wabidb-performance-benchmarks

name	wabidb-performance-benchmarks
description	Learn WabiDB's benchmark suite — 24 criterion benchmarks measuring projection read performance across get, list, and compact operations on 10k records.

WabiDB Performance Benchmarks

This skill covers WabiDB's Criterion-based benchmark suite in benches/projection_read.rs, which measures the read throughput of all major projection query methods.

When to Use

• Measuring the performance impact of changes to projection query methods
• Regression-testing read latency after adding new projection features
• Understanding the baseline performance characteristics of projection reads

Prerequisites

• Access to WabiDB source code at /var/home/Ronin/wabi
• Rust toolchain with nightly (for cargo bench)

How to Run

# Run all benchmarks
cargo bench -p wabidb

# Run only projection benchmarks
cargo bench -p wabidb -- projection_read

# Run a specific benchmark group
cargo bench -p wabidb -- "projection_get/"

# Run a single benchmark
cargo bench -p wabidb -- "projection_get/get_message"

Key File

benches/projection_read.rs (381 lines) — the entire benchmark suite.

Dataset

The populate() function creates 10k records across 100 groups (100 records per group) for each of 10 projection types:

Projection	Index Name	Records
messages	messages	10,000
channel_members	channel_members	10,000
dm_messages	dm_messages	10,000
dm_message_recipients	dm_message_recipients	30,000 (3 recipients per msg)
reactions	reactions	500 (5 emoji reactions on first msg per group)
wiki_pages	wiki_pages	10,000
forum_posts	forum_posts	10,000
incidents	incidents	10,000
albums	albums	100 (1 per group)
album_items	album_items	10,000

Approximately 10% of records are marked is_deleted = true for compaction benchmarks.

Benchmark Groups

projection_get (10 benchmarks)

Measures single-record lookup by exact key:

Benchmark	Method	Key
get_message	MessagesProjection::get_message()	channel_id + message_id
get_member	ChannelMembersProjection::get_member()	channel_id + user_id
get_dm_message	DmMessagesProjection::get_message()	dm_id + message_id
get_recipient	DmMessageRecipientsProjection::get_recipient()	dm_id + message_id + user_id
get_reaction	ReactionsProjection::get_reaction()	message_id + user_id + emoji
get_page	WikiProjection::get_page()	channel_id + page_id
get_post	ForumProjection::get_post()	channel_id + thread_id + post_id
get_incident	IncidentProjection::get_incident()	channel_id + incident_id
get_album	AlbumProjection::get_album()	scope_type + scope_id + album_id
get_album_item	AlbumItemsProjection::get_item()	album_id + item_id

projection_list (13 benchmarks)

Measures range scans with prefix lookup:

Benchmark	Method	Notes
list_messages	list_messages(ch, false)	Excludes deleted
list_messages_include_deleted	list_messages(ch, true)	Includes deleted
list_members	list_members(ch)	All members in channel
list_dm_messages	list_messages(dm_id)	All DMs in conversation
list_recipients	list_recipients(dm_id, msg_id)	All recipients of a DM
list_reactions	list_reactions(msg_id)	All reactions on a message
list_pages	list_pages(ch, false)	Wiki pages excluding deleted
list_posts	list_posts(ch, thread, false)	Forum posts in thread
list_threads	list_threads(ch, false)	Thread starters in channel
list_incidents	list_incidents(ch, false)	Incidents excluding deleted
list_albums	list_albums(scope_type, scope_id, false)	Albums in scope
list_album_items	list_items(album_id, false)	Items in album

projection_compact (1 benchmark)

Benchmark	Method	Effect
compact_messages	MessagesProjection::compact()	Removes ~1000 deleted records

Configuration

criterion_group! {
    name = projection_read;
    config = Criterion::default().sample_size(20).measurement_time(Duration::from_secs(3));
    targets = bench_get, bench_list, bench_compact
}

• 20 samples per benchmark
• 3-second measurement time
• Each benchmark uses black_box() to prevent compiler optimizations

Adding a New Benchmark

1. Import the projection module at the top of the file
2. Add data to the populate() function
3. Add a bench_* function or extend an existing one
4. Register in the criterion_group! macro