python-3.11%2B-blue)
+
 A toolkit for transforming molecular dynamics (MD) trajectories into rich graph representations, sampling
 random and self-avoiding walks, learning node embeddings, and visualising residue interaction networks (RINs). SAWNERGY
 keeps the full workflow — from `cpptraj` output to skip-gram embeddings (node2vec approach) — inside Python, backed by efficient Zarr-based archives and optional GPU acceleration.
@@ -72,8 +79,8 @@ node indexing, and RNG seeds stay consistent across the toolchain.
 ---
 
 ## Small visual example (constructed fully from trajectory and topology files)
-![rin](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/assets/rin.png)
-![embedding](https://github.com/Yehor-Mishchyriak/SAWNERGY/blob/main/assets/embedding.png)
+![RIN](https://raw.githubusercontent.com/Yehor-Mishchyriak/SAWNERGY/main/assets/rin.png)
+![Embedding](https://raw.githubusercontent.com/Yehor-Mishchyriak/SAWNERGY/main/assets/embedding.png)
 
 ---