MongoDB::Fast — Battle Test Results

Version: 0.1.0 Date: 2026-03-22 Environment: macOS 15.7.4, Rakudo v2024.10, MongoDB 8.2.2 Auth: github:Zer0-Tolerance

Summary

========================================================================
Results: 121 passed, 0 failed  (121 total)
========================================================================

121/121 tests passing. Zero failures.

Test Sections

Section Detail

1. Basic CRUD

• insert-one, find-one, update-one ($set), delete-one

• Verified: acknowledged flag, insertedId, matchedCount, modifiedCount, deletedCount, field values

2. Data Type Edge Cases

• Large int (2³¹−1), float (π), negative int

• Boolean True/False, zero int, false bool (falsy value round-trips)

• Unicode string (Japanese, Korean, Arabic, emoji)

• Empty string, empty array

• Deeply nested document (5 levels)

• 500-element array

• 64 KB string field

• String with tab, newline, quote, backslash

3. Bulk Operations

• insert-many 500 docs — insertedCount, insertedIds length

• count-documents total and with filter

• update-many with $inc

• delete-many with filter

• count after delete

4. Query Operators

• $gt, $in, $or, $exists, $ne

5. Sort / Skip / Limit / Projection

• sort descending + limit

• skip + limit with expected landing index

• projection include/exclude verified field-by-field

6. Aggregation Pipeline

• $match + $group + $sort

• $limit

• $project with computed field ($multiply)

7. Upsert Behavior

• Upsert on non-existent doc → new doc created, upsertedId returned

• Upsert on existing doc → modified, no upsertedId

8. Replace One

• Replaces full document, strips fields not in replacement

• modifiedCount == 1

9. Update Operators

• $inc positive and negative accumulation

• $set overwrite

• $unset field removal

• $push array append

10. Index Operations

• Single-field ascending index

• Compound index with custom name

11. Cursor Batch Iteration

• cursor.all over 400+ docs with batchSize=50 (multi-batch)

• cursor.next() one-by-one over 10 docs, order verified

12. Concurrent Stress Test (single connection)

• 50 concurrent insert-one — all land

• 50 concurrent find-one — no errors

• 50 concurrent update-one — all applied

• 50 concurrent mixed (insert + find + update per task) — no errors

13. Large Bulk Insert Stress

• insert-many 5,000 docs in one batch

• count-documents == 5,000

• aggregate $group by active over 5,000 docs — 2 groups, counts sum to 5,000

14. Connection Pool Stress

• 30 concurrent inserts via pool — all land

• 30 concurrent finds via pool — no errors

• 30 concurrent mixed (update + find) via pool — all applied

15. Error / Edge Case Handling

• find-one on empty collection → undefined (not crash)

• count on empty collection → 0

• drop non-existent collection → no throw (idempotent)

• update-one no-match (no upsert) → matchedCount 0, modifiedCount 0

• delete-one no-match → deletedCount 0

• duplicate _id insert → throws with E11000 message

• empty filter {} → matches all docs

Bugs Found & Fixed During Development

Bug #1: replace-one not pool-aware

Error: Cannot look up attributes in a MongoDB::Fast::Connection type object Cause: replace-one used $!connection.run-command(...) directly instead of self!run-command(...), bypassing the pool Fix: lib/MongoDB/Fast/Collection.rakumod:233 — changed to self!run-command(%command) Status: ✅ Fixed

Bug #2: Pool client connect not called

Error: Same type-object error triggered from pool section Cause: battle-test.raku created $pool-client but never called await $pool-client.connect before using collections Fix: Added await $pool-client.connect in section 14 Status: ✅ Fixed

Bug #3: $client.generate-object-id does not exist

Error: No such method 'generate-object-id' for invocant of type 'MongoDB::Fast' Cause: The method lives on MongoDB::Fast::BSON, not the top-level client Fix: Changed to MongoDB::Fast::BSON.new.generate-object-id Status: ✅ Fixed

Known Limitations

Performance Reference (compare-battle.raku, N=500, REPS=30, CONC=50)

Operation                              MF (ms)    MG (ms)    Winner
---------------------------------------------------------------------
insert-one                              36.3       33.8      MG  +8%   (statistical tie)
insert-many (500 docs)                  63.1      108.6      MF  +72%  (7,929 docs/s)
find-one (×30)                          27.2      210.9      MF +676%  (0.91 ms/op avg)
cursor.all (500 docs, batchSize=50)     29.1      463.0      MF +1492%
update-one (×30)                        29.8      179.7      MF +502%  (0.99 ms/op avg)
update-many (100 docs)                   4.4        8.6      MF  +95%
delete-one (×10)                        12.3       55.3      MF +350%
delete-many (100 docs)                   2.4        6.3      MF +160%
aggregate ($group+$sort)                 2.3        7.4      MF +227%
count-documents with filter (×10)        8.6       54.2      MF +533%
create-index (compound)                 42.9       40.4      MG   +6%  (statistical tie)
50× concurrent insert-one               43.8      256.1      MF +484%  (1,141 ops/s)
50× concurrent mixed ops                46.0      728.1      MF +1482%
upsert + $push + $inc                    2.6        7.4      MF +183%

Score: MF 12 wins — MG 2 wins — 0 ties
Correctness: ✓ Both drivers correct on all 14 operations

MF = MongoDB::Fast, MG = MongoDB (official driver)

Verdict

MongoDB::Fast v0.1.0 is production-ready for:

• ✅ Standard CRUD at high throughput

• ✅ Bulk insert/update/delete

• ✅ Aggregation pipelines

• ✅ Concurrent operations (single connection + pool)

• ✅ All core BSON types including unicode, large strings, nested docs

• ✅ Connection pooling under concurrent load

• ✅ Correct error propagation (duplicate keys, no-match ops, empty collections)

Not suitable for applications requiring authentication, GridFS, transactions, or change streams — use the official MongoDB driver for those.