Skip to content

Latest commit

 

History

History
114 lines (92 loc) · 4.39 KB

File metadata and controls

114 lines (92 loc) · 4.39 KB

Testing Report

中文版

Summary

Test Suite Tests Status
bitscout-core unit 122 All pass
rg conformance (e2e) 18 All pass
grep conformance (e2e) 8 All pass
find/fd conformance (e2e) 8 All pass
cat conformance (e2e) 10 All pass
full conformance (e2e) 44 All pass
semantic accuracy (e2e) 12 All pass
benchmark modes (e2e) 3 All pass
Total 181+ All pass

Running Tests

cargo test                                             # All tests
cargo test -p bitscout-core --lib                      # Core unit tests (122)
cargo test -p bitscout-e2e --test full_conformance     # Conformance (44)
cargo test -p bitscout-e2e --test semantic_accuracy    # Semantic accuracy (12)
cargo test -p bitscout-e2e --test benchmark_modes      # Performance benchmarks (3)

Conformance: 100% Compatible

44 conformance tests compare BitScout output against real rg/grep/find/fd/cat, line-by-line exact match:

Tool    Tests   Pass    Accuracy
------- ------- ------- --------
rg       18      18     100%
grep      8       8     100%
find      4       4     100%
fd        4       4     100%
cat      10      10     100%
------- ------- ------- --------
Total    44      44     100%

12 regex patterns compared against real rg, line-level accuracy 100%.

Speed: 3-81x Faster

Cold-start performance comparison (no daemon, independent scan per invocation):

Tool       real cmd    BitScout    Speedup
---------- ---------- ---------- ---------
cat          1.1ms     0.014ms     81.3x
find         1.5ms     0.12ms      12.9x
rg           6.1ms     0.77ms       8.0x
grep         2.6ms     0.82ms       3.2x

All Search Modes

Cold-start latency including FileTree scan + search + scoring:

Mode                    Avg(us)   Notes
----------------------- --------- -------------------------
find -name '*.rs'          123   Filename glob matching
fd -e rs                   117   Filename extension matching
cat file                    14   File read
rg literal                 769   Literal search
rg regex                  2456   Regex search
grep -rn                   823   Recursive line search
rg --bm25                  907   BM25 relevance scoring
rg --bm25=full             948   BM25 + IDF full scoring
rg --semantic              771   LSA semantic search

All modes < 2.5ms, meeting real-time interaction requirements for AI agents.

Semantic Search Accuracy: Top-1 100%

LSA semantic search (embedding-free, based on truncated SVD of TF-IDF):

Query                            Expected #1        Result         Score
-------------------------------- ------------------ -------------- ------
login password authenticate      auth_login.rs      auth_login.rs  0.214
jwt token generate validate      auth_jwt.rs        auth_jwt.rs    0.642
database query insert migrate    database.rs        database.rs    0.235
http server listen connection    http_server.rs     http_server.rs 0.168
cache evict lru store            cache.rs           cache.rs       0.244
-------------------------------- ------------------ ----------------------
Top-1 Accuracy: 5/5 (100%)
Top-3 Accuracy: 5/5 (100%)

RP unit-level discrimination: auth(0.70) > session(0.13) > math(-0.20) — relevant documents score significantly higher than unrelated ones.

Semantic Test Coverage

Test Description
RP cosine similarity ranking Verifies correct ordering: highly related > somewhat related > unrelated
RP code pattern ranking Error handling terms rank above database/config terms
Auth file ranking for token query Auth-focused files outrank tokenizer and config files
Database file ranking Database file ranks first for DB-related queries
Semantic reordering vs plain search Semantic results differ from filesystem order
Multi-word query ranking Auth files appear in top-2 for "authenticate session"
Score variance Meaningful spread between related and unrelated documents
Deterministic scoring Same query always produces same scores
Comprehensive accuracy report 5 diverse queries with expected rankings
LSA cross-vocabulary Co-occurring terms enable semantic bridging
LSA synonym discovery Co-occurring pairs (error/exception, success/ok) become similar
Pure LSA cross-vocabulary Auth files rank above DB files without external embeddings