Skip to main content
En Croissant provides a comprehensive database system for storing and organizing chess games. Built on SQLite with custom indexing, it efficiently handles databases with millions of games.

​
Database capabilities

The database system supports:
  • Fast position search across millions of games using memory-mapped binary indexes
  • Game filtering by player, tournament, date, ELO rating, and result
  • Player statistics with opening repertoire analysis and rating trends
  • Tournament data with event and site information
  • Full-text search for finding games by player name or tournament

​
Supported formats

En Croissant works with industry-standard chess database formats:

​
PGN (Portable Game Notation)

The application can import PGN files in various encodings:
  • Standard .pgn files
  • Compressed .pgn.zst (Zstandard compression)
  • Compressed .pgn.bz2 (Bzip2 compression)
PGN files are converted to optimized SQLite databases with custom binary move encoding for efficient storage and retrieval.

​
Database files

En Croissant uses .db3 (SQLite) database files that include:
  • Games table with move data stored as binary-encoded movetext
  • Players table with name normalization and ELO tracking
  • Events and Sites tables for tournament organization
  • Custom indexes for fast position and game lookup
  • Binary search index (.idx files) for position queries

​
Creating a database

You can create databases in multiple ways:
1

Import from PGN file

Click “Add Database” and select the “Local” tab, then choose a PGN file from your computer. The application will convert it to an optimized database format.
2

Download from web

Browse curated databases in the “Web” tab and click “Install” to download pre-built databases with millions of games.
3

Import from online platforms

Import your games directly from Lichess.org or Chess.com using OAuth authentication (see Importing games).

​
Database operations

​
Viewing database information

Access database details from the Databases page: 
// Database info includes:
// - Title and description
// - Total games, players, and events
// - Storage size on disk
// - Index status (indexed/unindexed)
// - Database file location
You can view this information in the database card on the main Databases page.

​
Managing indexes

Indexes dramatically improve search performance but increase file size:
Creating indexes on a large database can take several minutes but only needs to be done once.
Creating indexes:
  • Enables fast filtering by player, date, rating, and result
  • Required for optimal performance on databases with >10,000 games
  • Automatically created during PGN import
Deleting indexes:
  • Reduces database file size by ~30-40%
  • Useful for archival or when disk space is limited
  • Can be recreated at any time

​
Position search indexing

For position-based queries, En Croissant generates a binary search index (.idx file):
  • Automatic generation when first searching a database
  • Memory-mapped access for instant loading of millions of games
  • Parallel search using all CPU cores for maximum speed
  • Cached results to avoid redundant searches
The search index stores:
  • Material counts (for quick position reachability checks)
  • Pawn structure data (for exact position matching)
  • Move sequences in binary format
  • Player ratings and game results

​
Editing database metadata

Update database title and description through the database settings menu. This information is stored in the database’s internal Info table.

​
Deleting databases

Deleting a database removes both the .db3 file and associated .idx search index file. This operation cannot be undone.
Deleting a database permanently removes all games and cannot be reversed. Consider backing up important databases before deletion.

​
Database maintenance

​
Removing duplicate games

En Croissant can detect and remove duplicate games based on:
  • Same players (White and Black)
  • Same event and site
  • Same date and time
  • Identical move sequence
This is useful when merging multiple PGN sources that may contain overlapping games.

​
Removing empty games

Games with no moves (ply count = 0) can be filtered out to reduce database size and improve data quality.

​
Database optimization

After removing duplicates or empty games, the database automatically:
  • Deletes orphaned player records (players with no games)
  • Deletes orphaned event and site records
  • Updates game counts in the Info table
  • Maintains referential integrity

​
Performance characteristics

Small databases

< 100,000 games
  • Instant loading and search
  • Indexes optional
  • Suitable for personal game collections

Large databases

1M+ games
  • Memory-mapped index required
  • Parallel position search
  • Optimized for master game databases

​
Technical implementation

En Croissant’s database system uses:
  • SQLite with connection pooling (up to 16 concurrent connections)
  • Diesel ORM for type-safe database queries
  • Custom binary encoding for moves, reducing storage by ~60% vs. text
  • Rayon for parallel processing of large imports and searches
  • Memory-mapped I/O for instant access to multi-GB search indexes
Move encoding example from src-tauri/src/db/encoding.rs: 
// Each move is encoded as a single byte when possible
// This includes: square indices, piece type, and special flags
encode_move(&m, &position) // Chess move -> u8
The database schema tracks material changes and pawn structure to enable fast position reachability checks without decoding every move sequence.

Build docs developers (and LLMs) love

Get started for freeTalk to us