Need to search through your data? Heard about Meilisearch?

Transcript

1. Need to search through your data? Heard about Meilisearch?
2. Greetings
3. Hellcome Guillaume Loulier / @Guikingone Lead developer @SensioLabs Works mainly with PHP and Rust, just discovered Gleam Love motorsport, especially rallying because sliding in the snow is fun
4. Summary 01 - Searching is hard, to say the least 02 - Improving the wheel 03 - What about a demo? 04 - Tomorrow can wait 01
5. Searching is hard
6. To say the least
7. Searching the needle in the haystack We, as humans, split our knowledge since we’re born Knowledge is seen as a resource to index, split, filter, explore, learn from We, as human being, learn by splitting and exploring possibilities
8. Storing the world knowledge Storing for a long time is hard, ever heard of The Great Library of Alexandria? At the end, we mostly deal with indexes, documents and data types Indexing introduced a better learning experience / curve
9. Meilisearch
10. Improving the wheel
11. Experimenting Build for a specific use case at LVMH First iteration built in Go, rewritten in Rust, speaking of speed (< 50ms) Open-source (MIT) and french PHP SDK, Symfony bundle and many more
12. Built for tomorrow Support for “real-time” search, multi-search, geosearch, vector search, faceted search and search preview Can be fully synchronized with your database(s) via Meilisync Stop words, synonyms, automatic language detection (support for most of the languages), typo-tolerant, dashboard
13. But, how does it work?
14. Talking about architecture Built around JSON document structure (but also supports NDJSON and CSV), schemaless Automatic primary key detection, index settings and index swapping Up to 80 TiB index size, stored via LMDB and memory map Default document size to 100MB, can be changed
15. Bucket style Crawl a document, split each field then recursively (tokenizer phase via charabia) run pipelines Once a query is made, a bucket sort is applied to rank documents By default, Meilisearch defines a set of rules already applied during the sort phase Custom rules can be created / configured if needed
16. Show me
17. Fear none Built on top of symfony/demo Available via https://github.com/Guikingone/SymfonyConMeilisearch
18. Tomorrow can wait
19. Sell me this engine Full open-source (what about Elasticsearch / Algolia ?) No runtime dependencies, scheduled dumps, snapshots Support for one-click deployments 12 factor app? Of course
20. Losing grip High availability only via the cloud Symfony bundle opinionated Asynchronous operations No support for custom grouping, no built-in support for media, no embedded document querying
21. That’s all