supertable 2.2.3__tar.gz → 2.2.4__tar.gz

{supertable-2.2.3/supertable.egg-info → supertable-2.2.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: supertable
-Version: 2.2.3
+Version: 2.2.4
 Summary: SuperTable — versioned data lake library for SQL analytics on Parquet + Redis.
 Author: Levente Kupas
 Author-email: Levente Kupas <lkupas@kladnasoft.com>

{supertable-2.2.3 → supertable-2.2.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "supertable"
-version = "2.2.3"
+version = "2.2.4"
 description = "SuperTable — versioned data lake library for SQL analytics on Parquet + Redis."
 readme = "README.md"
 requires-python = ">=3.10"

{supertable-2.2.3 → supertable-2.2.4}/setup.py

@@ -19,7 +19,7 @@ long_description = readme.read_text(encoding="utf-8") if readme.exists() else ""
 
 setup(
     name="supertable",
-    version="2.2.3",
+    version="2.2.4",
     description="SuperTable — versioned data lake library for SQL analytics on Parquet + Redis.",
     long_description=long_description,
     long_description_content_type="text/markdown",

{supertable-2.2.3 → supertable-2.2.4}/supertable/__init__.py

@@ -25,7 +25,7 @@ See the ``supertable.demo`` package for runnable end-to-end demos and the
 project documentation for the full API surface.
 """
 
-__version__ = "2.2.3"
+__version__ = "2.2.4"
 
 # Re-export the core public surface so users can do ``from supertable import …``
 # instead of remembering submodule paths.

{supertable-2.2.3 → supertable-2.2.4}/supertable/data_writer.py

@@ -754,9 +754,14 @@ class DataWriter:
              qualify — large files are left untouched. ``small_only=False``
              rewrites every resource regardless of size.
           5. Commit the new snapshot via ``simple_table.update``,
-             leaf-CAS, ``bump_root``, and the same GC enqueue +
+             leaf-CAS, ``bump_root``, optional mirroring, and the same
              monitoring + audit pipeline as ``write``.
 
+        Compaction does **not** garbage-collect: the small files it
+        merges away are left in storage (still referenced by older
+        snapshot versions for time-travel) for an external GC to
+        reclaim later.
+
         Concurrency: compaction is serialised against writes through
         the same per-simple lock. A concurrent ``write()`` either
         runs first (compaction sees the updated snapshot) or waits.

{supertable-2.2.3 → supertable-2.2.4}/supertable/simple_table.py

@@ -260,6 +260,56 @@ class SimpleTable:
         data = self.storage.read_json(path)
         return data, path
 
+    def export_to(self, target_dir: str, compression_level: int = 3, small_only: bool = False):
+        """Write a standalone copy of the current data into ``target_dir``.
+
+        Reads the current snapshot's parquet resources and re-writes them
+        as memory-bounded parquet chunks (each ~``max_memory_chunk_size``,
+        from the per-table config or the global default) under
+        ``target_dir``.  This is a pure copy: it does NOT create a new
+        snapshot, advance the Redis leaf, or touch ``data/``/``snapshots/``.
+
+        Args:
+            target_dir: destination directory for the exported parquet
+                files (created if missing).  Typically an
+                ``export/<timestamp>/`` folder next to ``data/``.
+            compression_level: zstd level for the exported parquet.
+            small_only: when False (default) every resource is read and
+                re-chunked; when True only files smaller than
+                ``max_memory_chunk_size`` are included.
+
+        Returns:
+            ``dict`` with ``files`` (list of written paths),
+            ``files_written``, ``total_rows`` and ``total_bytes``.
+        """
+        from supertable.processing import compact_resources
+
+        snapshot, _path = self.get_simple_table_snapshot()
+        table_config = self.catalog.get_table_config(
+            self.super_table.organization,
+            self.super_table.super_name,
+            self.simple_name,
+        ) or {}
+
+        _considered, total_rows, new_resources, _sunset = compact_resources(
+            snapshot=snapshot,
+            data_dir=target_dir,
+            compression_level=compression_level,
+            table_config=table_config,
+            small_only=small_only,
+        )
+
+        files = [r.get("file") for r in new_resources if isinstance(r, dict) and r.get("file")]
+        total_bytes = sum(
+            int(r.get("file_size") or 0) for r in new_resources if isinstance(r, dict)
+        )
+        return {
+            "files": files,
+            "files_written": len(files),
+            "total_rows": int(total_rows),
+            "total_bytes": int(total_bytes),
+        }
+
     def update(self, new_resources, sunset_files, model_df, last_snapshot=None, last_snapshot_path=None, lineage=None, profiler: Optional[Profiler] = None):
         """
         Build and write a new heavy snapshot on storage.

{supertable-2.2.3 → supertable-2.2.4/supertable.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: supertable
-Version: 2.2.3
+Version: 2.2.4
 Summary: SuperTable — versioned data lake library for SQL analytics on Parquet + Redis.
 Author: Levente Kupas
 Author-email: Levente Kupas <lkupas@kladnasoft.com>