Skip to content

Commit

Permalink
Add iai versions of the multi-like evaluation benchmarks
Browse files Browse the repository at this point in the history
  • Loading branch information
@jpschorr
jpschorr committed Feb 10, 2023
1 parent b2ce0ce commit 203f77c
Show file tree
Hide file tree
Showing 4 changed files with 451 additions and 310 deletions.
6 changes: 6 additions & 0 deletions partiql/Cargo.toml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -32,8 +32,14 @@ partiql-eval = { path = "../partiql-eval" }

itertools = "0.10"
criterion = "0.4"
iai = { git = "https://github.com/bheisler/iai" }
rand = "0.8"
once_cell = "1.17"

[[bench]]
name = "bench_eval_multi_like"
harness = false

[[bench]]
name = "bench_eval_multi_like_iai"
harness = false
313 changes: 3 additions & 310 deletions partiql/benches/bench_eval_multi_like.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@ use partiql_eval::eval::EvalPlan;
use partiql_eval::plan::EvaluatorPlanner;
use partiql_logical::{BindingsOp, LogicalPlan};

use crate::multi_like_data::{employee_data, QUERY_1, QUERY_15, QUERY_30};
use partiql_parser::{Parser, ParserResult};
use partiql_value::{partiql_tuple, Bag, Tuple, Value};

Expand All @@ -21,315 +22,7 @@ use partiql_value::{partiql_tuple, Bag, Tuple, Value};
// of queries that filter against 1, 15, or 30 `OR`ed `LIKE` expressions
// over 10201 rows of tuples containing an id and a string

fn employee_data() -> Vec<Value> {
let name1 = vec![
"Bob",
"Madden",
"Brycen",
"Bryanna",
"Zayne",
"Jocelynn",
"Breanna",
"Margaret",
"Jasmine",
"Kenyon",
"Aryanna",
"Zackery",
"Jorden",
"Malia",
"Raven",
"Neveah",
"Finley",
"Austin",
"Jaxson",
"Tobias",
"Dominique",
"Devan",
"Colby",
"Tanner",
"Mckenna",
"Kristina",
"Cristal",
"River",
"Taliyah",
"Abagail",
"Spencer",
"Gage",
"Ronnie",
"Amari",
"Jabari",
"Alanna",
"Anderson",
"Saniya",
"Baylee",
"Elisa",
"Savannah",
"Jakobe",
"Sandra",
"Simone",
"Frank",
"Braedon",
"Clark",
"Francisco",
"Roman",
"Matias",
"Messi",
"Elisha",
"Alexander",
"Kadence",
"Karsyn",
"Adonis",
"Ishaan",
"Trevon",
"Ryan",
"Jaelynn",
"Marilyn",
"Emma",
"Avah",
"Jordan",
"Riley",
"Amelie",
"Denisse",
"Darion",
"Lydia",
"Marley",
"Brogan",
"Trace",
"Maeve",
"Elijah",
"Kareem",
"Erick",
"Hope",
"Elisabeth",
"Antwan",
"Francesca",
"Layla",
"Jase",
"Angel",
"Addyson",
"Mckinley",
"Julianna",
"Winston",
"Royce",
"Paola",
"Issac",
"Zachary",
"Niko",
"Shania",
"Colin",
"Jesse",
"Pedro",
"Cheyenne",
"Ashley",
"Karli",
"Bianca",
"Mario",
];
let name2 = vec![
"Smith",
"Oconnell",
"Whitehead",
"Carrillo",
"Parrish",
"Monroe",
"Summers",
"Hurst",
"Durham",
"Hardin",
"Hunt",
"Mitchell",
"Pennington",
"Woodward",
"Franklin",
"Martinez",
"Shepard",
"Khan",
"Mcfarland",
"Frey",
"Mckenzie",
"Blair",
"Mercer",
"Callahan",
"Cameron",
"Gilmore",
"Bowers",
"Donovan",
"Meyers",
"Horne",
"Rice",
"Castillo",
"Cain",
"Dickson",
"Valenzuela",
"Silva",
"Prince",
"Vance",
"Berry",
"Coffey",
"Young",
"Walker",
"Burch",
"Ross",
"Mejia",
"Zuniga",
"Haney",
"Jordan",
"Love",
"Larsen",
"Bowman",
"Werner",
"Greer",
"Krause",
"Bishop",
"Day",
"Luna",
"Patrick",
"Adkins",
"Benson",
"Mcconnell",
"Sanchez",
"Villa",
"Wu",
"Duke",
"Fisher",
"Hess",
"Lawrence",
"Perry",
"Hardy",
"Wyatt",
"Mcknight",
"Thomas",
"Trevino",
"Flowers",
"Cisneros",
"Coleman",
"Sanders",
"Good",
"Newton",
"Carpenter",
"Garza",
"Barber",
"Swanson",
"Owen",
"Anderson",
"Bright",
"Beck",
"Lawson",
"Jones",
"Davila",
"Porter",
"Dougherty",
"Stevenson",
"Malone",
"Garrison",
"Bates",
"Wheeler",
"Petty",
"Rojas",
"Townsend",
];

// cartesian product of name1 x name2 (e.g., "Bob Smith", ... "Mario Townsend")
let combined = name1
.iter()
.cartesian_product(name2.iter())
.map(|(n1, n2)| format!("{n1} {n2}"));

// seed the rng with a known value to assure same data across runs
let mut rng = rand::rngs::StdRng::from_seed([42; 32]);
use rand::distributions::Distribution;
let chars = rand::distributions::Alphanumeric;
let random_size = rand::distributions::uniform::Uniform::from(5..=100);

// add random string prefix and suffix to each combined name
let employee_data: Vec<Value> = combined
.enumerate()
.map(|(id, person)| {
let prefix_size = random_size.sample(&mut rng);
let suffix_size = random_size.sample(&mut rng);
let prefix: String = (0..prefix_size)
.map(|_| rng.sample(chars) as char)
.collect();
let suffix: String = (0..suffix_size)
.map(|_| rng.sample(chars) as char)
.collect();
let full_name = format!("{prefix} {person} {suffix}");
partiql_tuple![("id", id), ("name", full_name)].into()
})
.collect_vec();

employee_data
}

fn data() -> MapBindings<Value> {
let data = partiql_tuple![(
"hr",
partiql_tuple![("employees", Bag::from(employee_data()))]
)];

data.into()
}

const QUERY_1: &str = "
SELECT *
FROM hr.employees as emp
WHERE lower(emp.name) LIKE '%bob smith%'
";

const QUERY_15: &str = "
SELECT *
FROM hr.employees as emp
WHERE lower(emp.name) LIKE '%bob smith%'
OR lower(emp.name) LIKE '%gage swanson%'
OR lower(emp.name) LIKE '%riley perry%'
OR lower(emp.name) LIKE '%sandra woodward%'
OR lower(emp.name) LIKE '%abagail oconnell%'
OR lower(emp.name) LIKE '%amari duke%'
OR lower(emp.name) LIKE '%elisha wyatt%'
OR lower(emp.name) LIKE '%aryanna hess%'
OR lower(emp.name) LIKE '%bryanna jones%'
OR lower(emp.name) LIKE '%trace gilmore%'
OR lower(emp.name) LIKE '%antwan stevenson%'
OR lower(emp.name) LIKE '%julianna callahan%'
OR lower(emp.name) LIKE '%jaelynn trevino%'
OR lower(emp.name) LIKE '%kadence bates%'
OR lower(emp.name) LIKE '%jakobe townsend%'
";

const QUERY_30: &str = "
SELECT *
FROM hr.employees as emp
WHERE lower(emp.name) LIKE '%bob smith%'
OR lower(emp.name) LIKE '%gage swanson%'
OR lower(emp.name) LIKE '%riley perry%'
OR lower(emp.name) LIKE '%sandra woodward%'
OR lower(emp.name) LIKE '%abagail oconnell%'
OR lower(emp.name) LIKE '%amari duke%'
OR lower(emp.name) LIKE '%elisha wyatt%'
OR lower(emp.name) LIKE '%aryanna hess%'
OR lower(emp.name) LIKE '%bryanna jones%'
OR lower(emp.name) LIKE '%trace gilmore%'
OR lower(emp.name) LIKE '%antwan stevenson%'
OR lower(emp.name) LIKE '%julianna callahan%'
OR lower(emp.name) LIKE '%jaelynn trevino%'
OR lower(emp.name) LIKE '%kadence bates%'
OR lower(emp.name) LIKE '%jakobe townsend%'
OR lower(emp.name) LIKE '%austin pennington%'
OR lower(emp.name) LIKE '%colby woodward%'
OR lower(emp.name) LIKE '%brycen blair%'
OR lower(emp.name) LIKE '%cristal mercer%'
OR lower(emp.name) LIKE '%river gilmore%'
OR lower(emp.name) LIKE '%saniya bowers%'
OR lower(emp.name) LIKE '%braedon ross%'
OR lower(emp.name) LIKE '%clark mejia%'
OR lower(emp.name) LIKE '%ryan day%'
OR lower(emp.name) LIKE '%marilyn luna%'
OR lower(emp.name) LIKE '%avah sanchez%'
OR lower(emp.name) LIKE '%amelie wu%'
OR lower(emp.name) LIKE '%paola duke%'
OR lower(emp.name) LIKE '%jesse trevino%'
OR lower(emp.name) LIKE '%bianca cisneros%'
";
mod multi_like_data;

#[inline]
fn parse(text: &str) -> ParserResult {
Expand Down Expand Up @@ -413,7 +106,7 @@ fn bench_eval(c: &mut Criterion) {
let compiled_15 = compile(&parse(QUERY_15).unwrap());
let compiled_30 = compile(&parse(QUERY_30).unwrap());

let bindings = data();
let bindings = employee_data();

c.bench_function("eval-1", |b| {
b.iter(|| {
Expand Down
Loading

0 comments on commit 203f77c

Please sign in to comment.