analyze-text-async

analyze-text-async is a demo app exposing REST API for analyzing texts using Vert.x

Explanation

The app exposes an endpoint at /analyze which expects a JSON object with a string property text containing letters. The endpoint returns a json object with two string properties containing one of the previously sent texts:

• value: the text closest in terms of the sum of all character values of the text, where character values are listed as a = 1, b = 2 and so on, in a non-sensitive case way. If there are ties, the higher value is picked. If there are ties again, the higher sorted in ascendent lexical order is picked.
• lexical: the text closest in terms of lexical closeness, in a non-sensitive case way. If there are ties, the lower sorted in ascendent lexical order is picked.

The server persists the texts to a postgres database. It does so after it returns the result, for better latency, it also means the text may not be persisted.

The server caches the texts already sent, so it does not go to the database in a request. Persisted texts are loaded on start up.

The closest value and lexical texts are calculated concurrently, outside the verticle event loop. Binary search are used to find them.

Building

To build the app run mvn clean package. This creates a fat jar at ./target.

Running

The server uses the following environment variables:

• PORT: port the server listens. Default 8888.
• PGHOST: host of postgres database. Default localhost.
• PGPORT: port of postgres database. Default 5432.
• PGDATABASE: database of postgres database. Default postgres.
• PGUSER: user of postgres database. Default postgres.
• PGPASSWORD: password of postgres database. Default postgres.
• PGCONNECTIONS: number of pooled connections. Default 20.

After the database is up, run the fat jar with java -jar target/*-fat.jar or mvn exec:java to start the server.

There is docker-compose.yaml that sets up a postgres database, exposes it at port 5432, starts the server and exposes it at port 8888. Run docker-compose up to deploy.

Testing

To test the API, send requests to /analyze, for example using curl:

$ curl localhost:8888/analyze -d '{"text":"word"}'

There is also a K6 script to run stress tests on the server localhost:8888, run the following command:

$ docker run --network="host" -i grafana/k6 run --vus 10 --duration 30s - < tests/k6/random-text.js

It continuously sends random texts to the server for 30 seconds on 10 different virtual users. At the end, K6 reports statistics about the API.