558: Deletion benchmarks r=ManyTheFish a=ManyTheFish

Add benchmarks on the deletion and start rethinking benchmark names.

Co-authored-by: ManyTheFish <many@meilisearch.com>
This commit is contained in:
bors[bot] 2022-06-16 09:34:37 +00:00 committed by GitHub
commit a59ae19842
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 519 additions and 255 deletions

View File

@ -16,6 +16,9 @@ jemallocator = "0.3.2"
[dev-dependencies] [dev-dependencies]
heed = { git = "https://github.com/meilisearch/heed", tag = "v0.12.1" } heed = { git = "https://github.com/meilisearch/heed", tag = "v0.12.1" }
criterion = { version = "0.3.5", features = ["html_reports"] } criterion = { version = "0.3.5", features = ["html_reports"] }
rand = "0.8.5"
rand_chacha = "0.3.1"
roaring = "0.9.0"
[build-dependencies] [build-dependencies]
anyhow = "1.0.56" anyhow = "1.0.56"

View File

@ -5,14 +5,21 @@ use std::fs::{create_dir_all, remove_dir_all};
use std::path::Path; use std::path::Path;
use criterion::{criterion_group, criterion_main, Criterion}; use criterion::{criterion_group, criterion_main, Criterion};
use heed::EnvOpenOptions; use heed::{EnvOpenOptions, RwTxn};
use milli::update::{IndexDocuments, IndexDocumentsConfig, IndexerConfig, Settings}; use milli::update::{
DeleteDocuments, IndexDocuments, IndexDocumentsConfig, IndexerConfig, Settings,
};
use milli::Index; use milli::Index;
use rand::seq::SliceRandom;
use rand_chacha::rand_core::SeedableRng;
use roaring::RoaringBitmap;
#[cfg(target_os = "linux")] #[cfg(target_os = "linux")]
#[global_allocator] #[global_allocator]
static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc; static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc;
const BENCHMARK_ITERATION: usize = 10;
fn setup_dir(path: impl AsRef<Path>) { fn setup_dir(path: impl AsRef<Path>) {
match remove_dir_all(path.as_ref()) { match remove_dir_all(path.as_ref()) {
Ok(_) => (), Ok(_) => (),
@ -31,39 +38,95 @@ fn setup_index() -> Index {
Index::new(options, path).unwrap() Index::new(options, path).unwrap()
} }
fn setup_settings<'t>(
wtxn: &mut RwTxn<'t, '_>,
index: &'t Index,
primary_key: &str,
searchable_fields: &[&str],
filterable_fields: &[&str],
sortable_fields: &[&str],
) {
let config = IndexerConfig::default();
let mut builder = Settings::new(wtxn, index, &config);
builder.set_primary_key(primary_key.to_owned());
let searchable_fields = searchable_fields.iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
let filterable_fields = filterable_fields.iter().map(|s| s.to_string()).collect();
builder.set_filterable_fields(filterable_fields);
let sortable_fields = sortable_fields.iter().map(|s| s.to_string()).collect();
builder.set_sortable_fields(sortable_fields);
builder.execute(|_| ()).unwrap();
}
fn setup_index_with_settings<'t>(
primary_key: &str,
searchable_fields: &[&str],
filterable_fields: &[&str],
sortable_fields: &[&str],
) -> milli::Index {
let index = setup_index();
let mut wtxn = index.write_txn().unwrap();
setup_settings(
&mut wtxn,
&index,
primary_key,
searchable_fields,
filterable_fields,
sortable_fields,
);
wtxn.commit().unwrap();
index
}
fn choose_document_ids_from_index_batched(
index: &Index,
count: usize,
batch_size: usize,
) -> Vec<RoaringBitmap> {
let rtxn = index.read_txn().unwrap();
// create batch of document ids to delete
let mut rng = rand_chacha::ChaCha8Rng::seed_from_u64(7700);
let document_ids: Vec<_> = index.documents_ids(&rtxn).unwrap().into_iter().collect();
let document_ids_to_delete: Vec<_> =
document_ids.choose_multiple(&mut rng, count).map(Clone::clone).collect();
document_ids_to_delete
.chunks(batch_size)
.map(|c| {
let mut batch = RoaringBitmap::new();
for id in c {
batch.insert(*id);
}
batch
})
.collect()
}
fn indexing_songs_default(c: &mut Criterion) { fn indexing_songs_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing songs with default settings", |b| { group.bench_function("Indexing songs with default settings", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "album", "artist"];
let filterable_fields =
["released-timestamp", "duration-float", "genre", "country", "artist"];
let sortable_fields = [];
let config = IndexerConfig::default(); setup_index_with_settings(
let mut wtxn = index.write_txn().unwrap(); &primary_key,
let mut builder = Settings::new(&mut wtxn, &index, &config); &searchable_fields,
&filterable_fields,
builder.set_primary_key("id".to_owned()); &sortable_fields,
let displayed_fields = )
["title", "album", "artist", "genre", "country", "released", "duration"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields =
["title", "album", "artist"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
let faceted_fields =
["released-timestamp", "duration-float", "genre", "country", "artist"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_filterable_fields(faceted_fields);
builder.execute(|_| ()).unwrap();
wtxn.commit().unwrap();
index
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -84,41 +147,85 @@ fn indexing_songs_default(c: &mut Criterion) {
}); });
} }
fn indexing_songs_in_three_batches_default(c: &mut Criterion) { fn deleting_songs_in_batches_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing songs in three batches with default settings", |b| { group.bench_function("-songs-delete-facetedString-facetedNumber-searchable-", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "album", "artist"];
let filterable_fields =
["released-timestamp", "duration-float", "genre", "country", "artist"];
let sortable_fields = [];
let config = IndexerConfig::default(); let index = setup_index_with_settings(
let mut wtxn = index.write_txn().unwrap(); &primary_key,
let mut builder = Settings::new(&mut wtxn, &index, &config); &searchable_fields,
&filterable_fields,
builder.set_primary_key("id".to_owned()); &sortable_fields,
let displayed_fields = );
["title", "album", "artist", "genre", "country", "released", "duration"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields =
["title", "album", "artist"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
let faceted_fields =
["released-timestamp", "duration-float", "genre", "country", "artist"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_filterable_fields(faceted_fields);
builder.execute(|_| ()).unwrap();
// We index only one half of the dataset in the setup part // We index only one half of the dataset in the setup part
// as we don't care about the time it takes. // as we don't care about the time it takes.
let config = IndexerConfig::default(); let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let indexing_config = IndexDocumentsConfig::default();
let mut builder =
IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| ())
.unwrap();
let documents = utils::documents_from(datasets_paths::SMOL_SONGS, "csv");
builder.add_documents(documents).unwrap();
builder.execute().unwrap();
wtxn.commit().unwrap();
let count = 1250;
let batch_size = 250;
let document_ids_to_delete =
choose_document_ids_from_index_batched(&index, count, batch_size);
(index, document_ids_to_delete)
},
move |(index, document_ids_to_delete)| {
let mut wtxn = index.write_txn().unwrap();
for ids in document_ids_to_delete {
let mut builder = DeleteDocuments::new(&mut wtxn, &index).unwrap();
builder.delete_documents(&ids);
builder.execute().unwrap();
}
wtxn.commit().unwrap();
index.prepare_for_closing().wait();
},
)
});
}
fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing");
group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing songs in three batches with default settings", |b| {
b.iter_with_setup(
move || {
let primary_key = "id";
let searchable_fields = ["title", "album", "artist"];
let filterable_fields =
["released-timestamp", "duration-float", "genre", "country", "artist"];
let sortable_fields = [];
let index = setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
);
// We index only one half of the dataset in the setup part
// as we don't care about the time it takes.
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let indexing_config = IndexDocumentsConfig::default(); let indexing_config = IndexDocumentsConfig::default();
let mut builder = let mut builder =
IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| ()) IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| ())
@ -160,34 +267,21 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
fn indexing_songs_without_faceted_numbers(c: &mut Criterion) { fn indexing_songs_without_faceted_numbers(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing songs without faceted numbers", |b| { group.bench_function("Indexing songs without faceted numbers", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "album", "artist"];
let filterable_fields = ["genre", "country", "artist"];
let sortable_fields = [];
let config = IndexerConfig::default(); setup_index_with_settings(
let mut wtxn = index.write_txn().unwrap(); &primary_key,
let mut builder = Settings::new(&mut wtxn, &index, &config); &searchable_fields,
&filterable_fields,
builder.set_primary_key("id".to_owned()); &sortable_fields,
let displayed_fields = )
["title", "album", "artist", "genre", "country", "released", "duration"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields =
["title", "album", "artist"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
let faceted_fields =
["genre", "country", "artist"].iter().map(|s| s.to_string()).collect();
builder.set_filterable_fields(faceted_fields);
builder.execute(|_| ()).unwrap();
wtxn.commit().unwrap();
index
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -211,30 +305,21 @@ fn indexing_songs_without_faceted_numbers(c: &mut Criterion) {
fn indexing_songs_without_faceted_fields(c: &mut Criterion) { fn indexing_songs_without_faceted_fields(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing songs without any facets", |b| { group.bench_function("Indexing songs without any facets", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "album", "artist"];
let filterable_fields = [];
let sortable_fields = [];
let config = IndexerConfig::default(); setup_index_with_settings(
let mut wtxn = index.write_txn().unwrap(); &primary_key,
let mut builder = Settings::new(&mut wtxn, &index, &config); &searchable_fields,
&filterable_fields,
builder.set_primary_key("id".to_owned()); &sortable_fields,
let displayed_fields = )
["title", "album", "artist", "genre", "country", "released", "duration"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields =
["title", "album", "artist"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
builder.execute(|_| ()).unwrap();
wtxn.commit().unwrap();
index
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -257,29 +342,21 @@ fn indexing_songs_without_faceted_fields(c: &mut Criterion) {
fn indexing_wiki(c: &mut Criterion) { fn indexing_wiki(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing wiki", |b| { group.bench_function("Indexing wiki", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "body"];
let filterable_fields = [];
let sortable_fields = [];
let config = IndexerConfig::default(); setup_index_with_settings(
let mut wtxn = index.write_txn().unwrap(); &primary_key,
let mut builder = Settings::new(&mut wtxn, &index, &config); &searchable_fields,
&filterable_fields,
builder.set_primary_key("id".to_owned()); &sortable_fields,
let displayed_fields = )
["title", "body", "url"].iter().map(|s| s.to_string()).collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields = ["title", "body"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
// there is NO faceted fields at all
builder.execute(|_| ()).unwrap();
wtxn.commit().unwrap();
index
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -301,28 +378,81 @@ fn indexing_wiki(c: &mut Criterion) {
}); });
} }
fn deleting_wiki_in_batches_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing");
group.sample_size(BENCHMARK_ITERATION);
group.bench_function("-wiki-delete-searchable-", |b| {
b.iter_with_setup(
move || {
let primary_key = "id";
let searchable_fields = ["title", "body"];
let filterable_fields = [];
let sortable_fields = [];
let index = setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
);
// We index only one half of the dataset in the setup part
// as we don't care about the time it takes.
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let indexing_config =
IndexDocumentsConfig { autogenerate_docids: true, ..Default::default() };
let mut builder =
IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| ())
.unwrap();
let documents = utils::documents_from(datasets_paths::SMOL_WIKI_ARTICLES, "csv");
builder.add_documents(documents).unwrap();
builder.execute().unwrap();
wtxn.commit().unwrap();
let count = 1250;
let batch_size = 250;
let document_ids_to_delete =
choose_document_ids_from_index_batched(&index, count, batch_size);
(index, document_ids_to_delete)
},
move |(index, document_ids_to_delete)| {
let mut wtxn = index.write_txn().unwrap();
for ids in document_ids_to_delete {
let mut builder = DeleteDocuments::new(&mut wtxn, &index).unwrap();
builder.delete_documents(&ids);
builder.execute().unwrap();
}
wtxn.commit().unwrap();
index.prepare_for_closing().wait();
},
)
});
}
fn indexing_wiki_in_three_batches(c: &mut Criterion) { fn indexing_wiki_in_three_batches(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing wiki in three batches", |b| { group.bench_function("Indexing wiki in three batches", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "body"];
let filterable_fields = [];
let sortable_fields = [];
let index = setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
);
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap(); let mut wtxn = index.write_txn().unwrap();
let mut builder = Settings::new(&mut wtxn, &index, &config);
builder.set_primary_key("id".to_owned());
let displayed_fields =
["title", "body", "url"].iter().map(|s| s.to_string()).collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields = ["title", "body"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
// there is NO faceted fields at all
builder.execute(|_| ()).unwrap();
// We index only one half of the dataset in the setup part // We index only one half of the dataset in the setup part
// as we don't care about the time it takes. // as we don't care about the time it takes.
@ -376,34 +506,21 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
fn indexing_movies_default(c: &mut Criterion) { fn indexing_movies_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing movies with default settings", |b| { group.bench_function("Indexing movies with default settings", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "overview"];
let filterable_fields = ["released_date", "genres"];
let sortable_fields = [];
let config = IndexerConfig::default(); setup_index_with_settings(
let mut wtxn = index.write_txn().unwrap(); &primary_key,
let mut builder = Settings::new(&mut wtxn, &index, &config); &searchable_fields,
&filterable_fields,
builder.set_primary_key("id".to_owned()); &sortable_fields,
let displayed_fields = ["title", "poster", "overview", "release_date", "genres"] )
.iter()
.map(|s| s.to_string())
.collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields =
["title", "overview"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
let faceted_fields =
["released_date", "genres"].iter().map(|s| s.to_string()).collect();
builder.set_filterable_fields(faceted_fields);
builder.execute(|_| ()).unwrap();
wtxn.commit().unwrap();
index
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -424,35 +541,80 @@ fn indexing_movies_default(c: &mut Criterion) {
}); });
} }
fn deleting_movies_in_batches_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing");
group.sample_size(BENCHMARK_ITERATION);
group.bench_function("-movies-delete-facetedString-facetedNumber-searchable-", |b| {
b.iter_with_setup(
move || {
let primary_key = "id";
let searchable_fields = ["title", "overview"];
let filterable_fields = ["released_date", "genres"];
let sortable_fields = [];
let index = setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
);
// We index only one half of the dataset in the setup part
// as we don't care about the time it takes.
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let indexing_config = IndexDocumentsConfig::default();
let mut builder =
IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| ())
.unwrap();
let documents = utils::documents_from(datasets_paths::MOVIES, "json");
builder.add_documents(documents).unwrap();
builder.execute().unwrap();
wtxn.commit().unwrap();
let count = 1250;
let batch_size = 250;
let document_ids_to_delete =
choose_document_ids_from_index_batched(&index, count, batch_size);
(index, document_ids_to_delete)
},
move |(index, document_ids_to_delete)| {
let mut wtxn = index.write_txn().unwrap();
for ids in document_ids_to_delete {
let mut builder = DeleteDocuments::new(&mut wtxn, &index).unwrap();
builder.delete_documents(&ids);
builder.execute().unwrap();
}
wtxn.commit().unwrap();
index.prepare_for_closing().wait();
},
)
});
}
fn indexing_movies_in_three_batches(c: &mut Criterion) { fn indexing_movies_in_three_batches(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing movies in three batches", |b| { group.bench_function("Indexing movies in three batches", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let searchable_fields = ["title", "overview"];
let filterable_fields = ["released_date", "genres"];
let sortable_fields = [];
let index = setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
);
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap(); let mut wtxn = index.write_txn().unwrap();
let mut builder = Settings::new(&mut wtxn, &index, &config);
builder.set_primary_key("id".to_owned());
let displayed_fields = ["title", "poster", "overview", "release_date", "genres"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields =
["title", "overview"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
let faceted_fields =
["released_date", "genres"].iter().map(|s| s.to_string()).collect();
builder.set_filterable_fields(faceted_fields);
builder.execute(|_| ()).unwrap();
// We index only one half of the dataset in the setup part // We index only one half of the dataset in the setup part
// as we don't care about the time it takes. // as we don't care about the time it takes.
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -500,17 +662,11 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
fn indexing_nested_movies_default(c: &mut Criterion) { fn indexing_nested_movies_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing nested movies with default settings", |b| { group.bench_function("Indexing nested movies with default settings", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let mut builder = Settings::new(&mut wtxn, &index, &config);
builder.set_primary_key("id".to_owned());
let searchable_fields = [ let searchable_fields = [
"title", "title",
"overview", "overview",
@ -519,12 +675,7 @@ fn indexing_nested_movies_default(c: &mut Criterion) {
"crew.name", "crew.name",
"cast.character", "cast.character",
"cast.name", "cast.name",
] ];
.iter()
.map(|s| s.to_string())
.collect();
builder.set_searchable_fields(searchable_fields);
let filterable_fields = [ let filterable_fields = [
"popularity", "popularity",
"release_date", "release_date",
@ -540,21 +691,15 @@ fn indexing_nested_movies_default(c: &mut Criterion) {
"crew.name", "crew.name",
"cast.character", "cast.character",
"cast.name", "cast.name",
] ];
.iter() let sortable_fields = ["popularity", "runtime", "vote_average", "release_date"];
.map(|s| s.to_string())
.collect();
builder.set_filterable_fields(filterable_fields);
let sortable_fields = ["popularity", "runtime", "vote_average", "release_date"] setup_index_with_settings(
.iter() &primary_key,
.map(|s| s.to_string()) &searchable_fields,
.collect(); &filterable_fields,
builder.set_sortable_fields(sortable_fields); &sortable_fields,
)
builder.execute(|_| ()).unwrap();
wtxn.commit().unwrap();
index
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -575,19 +720,13 @@ fn indexing_nested_movies_default(c: &mut Criterion) {
}); });
} }
fn indexing_nested_movies_without_faceted_fields(c: &mut Criterion) { fn deleting_nested_movies_in_batches_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing nested movies without any facets", |b| { group.bench_function("-movies-delete-facetedString-facetedNumber-searchable-nested-", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "id";
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let mut builder = Settings::new(&mut wtxn, &index, &config);
builder.set_primary_key("id".to_owned());
let searchable_fields = [ let searchable_fields = [
"title", "title",
"overview", "overview",
@ -596,14 +735,94 @@ fn indexing_nested_movies_without_faceted_fields(c: &mut Criterion) {
"crew.name", "crew.name",
"cast.character", "cast.character",
"cast.name", "cast.name",
] ];
.iter() let filterable_fields = [
.map(|s| s.to_string()) "popularity",
.collect(); "release_date",
builder.set_searchable_fields(searchable_fields); "runtime",
builder.execute(|_| ()).unwrap(); "vote_average",
"external_ids",
"keywords",
"providers.buy.name",
"providers.rent.name",
"providers.flatrate.name",
"provider_names",
"genres",
"crew.name",
"cast.character",
"cast.name",
];
let sortable_fields = ["popularity", "runtime", "vote_average", "release_date"];
let index = setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
);
// We index only one half of the dataset in the setup part
// as we don't care about the time it takes.
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let indexing_config = IndexDocumentsConfig::default();
let mut builder =
IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| ())
.unwrap();
let documents = utils::documents_from(datasets_paths::NESTED_MOVIES, "json");
builder.add_documents(documents).unwrap();
builder.execute().unwrap();
wtxn.commit().unwrap(); wtxn.commit().unwrap();
index
let count = 1250;
let batch_size = 250;
let document_ids_to_delete =
choose_document_ids_from_index_batched(&index, count, batch_size);
(index, document_ids_to_delete)
},
move |(index, document_ids_to_delete)| {
let mut wtxn = index.write_txn().unwrap();
for ids in document_ids_to_delete {
let mut builder = DeleteDocuments::new(&mut wtxn, &index).unwrap();
builder.delete_documents(&ids);
builder.execute().unwrap();
}
wtxn.commit().unwrap();
index.prepare_for_closing().wait();
},
)
});
}
fn indexing_nested_movies_without_faceted_fields(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing");
group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing nested movies without any facets", |b| {
b.iter_with_setup(
move || {
let primary_key = "id";
let searchable_fields = [
"title",
"overview",
"provider_names",
"genres",
"crew.name",
"cast.character",
"cast.name",
];
let filterable_fields = [];
let sortable_fields = [];
setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
)
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -626,39 +845,21 @@ fn indexing_nested_movies_without_faceted_fields(c: &mut Criterion) {
fn indexing_geo(c: &mut Criterion) { fn indexing_geo(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing"); let mut group = c.benchmark_group("indexing");
group.sample_size(10); group.sample_size(BENCHMARK_ITERATION);
group.bench_function("Indexing geo_point", |b| { group.bench_function("Indexing geo_point", |b| {
b.iter_with_setup( b.iter_with_setup(
move || { move || {
let index = setup_index(); let primary_key = "geonameid";
let searchable_fields = ["name", "alternatenames", "elevation"];
let filterable_fields = ["_geo", "population", "elevation"];
let sortable_fields = ["_geo", "population", "elevation"];
let config = IndexerConfig::default(); setup_index_with_settings(
let mut wtxn = index.write_txn().unwrap(); &primary_key,
let mut builder = Settings::new(&mut wtxn, &index, &config); &searchable_fields,
&filterable_fields,
builder.set_primary_key("geonameid".to_owned()); &sortable_fields,
let displayed_fields = )
["geonameid", "name", "asciiname", "alternatenames", "_geo", "population"]
.iter()
.map(|s| s.to_string())
.collect();
builder.set_displayed_fields(displayed_fields);
let searchable_fields =
["name", "alternatenames", "elevation"].iter().map(|s| s.to_string()).collect();
builder.set_searchable_fields(searchable_fields);
let filterable_fields =
["_geo", "population", "elevation"].iter().map(|s| s.to_string()).collect();
builder.set_filterable_fields(filterable_fields);
let sortable_fields =
["_geo", "population", "elevation"].iter().map(|s| s.to_string()).collect();
builder.set_sortable_fields(sortable_fields);
builder.execute(|_| ()).unwrap();
wtxn.commit().unwrap();
index
}, },
move |index| { move |index| {
let config = IndexerConfig::default(); let config = IndexerConfig::default();
@ -680,18 +881,78 @@ fn indexing_geo(c: &mut Criterion) {
}); });
} }
fn deleting_geo_in_batches_default(c: &mut Criterion) {
let mut group = c.benchmark_group("indexing");
group.sample_size(BENCHMARK_ITERATION);
group.bench_function("-geo-delete-facetedNumber-facetedGeo-searchable-", |b| {
b.iter_with_setup(
move || {
let primary_key = "geonameid";
let searchable_fields = ["name", "alternatenames", "elevation"];
let filterable_fields = ["_geo", "population", "elevation"];
let sortable_fields = ["_geo", "population", "elevation"];
let index = setup_index_with_settings(
&primary_key,
&searchable_fields,
&filterable_fields,
&sortable_fields,
);
// We index only one half of the dataset in the setup part
// as we don't care about the time it takes.
let config = IndexerConfig::default();
let mut wtxn = index.write_txn().unwrap();
let indexing_config = IndexDocumentsConfig::default();
let mut builder =
IndexDocuments::new(&mut wtxn, &index, &config, indexing_config, |_| ())
.unwrap();
let documents = utils::documents_from(datasets_paths::SMOL_ALL_COUNTRIES, "json");
builder.add_documents(documents).unwrap();
builder.execute().unwrap();
wtxn.commit().unwrap();
let count = 1250;
let batch_size = 250;
let document_ids_to_delete =
choose_document_ids_from_index_batched(&index, count, batch_size);
(index, document_ids_to_delete)
},
move |(index, document_ids_to_delete)| {
let mut wtxn = index.write_txn().unwrap();
for ids in document_ids_to_delete {
let mut builder = DeleteDocuments::new(&mut wtxn, &index).unwrap();
builder.delete_documents(&ids);
builder.execute().unwrap();
}
wtxn.commit().unwrap();
index.prepare_for_closing().wait();
},
)
});
}
criterion_group!( criterion_group!(
benches, benches,
indexing_songs_default, indexing_songs_default,
deleting_songs_in_batches_default,
indexing_songs_without_faceted_numbers, indexing_songs_without_faceted_numbers,
indexing_songs_without_faceted_fields, indexing_songs_without_faceted_fields,
indexing_songs_in_three_batches_default, indexing_songs_in_three_batches_default,
indexing_wiki, indexing_wiki,
deleting_wiki_in_batches_default,
indexing_wiki_in_three_batches, indexing_wiki_in_three_batches,
indexing_movies_default, indexing_movies_default,
deleting_movies_in_batches_default,
indexing_movies_in_three_batches, indexing_movies_in_three_batches,
indexing_nested_movies_default, indexing_nested_movies_default,
deleting_nested_movies_in_batches_default,
indexing_nested_movies_without_faceted_fields, indexing_nested_movies_without_faceted_fields,
indexing_geo indexing_geo,
deleting_geo_in_batches_default
); );
criterion_main!(benches); criterion_main!(benches);