Summary

mempalace mine --dry-run crashes with a TypeError in the summary printer when any file in the scan gets room=None instead of falling through to "general". The crash happens only in --dry-run mode because the real mine path skips the problematic bookkeeping. Harmless (dry-run), but confusing when you're trying to figure out what will get filed before committing.

Environment

• Ubuntu 24.04, Python 3.12
• mempalace from PyPI, ~/.local/lib/python3.12/site-packages/mempalace/
• mempalace init run against ~/.hermes (a large, multi-topic directory)

Reproduce

cd ~/.hermes
mempalace init . --yes
mempalace mine . --limit 500 --dry-run

Output

[DRY RUN] ... → room:profiles (25 drawers)
...
=======================================================
  Done.
  Files processed: 500
  Files skipped (already filed): 0
  Drawers filed: 27501

  By room:
    profiles             479 files
    general              4 files
    scripts              4 files
Traceback (most recent call last):
  File ".../bin/mempalace", line 8, in <module>
    sys.exit(main())
  File ".../site-packages/mempalace/cli.py", line 540, in main
    dispatch[args.command](args)
  File ".../site-packages/mempalace/cli.py", line 87, in cmd_mine
    mine(...)
  File ".../site-packages/mempalace/miner.py", line 605, in mine
    print(f"    {room:20} {count} files")
                ^^^^^^^^^
TypeError: unsupported format string passed to NoneType.__format__

Root cause

In mempalace/miner.py, process_file() can return (0, None) when a file is unreadable or falls below MIN_CHUNK_SIZE:

# miner.py ~line 425-431
except OSError:
    return 0, None
content = content.strip()
if len(content) < MIN_CHUNK_SIZE:
    return 0, None

The caller then does:

drawers, room = process_file(...)  # room can be None here
if drawers == 0 and not dry_run:
    files_skipped += 1
else:
    total_drawers += drawers
    room_counts[room] += 1          # ← room=None lands in the counter

In real mine mode the drawers == 0 and not dry_run branch is taken, so None rooms get skipped before reaching the counter. In dry-run mode the second clause is always entered, so None can make it into room_counts. Then the summary print:

# miner.py line 605
for room, count in sorted(room_counts.items(), key=lambda x: x[1], reverse=True):
    print(f"    {room:20} {count} files")   # crashes on room=None

Suggested fix

Any one of:

1. Defensive print (smallest diff):

   print(f"    {(room or 'general'):20} {count} files")

2. Normalize at the source — guarantee process_file() never returns None for the room; return "general" or a sentinel like "_unreadable":

   except OSError:
       return 0, "general"
   if len(content) < MIN_CHUNK_SIZE:
       return 0, "general"

3. Filter the counter — skip None rooms before printing:

   for room, count in sorted(
       ((r, c) for r, c in room_counts.items() if r is not None),
       key=lambda x: x[1], reverse=True,
   ):
       ...

Option 2 is probably the cleanest because it also fixes any other downstream consumer that assumes room is always a string.

Severity

Cosmetic / workflow bug — real mines work fine. But --dry-run is exactly the mode you'd use to preview a large mine before committing, so crashing at the end of the preview defeats the point.

Reporter

Filed by @mssteuer on behalf of Jean Clawd, a Hermes agent.