CREATE#

CREATE DATABASE#

The CREATE DATABASE statement allows us to connect to an external structured data store in EvaDB.

CREATE DATABASE [database_connection]
     WITH ENGINE = [database_engine],
     PARAMETERS = [key_value_parameters];

[database_connection] is the name of the database connection. [database_connection].[table_name] will be used as table name to compose SQL queries in EvaDB.
[database_engine] is the supported database engine. Check supported data sources for all engine and their available configuration parameters.
[key_value_parameters] is a list of key-value pairs as arguments to establish a connection.

Examples#

CREATE DATABASE postgres_data WITH ENGINE = 'postgres', PARAMETERS = {
     "user": "eva",
     "password": "password",
     "host": "localhost",
     "port": "5432",
     "database": "evadb"
};

CREATE TABLE#

To create a table, specify the schema of the table.

CREATE TABLE IF NOT EXISTS MyCSV (
                id INTEGER UNIQUE,
                frame_id INTEGER,
                video_id INTEGER,
                dataset_name TEXT(30),
                label TEXT(30),
                bbox NDARRAY FLOAT32(4),
                object_id INTEGER
 );

CREATE INDEX#

The CREATE INDEX statement allows us to construct an EvaDB based index to accelerate semantic based searching. The index can be created on either a column of a table directly or outputs from a function running on a column of a table.

CREATE INDEX [index_name]
   ON [table_name] ([column_name])
   USING [index_method]

CREATE INDEX [index_name]
   ON [table_name] ([function_name]([column_name]))
   USING [index_method]

[index_name] is the name the of constructed index.
[table_name] is the name of the table, on which the index is created.
[column_name] is the name of one of the column in the table. We currently only support creating index on single column of a table.
[function_name] is an optional parameter that can be added if the index needs to be construsted on results of a funciton.

Examples#

CREATE INDEX reddit_index
ON reddit_dataset (data)
USING FAISS

CREATE INDEX func_reddit_index
ON reddit_dataset (SiftFeatureExtractor(data))
USING QDRANT

You can check out similarity search use case about how to use index automatically.

CREATE FUNCTION#

To register an user-defined function, specify the implementation details of the function.

CREATE FUNCTION IF NOT EXISTS FastRCNNObjectDetector
INPUT  (frame NDARRAY UINT8(3, ANYDIM, ANYDIM))
OUTPUT (labels NDARRAY STR(ANYDIM), bboxes NDARRAY FLOAT32(ANYDIM, 4),
        scores NDARRAY FLOAT32(ANYDIM))
TYPE  Classification
IMPL  'evadb/functions/fastrcnn_object_detector.py';

CREATE FUNCTION via Type#

CREATE [OR REPALCE] FUNCTION [IF NOT EXISTS] function_name
[ FROM ( select ) ]
TYPE function_type
[ parameter [ ...] ]

Where the parameter is key value pair.

Warning

For one CREATE FUNCTION query, we can specify OR REPLACE or IF NOT EXISTS or neither, but not both.

Note

Go over HuggingFace Models, Model Training with Ludwig, and Time Series Forecasting to check examples for creating function via type.

CREATE MATERIALIZED VIEW#

To create a view with materialized results – like the outputs of deep learning model, use the following template:

CREATE MATERIALIZED VIEW UADETRAC_FastRCNN (id, labels) AS
SELECT id, FastRCNNObjectDetector(frame).labels
FROM UADETRAC
WHERE id<5;