This is a proof of concept of using n‑grams for partial string match in Azure Cosmos DB.
Since Cosmos DB does not support efficient partial text match out of the box, the idea was to test if n‑grams can be used for this purpose in a read-heavy application instead of having to connect to another external service like Azure Cognitive Search.
| Method | Mean | Avg. RU charge |
|---|---|---|
| AddItem_Plain | 24,723.0 us | 7.81 |
| AddItem_NGram | 98,705.9 us | 125.14 |
| QueryItem_Plain | 133,065.5 us | 207.74 |
| QueryItem_NGram | 938.9 us | 2.99 |
| Container | Document Count | Container Size | Avg. Document Size |
|---|---|---|---|
| UsersPlain | 100,000 | 120,925 KiB | 0.95 KiB |
| UsersNGram | 100,000 | 1,273,532 KiB | 12.47 KiB |