Skip to content

Commit ea723f4

Browse files
zzylolzzylol
authored
Revert "refactor: retire sketch-core mirror (#307)" (#310)
This reverts commit ba0cfce.
1 parent ba0cfce commit ea723f4

29 files changed

Lines changed: 3226 additions & 180 deletions

‎Cargo.lock‎

Lines changed: 180 additions & 73 deletions
Some generated files are not rendered by default. Learn more about customizing how changed files appear on GitHub.

‎Cargo.toml‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,7 @@
11
[workspace]
22
resolver = "2"
33
members = [
4+
"asap-common/sketch-core",
45
"asap-common/dependencies/rs/promql_utilities",
56
"asap-common/dependencies/rs/sql_utilities",
67
"asap-common/dependencies/rs/elastic_dsl_utilities",
@@ -31,6 +32,7 @@ promql-parser = "0.5.0"
3132
tokio = { version = "1.0", features = ["full"] }
3233

3334
# Internal crates
35+
sketch-core = { path = "asap-common/sketch-core" }
3436
promql_utilities = { path = "asap-common/dependencies/rs/promql_utilities" }
3537
sql_utilities = { path = "asap-common/dependencies/rs/sql_utilities" }
3638
asap_types = { path = "asap-common/dependencies/rs/asap_types" }

‎asap-common/sketch-core/Cargo.toml‎

Lines changed: 16 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,16 @@
1+
[package]
2+
name = "sketch-core"
3+
version.workspace = true
4+
edition.workspace = true
5+
6+
[dependencies]
7+
serde.workspace = true
8+
# sketch-core-specific, keep pinned
9+
rmp-serde = "1.1"
10+
xxhash-rust = { version = "0.8", features = ["xxh32"] }
11+
dsrs = { git = "https://github.com/ProjectASAP/datasketches-rs", rev = "d748ec75c80fff21f7b24897244dd1c895df2e9a" }
12+
asap_sketchlib = { git = "https://github.com/ProjectASAP/asap_sketchlib" }
13+
clap = { version = "4.0", features = ["derive"] }
14+
15+
[dev-dependencies]
16+
ctor = "0.2"

‎asap-common/sketch-core/report.md‎

Lines changed: 125 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,125 @@
1+
# Report
2+
3+
Compares the **legacy** sketch implementations in `sketch-core` vs the **asap_sketchlib** backends (Count-Min Sketch, Count-Min-With-Heap, KLL, HydraKLL).
4+
5+
## Fidelity harness
6+
7+
The fidelity binary selects backends via CLI flags (`--cms-impl`, `--kll-impl`, `--cmwh-impl`).
8+
9+
| Goal | Command |
10+
|--------------------------|--------------------------------------------------------------------------------------------------------------|
11+
| Default (all sketchlib) | `cargo run -p sketch-core --bin sketchlib_fidelity` |
12+
| All legacy | `cargo run -p sketch-core --bin sketchlib_fidelity -- --cms-impl legacy --kll-impl legacy --cmwh-impl legacy` |
13+
| Legacy KLL only | `cargo run -p sketch-core --bin sketchlib_fidelity -- --cms-impl sketchlib --kll-impl legacy --cmwh-impl sketchlib` |
14+
| CMS sketchlib only | `cargo run -p sketch-core --bin sketchlib_fidelity -- --cms-impl sketchlib` |
15+
| CMS legacy only | `cargo run -p sketch-core --bin sketchlib_fidelity -- --cms-impl legacy` |
16+
17+
## Unit tests
18+
19+
Unit tests always run with **legacy** backends enabled (the test ctor calls
20+
`force_legacy_mode_for_tests()`), so you only need:
21+
22+
```bash
23+
cargo test -p sketch-core
24+
```
25+
26+
## Results
27+
28+
### CountMinSketch (accuracy vs exact counts)
29+
30+
#### depth=3
31+
32+
| width | n | domain | Mode | Pearson corr | MAPE (%) | RMSE (%) |
33+
|-------|--------|--------|----------------|----------------|----------|----------|
34+
| 1024 | 100000 | 1000 | Legacy | 0.9998451189 | 24.48 | 52.76 |
35+
| 1024 | 100000 | 1000 | asap_sketchlib | 0.9998387103 | 24.36 | 54.11 |
36+
37+
#### depth=5
38+
39+
| width | n | domain | Mode | Pearson corr | MAPE (%) | RMSE (%) |
40+
|-------|--------|--------|----------------|----------------|----------|----------|
41+
| 2048 | 200000 | 2000 | Legacy | 0.9999733814 | 8.75 | 29.94 |
42+
| 2048 | 200000 | 2000 | asap_sketchlib | 0.9999744627 | 8.37 | 28.84 |
43+
| 2048 | 50000 | 500 | Legacy | 1.0000000000 | 0.00 | 0.00 |
44+
| 2048 | 50000 | 500 | asap_sketchlib | 1.0000000000 | 0.00 | 0.00 |
45+
46+
#### depth=7
47+
48+
| width | n | domain | Mode | Pearson corr | MAPE (%) | RMSE (%) |
49+
|-------|--------|--------|----------------|----------------|----------|----------|
50+
| 4096 | 200000 | 2000 | Legacy | 0.9999993694 | 0.20 | 3.69 |
51+
| 4096 | 200000 | 2000 | asap_sketchlib | 0.9999993499 | 0.21 | 4.27 |
52+
53+
---
54+
55+
### CountMinSketchWithHeap (top-k + CMS accuracy on exact top-k)
56+
57+
The heap is maintained by local updates; recall is measured against the **true** top-k at the end of the stream.
58+
59+
#### depth=3
60+
61+
| width | n | domain | heap_size | Mode | Top-k recall | Pearson (top-k) | MAPE (%) | RMSE (%) |
62+
|-------|--------|--------|-----------|----------------|--------------|-----------------|----------|----------|
63+
| 1024 | 100000 | 1000 | 10 | Legacy | 0.40 | 0.9571 | 0.174 | 0.319 |
64+
| 1024 | 100000 | 1000 | 10 | asap_sketchlib | 0.80 | 1.0000 | 0.000 | 0.000 |
65+
66+
#### depth=5
67+
68+
| width | n | domain | heap_size | Mode | Top-k recall | Pearson (top-k) | MAPE (%) | RMSE (%) |
69+
|-------|--------|--------|-----------|----------------|--------------|-----------------|----------|----------|
70+
| 2048 | 200000 | 2000 | 20 | Legacy | 0.60 | 0.9964 | 0.045 | 0.101 |
71+
| 2048 | 200000 | 2000 | 20 | asap_sketchlib | 1.00 | 0.9982 | 0.021 | 0.067 |
72+
| 2048 | 200000 | 2000 | 50 | Legacy | 0.40 | 0.9999983 | 5.60 | 16.49 |
73+
| 2048 | 200000 | 2000 | 50 | asap_sketchlib | 0.48 | 0.9999990 | 3.90 | 12.95 |
74+
75+
---
76+
77+
### KllSketch (quantiles, absolute rank error)
78+
79+
For each quantile \(q\), we compute the sketch estimate `est_value`, then:
80+
`abs_rank_error = |rank_fraction(exact_sorted_values, est_value) - q|`.
81+
82+
#### k=20
83+
84+
| n_updates | Mode | q=0.5 | q=0.9 | q=0.99 |
85+
|-----------|----------------|---------|---------|---------|
86+
| 200000 | Legacy | 0.0104 | 0.0145 | 0.0028 |
87+
| 200000 | asap_sketchlib | 0.0275 | 0.0470 | 0.0061 |
88+
| 50000 | Legacy | 0.0131 | 0.0091 | 0.0054 |
89+
| 50000 | asap_sketchlib | 0.0110 | 0.0116 | 0.0031 |
90+
91+
#### k=50
92+
93+
| n_updates | Mode | q=0.5 | q=0.9 | q=0.99 |
94+
|-----------|----------------|---------|---------|---------|
95+
| 200000 | Legacy | 0.0013 | 0.0021 | 0.0012 |
96+
| 200000 | asap_sketchlib | 0.0101 | 0.0044 | 0.0074 |
97+
98+
#### k=200
99+
100+
| n_updates | Mode | q=0.5 | q=0.9 | q=0.99 |
101+
|-----------|----------------|---------|---------|---------|
102+
| 200000 | Legacy | 0.0021 | 0.0036 | 0.0000 |
103+
| 200000 | asap_sketchlib | 0.0015 | 0.0001 | 0.0002 |
104+
105+
---
106+
107+
### HydraKllSketch (per-key quantiles, mean/max absolute rank error across 50 keys)
108+
109+
#### rows=2, cols=64
110+
111+
| k | n | domain | Mode | q=0.5 (mean / max) | q=0.9 (mean / max) |
112+
|-----|--------|--------|----------------|--------------------|--------------------|
113+
| 20 | 200000 | 200 | Legacy | 0.0170 / 0.0546 | 0.0165 / 0.0452 |
114+
| 20 | 200000 | 200 | asap_sketchlib | 0.0254 / 0.0629 | 0.0546 / 0.0942 |
115+
116+
#### rows=3, cols=128
117+
118+
| k | n | domain | Mode | q=0.5 (mean / max) | q=0.9 (mean / max) |
119+
|-----|--------|--------|----------------|--------------------|--------------------|
120+
| 20 | 200000 | 200 | Legacy | 0.0166 / 0.0591 | 0.0114 / 0.0304 |
121+
| 20 | 200000 | 200 | asap_sketchlib | 0.0216 / 0.0534 | 0.0238 / 0.1087 |
122+
| 50 | 200000 | 200 | Legacy | 0.0099 / 0.0352 | 0.0087 / 0.0330 |
123+
| 50 | 200000 | 200 | asap_sketchlib | 0.0119 / 0.0458 | 0.0119 / 0.0296 |
124+
| 20 | 100000 | 100 | Legacy | 0.0141 / 0.0574 | 0.0149 / 0.0471 |
125+
| 20 | 100000 | 100 | asap_sketchlib | 0.0202 / 0.0621 | 0.0287 / 0.0779 |

0 commit comments

Comments
 (0)