33 lines
846 B
Markdown
33 lines
846 B
Markdown
|
# Tantivy ADR
|
||
|
|
|
||
|
|
**Author**
|
||
|
|
Marko Budiselic (github.com/gitbuda)
|
||
|
|
|
||
|
|
**Status**
|
||
|
|
APPROVED
|
||
|
|
|
||
|
|
**Date**
|
||
|
|
January 5, 2024
|
||
|
|
|
||
|
|
**Problem**
|
||
|
|
|
||
|
|
For some of Memgraph workloads, text search is a required feature. We don't
|
||
|
|
want to build a new text search engine because that's not Memgraph's core
|
||
|
|
value.
|
||
|
|
|
||
|
|
**Criteria**
|
||
|
|
|
||
|
|
- easy integration with our C++ codebase
|
||
|
|
- ability to operate in-memory and on-disk
|
||
|
|
- sufficient features (regex, full-text search, fuzzy search, aggregations over
|
||
|
|
text data)
|
||
|
|
- production-ready
|
||
|
|
|
||
|
|
**Decision**
|
||
|
|
|
||
|
|
All known C++ libraries are not production-ready. Recent Rust libraries, in
|
||
|
|
particular [Tantivy](https://github.com/quickwit-oss/tantivy), seem to provide
|
||
|
|
much more features, it is production ready. The way how we'll integrate Tantivy
|
||
|
|
into the current Memgraph codebase is via
|
||
|
|
[cxx](https://github.com/dtolnay/cxx). **We select Tantivy.**
|