From d7d9d30180df45061311dd7c65008e3945db967c Mon Sep 17 00:00:00 2001
From: Jim Pivarski <jpivarski@users.noreply.github.com>
Date: Tue, 5 Nov 2024 17:16:41 -0600
Subject: [PATCH] perf: removing output arrays from reference cycles so they
 don't have to wait for GC (#1305)

* perf: desperate attempts to unlink output arrays

* implement https://github.com/scikit-hep/uproot5/pull/1305#issuecomment-2407868371

* clear values dict to avoid holding onto computed arrays

* clear scope to avoid holding onto big arrays & add a note for the future...

---------

Co-authored-by: Peter Fackeldey <peter.fackeldey@rwth-aachen.de>
---
 src/uproot/behaviors/TBranch.py | 23 ++++++++++++++++++-----
 src/uproot/language/python.py   |  4 ++++
 2 files changed, 22 insertions(+), 5 deletions(-)

diff --git a/src/uproot/behaviors/TBranch.py b/src/uproot/behaviors/TBranch.py
index 2d2778564..3fb57ab7e 100644
--- a/src/uproot/behaviors/TBranch.py
+++ b/src/uproot/behaviors/TBranch.py
@@ -210,10 +210,16 @@ def iterate(
                             arrays, report = item
                             arrays = library.global_index(arrays, global_offset)
                             report = report.to_global(global_offset)
-                            yield arrays, report
+                            popper = [arrays]
+                            del arrays
+                            del item
+                            yield popper.pop(), report
+
                         else:
-                            arrays = library.global_index(item, global_offset)
-                            yield arrays
+                            popper = [library.global_index(item, global_offset)]
+                            del item
+                            yield popper.pop()
+
                 except uproot.exceptions.KeyInFileError:
                     if allow_missing:
                         continue
@@ -1111,6 +1117,9 @@ def iterate(
                     ak_add_doc,
                 )
 
+                # no longer needed; save memory
+                del output
+
                 next_baskets = {}
                 for branch, basket_num, basket in ranges_or_baskets:
                     basket_entry_start, basket_entry_stop = basket.entry_start_stop
@@ -1119,10 +1128,14 @@ def iterate(
 
                 previous_baskets = next_baskets
 
+                # no longer needed; save memory
+                popper = [out]
+                del out
+
                 if report:
-                    yield out, Report(self, sub_entry_start, sub_entry_stop)
+                    yield popper.pop(), Report(self, sub_entry_start, sub_entry_stop)
                 else:
-                    yield out
+                    yield popper.pop()
 
     def keys(
         self,
diff --git a/src/uproot/language/python.py b/src/uproot/language/python.py
index e3ce8838a..f72f68f65 100644
--- a/src/uproot/language/python.py
+++ b/src/uproot/language/python.py
@@ -516,6 +516,10 @@ def getter(name):
                 else:
                     output[name] = output[name][cut]
 
+        # clear dicts to get rid of big arrays.
+        # note: without this these arrays are not properly released from memory!
+        values.clear()
+        scope.clear()
         return output