This extension provides a UDF and TVF that perform sentiment analysis using VADER.

Usage: sentiment(input TEXT)

Parameters:

• input: The text to analyze.

Returns: A record four fields, each representing a certain sentiment score as a DOUBLE: compound, positive, negative, and neutral.

Description: This function will return a record with four sentiment scores for the given input.

Example:

select to_json(sentiment("i like ice cream"));

+---------------------------------------------------------------------------------------------------------+
| to_json(sentiment("i like ice cream"))                                                                  |
+---------------------------------------------------------------------------------------------------------+
| {"compound":0.3611575592573076,"negative":0,"neutral":0.4444444444444444,"positive":0.5555555555555556} |
+---------------------------------------------------------------------------------------------------------+

Usage: sentiment_tvf(input TEXT)

Parameters:

• input: The text to analyze.

Returns: A table with four columns, each representing a certain sentiment score as a DOUBLE: compound, positive, negative, and neutral.

Description: This function is identical to the UDF above, except that it will return a table with the four sentiment scores as columns.

Example:

select * from sentiment_tvf('i like ice cream');

+--------------------+--------------------+----------+--------------------+
| compound           | positive           | negative | neutral            |
+--------------------+--------------------+----------+--------------------+
| 0.3611575592573076 | 0.5555555555555556 |        0 | 0.4444444444444444 |
+--------------------+--------------------+----------+--------------------+

To install these functions using the MySQL CLI, you can use the following command. Replace $DBUSER, $DBHOST, $DBPORT, and $DBNAME with, respectively, your database username, hostname, port, and the name of the database where you want to deploy the functions.

mysql -u $DBUSER -h $DBHOST -P $DBPORT -D $DBNAME -p < load_standalone.sql

The example directory provides a data set (YouTube trending data from 2020) that can be used to explore this function. To load it, simply source the examples/load.sql file.

Then, you can run sentiment analysis like this:

select y.video_description, s.compound from youtube y, sentiment_tvf(y.video_description) s;

The load_* scripts in this repo should be sufficient to load the pre-built extension into SingleStore. However, if you wish to rebuild it, do the following.

1. Install Rust and Cargo
2. Install the Rust wasm32-wasi target
3. Run cargo build --target wasm32-wasi

To build the extension from the Wasm locally, run the following.

cp ./target/wasm32-wasi/release/json-flatten.wasm ./json-flatten.wasm
tar cvf json-flatten.tar json-flatten.sql json-flatten.wasm json-flatten.wit 

There are automated script tests in the db-tests directory. They are run against the singlestoredb-dev-image.

In order to run the tests

1. Install Python 3
2. Install singlestoredb[pytest] (and optionally pytest-xdist) using pip
3. Set the SINGLESTORE_LICENSE environment variable
4. Run the pytest command

If you installed pytest-xdist, you can also distribute the tests to multiple workers by running pytest -n auto or giving a specific number instead of auto

This repo repackages code originally written by Chris Wong (https://github.com/ckw017/vader-sentiment-rust). Thanks, Chris!