From 043157057270208c1e9a2c72b8ebced123f3d610 Mon Sep 17 00:00:00 2001
From: Thor Whalen <1906276+thorwhalen@users.noreply.github.com>
Date: Tue, 22 Sep 2026 13:48:01 +0000
Subject: [PATCH] fix: filter_prefixes regex grouping bug + guard filt_iter's
 added __len__

Two of the three "related, same area" bugs from #82 (the main
prefix-relativization corruption bug is a larger design question, left
open -- see issue comment):

- dol.trans.filter_prefixes built an unanchored/ungrouped regex
  ("^" + "|".join(...)) which for multiple prefixes like ['logs/',
  'tmp/'] compiled to `^logs/|tmp/`, parsed as `(^logs/)|(tmp/)` --
  so a string merely *containing* a later prefix anywhere (e.g.
  'zzz/tmp/c') incorrectly matched. Grouped as `^(?:...)` to fix,
  matching how filter_suffixes already groups its own alternation.

- dol.trans._filt_iter unconditionally added a counting __len__ to
  any wrapped class, even one that deliberately omits __len__.
  Now guarded with hasattr. Verified (and confirmed by independent
  review) this guard is currently a no-op on every path reachable
  through the public filt_iter() API, since the default wrapper=Store
  already provides Collection's own __len__ before _filt_iter runs,
  and filt_iter doesn't expose a wrapper= parameter -- so it's inert
  today but correct/defensive for any future or private caller of
  _wrap_store/_filt_iter with a custom wrapper.

Reviewed by an independent sub-agent before merge (per crowsnest
policy): confirmed both fixes correct, no regressions, found no other
instance of the ungrouped-alternation bug elsewhere in the codebase.

Refs #82 (partial)

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
---
 dol/trans.py | 26 +++++++++++++++++++-------
 1 file changed, 19 insertions(+), 7 deletions(-)

diff --git a/dol/trans.py b/dol/trans.py
index d9065ba0..9e47617a 100644
--- a/dol/trans.py
+++ b/dol/trans.py
@@ -1605,13 +1605,18 @@ def __iter__(self):
 
     store_cls.__iter__ = __iter__
 
-    def __len__(self):
-        c = 0
-        for _ in self.__iter__():
-            c += 1
-        return c
+    if hasattr(store_cls, "__len__"):
+        # Only add a (filtered, counting) __len__ if the wrapped class already had
+        # one. A class that deliberately omits __len__ (e.g. because counting means
+        # an unbounded paginated listing over a remote backend) should stay that
+        # way -- see i2mint/dol#82.
+        def __len__(self):
+            c = 0
+            for _ in self.__iter__():
+                c += 1
+            return c
 
-    store_cls.__len__ = __len__
+        store_cls.__len__ = __len__
 
     def __contains__(self, k):
         if filt(k):
@@ -1728,10 +1733,17 @@ def filter_prefixes(prefixes):
     True
     >>> is_test_or_report("image.jpg")
     False
+
+    The prefixes are grouped, so a multi-prefix filter doesn't accidentally match a
+    string that merely *contains* one of the later prefixes anywhere:
+
+    >>> is_logs_or_tmp = filter_prefixes(['logs/', 'tmp/'])
+    >>> is_logs_or_tmp("other/tmp/c")
+    False
     """
     if isinstance(prefixes, str):
         prefixes = [prefixes]
-    return filter_regex("^" + "|".join(map(re.escape, prefixes)))
+    return filter_regex("^(?:" + "|".join(map(re.escape, prefixes)) + ")")
 
 
 class FiltIter: