From e21b8d54f822b80bcc4b07e373ecb187e1a34faa Mon Sep 17 00:00:00 2001
From: Matthew Roeschke <10647082+mroeschke@users.noreply.github.com>
Date: Sun, 2 Nov 2025 19:29:11 -0800
Subject: [PATCH 1/3] Make _merge_blocks lazy

---
 pandas/core/internals/managers.py | 31 +++++++++++--------------------
 1 file changed, 11 insertions(+), 20 deletions(-)

diff --git a/pandas/core/internals/managers.py b/pandas/core/internals/managers.py
index 40ba74fed49d1..8d94606660639 100644
--- a/pandas/core/internals/managers.py
+++ b/pandas/core/internals/managers.py
@@ -1938,7 +1938,7 @@ def _consolidate_check(self) -> None:
 
     def _consolidate_inplace(self) -> None:
         if not self.is_consolidated():
-            self.blocks = _consolidate(self.blocks)
+            self.blocks = tuple(_consolidate(self.blocks))
             self._is_consolidated = True
             self._known_consolidated = True
             self._rebuild_blknos_and_blklocs()
@@ -2430,30 +2430,21 @@ def _stack_arrays(tuples, dtype: np.dtype):
     return stacked, placement
 
 
-def _consolidate(blocks: tuple[Block, ...]) -> tuple[Block, ...]:
+def _consolidate(blocks: tuple[Block, ...]) -> Generator[Block]:
     """
     Merge blocks having same dtype, exclude non-consolidating blocks
     """
     # sort by _can_consolidate, dtype
     gkey = lambda x: x._consolidate_key
     grouper = itertools.groupby(sorted(blocks, key=gkey), gkey)
-
-    new_blocks: list[Block] = []
-    for (_can_consolidate, dtype), group_blocks in grouper:
-        merged_blocks, _ = _merge_blocks(
-            list(group_blocks), dtype=dtype, can_consolidate=_can_consolidate
-        )
-        new_blocks = extend_blocks(merged_blocks, new_blocks)
-    return tuple(new_blocks)
+    for (_can_consolidate, _), group_blocks in grouper:
+        yield from _merge_blocks(tuple(group_blocks), can_consolidate=_can_consolidate)
 
 
-def _merge_blocks(
-    blocks: list[Block], dtype: DtypeObj, can_consolidate: bool
-) -> tuple[list[Block], bool]:
+def _merge_blocks(blocks: tuple[Block], can_consolidate: bool) -> Generator[Block]:
     if len(blocks) == 1:
-        return blocks, False
-
-    if can_consolidate:
+        yield from blocks
+    elif can_consolidate:
         # TODO: optimization potential in case all mgrs contain slices and
         # combination of those slices is a slice, too.
         new_mgr_locs = np.concatenate([b.mgr_locs.as_array for b in blocks])
@@ -2476,10 +2467,10 @@ def _merge_blocks(
         new_mgr_locs = new_mgr_locs[argsort]
 
         bp = BlockPlacement(new_mgr_locs)
-        return [new_block_2d(new_values, placement=bp)], True
-
-    # can't consolidate --> no merge
-    return blocks, False
+        yield new_block_2d(new_values, placement=bp)
+    else:
+        # can't consolidate --> no merge
+        yield from blocks
 
 
 def _preprocess_slice_or_indexer(

From beb6649913e44bb9185d33d60b9cfaf535f98a3c Mon Sep 17 00:00:00 2001
From: Matthew Roeschke <10647082+mroeschke@users.noreply.github.com>
Date: Mon, 3 Nov 2025 12:08:03 -0800
Subject: [PATCH 2/3] Fix typing checks

---
 pandas/core/internals/managers.py | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/pandas/core/internals/managers.py b/pandas/core/internals/managers.py
index 8d94606660639..5cfc3d4782a95 100644
--- a/pandas/core/internals/managers.py
+++ b/pandas/core/internals/managers.py
@@ -2441,7 +2441,7 @@ def _consolidate(blocks: tuple[Block, ...]) -> Generator[Block]:
         yield from _merge_blocks(tuple(group_blocks), can_consolidate=_can_consolidate)
 
 
-def _merge_blocks(blocks: tuple[Block], can_consolidate: bool) -> Generator[Block]:
+def _merge_blocks(blocks: tuple[Block, ...], can_consolidate: bool) -> Generator[Block]:
     if len(blocks) == 1:
         yield from blocks
     elif can_consolidate:
@@ -2452,11 +2452,7 @@ def _merge_blocks(blocks: tuple[Block], can_consolidate: bool) -> Generator[Bloc
         new_values: ArrayLike
 
         if isinstance(blocks[0].dtype, np.dtype):
-            # error: List comprehension has incompatible type List[Union[ndarray,
-            # ExtensionArray]]; expected List[Union[complex, generic,
-            # Sequence[Union[int, float, complex, str, bytes, generic]],
-            # Sequence[Sequence[Any]], SupportsArray]]
-            new_values = np.vstack([b.values for b in blocks])  # type: ignore[misc]
+            new_values = np.vstack([b.values for b in blocks])
         else:
             bvals = [blk.values for blk in blocks]
             bvals2 = cast(Sequence[NDArrayBackedExtensionArray], bvals)

From 96e5d3f1f24e9365c62162037a8ee74539b3f8d6 Mon Sep 17 00:00:00 2001
From: Matthew Roeschke <10647082+mroeschke@users.noreply.github.com>
Date: Mon, 3 Nov 2025 13:12:46 -0800
Subject: [PATCH 3/3] Add back type ignore

---
 pandas/core/internals/managers.py | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/pandas/core/internals/managers.py b/pandas/core/internals/managers.py
index 5cfc3d4782a95..a347478413ec7 100644
--- a/pandas/core/internals/managers.py
+++ b/pandas/core/internals/managers.py
@@ -2452,7 +2452,11 @@ def _merge_blocks(blocks: tuple[Block, ...], can_consolidate: bool) -> Generator
         new_values: ArrayLike
 
         if isinstance(blocks[0].dtype, np.dtype):
-            new_values = np.vstack([b.values for b in blocks])
+            # error: List comprehension has incompatible type List[Union[ndarray,
+            # ExtensionArray]]; expected List[Union[complex, generic,
+            # Sequence[Union[int, float, complex, str, bytes, generic]],
+            # Sequence[Sequence[Any]], SupportsArray]]
+            new_values = np.vstack([b.values for b in blocks])  # type: ignore[misc]
         else:
             bvals = [blk.values for blk in blocks]
             bvals2 = cast(Sequence[NDArrayBackedExtensionArray], bvals)