Lesson 24: Executing SQL During App Deployment

App Use Case Scenario

In Lesson 8, we introduced a feature that automatically creates stock items when none are present in the system. This was done to avoid having to create these default stock items each time the app is newly deployed. The invocation for this logic was included in the invite function for the "System Admin" role. This means that when a System Admin is invited to the app using the Helium core web app or the HeliumDev client, the default stock items are created if they do not already exist.

A better way to achieve the same result is to have the default records created during deployment instead of during user invitation. This can be achieved by packaging SQL for this purpose with the DSL source code.

In addition to this use case we can also make use of the same Helium feature to provide a mechanism by which to efficiently report on all purchases in the system. For this we will include a SQL script that drops an existing table, and creates a materialized view in its place. Queries can then be performed directly on the newly created view by using the native DSL SQL features or selectors.

Packaging SQL Script Files With App Source Code

SQL scripts are packaged with DSL app source code by including them in a folder named sql-scripts. This folder needs to be in the app root directory. This folder is processed recursively for any files with a .sql extension. This means that you can have files in nested folders with no limit on the depth. As long as the correct file extension of .sql is used, these files will be included in your app source code that gets sent to the Helium server.

Script Execution

Order of Script Execution

Scripts are executed in alphabetical order of the path to that file. In other words, when execution order is determined, the path to that file in your project and the file name is taken into account. It is highly recommended that a strict naming convention and folder structure is followed when adding SQL script files to your project as out of order execution might result is SQL errors which will cause the app deployment to fail. One example of such a naming scheme is to prepend the file name with the date on which it was added for example:

2017-09-12 create_default_records.sql

2017-09-18 update_default_records.sql

Note that Helium will attempt to execute the scripts each time the app is deployed. SQL scripts should therefore make provision for this to avoid SQL errors or the execution of SQL that might lead to inconsistent app data. This is discussed along with examples in following sections.

Script Errors & Debugging

If a SQL script that is packaged with app source fails to execute, the error along with file name will be reported in the HeliumDev client. The packaged SQL scripts are seen as part of the app model and is executed as part of the schema upgrade for existing apps and schema creation for new apps. For this reason a script that results in an error will cause the entire deployment to fail and be rolled back. Once the script has been fixed the deployment can proceed. Also note that no SQL select queries should be packaged with app source code. Helium executes SQL scripts without expecting any result from the query. For this reason a select query will result in a failed schema upgrade and thus a failed deployment.

Be sure to follow a strict naming convention for SQL scripts that will ensure the correct order of execution.

Initial App Data Using SQL Scripts

In order to create default app data for the stock object we simply include a SQL script file in our sql-scripts folder with insert statements for the default data. Keep in mind that the script will be run regardless of whether there is already data for default stock items or not. The SQL script itself should therefore make provision for this. See the two statements below for examples on how this can be achieved:

One way of achieving this is to use of a conditional insert that makes use of NOT EXISTS.

INSERT INTO stock (_id_, _tstamp_, name, stocktype, deleted)
SELECT '80dc5655-9600-440b-86c0-614ccaef11fe', current_date, 'Corn', 'crop_seed', false
WHERE NOT EXISTS (
    SELECT _id_ FROM stock WHERE _id_ = '80dc5655-9600-440b-86c0-614ccaef11fe'
);

Another, more preferable, way of achieving the same result is to make use of the so called upsert functionality provided in PostgreSQL 9.5. Using this, the desired behaviour on insert conflicts can be specified. In this case the statements are as follows:

 INSERT INTO stock (_id_, _tstamp_, name, stocktype, deleted) VALUES ('80dc5655-9600-440b-86c0-614ccaef11fe', current_date, 'Corn', 'crop_seed', false) ON CONFLICT DO NOTHING;

Materialised View for Reporting Using SQL Scripts

Data Model Additions

In order to make use of a materialized view for reporting purposes, we first create a matching persistent object in our app. After deployment, this will be dropped and a materialised view will be created in its place. The result is an object that can be queried in the same manner as a normal object using selectors but where the data is backed by our newly created materialised view.

persistent object FarmerPurchaseSummary{
    
    // When was the purchase made
    datetime purchasedOn;
    
    // Quantity and cost of purchase
    int quantity;
    decimal unitPrice;
    decimal goodsCost;
    decimal discount;
    int finalCost;
    
    // Helium provided payment status and id
    datetime paymentStatusUpdatedOn;
    PAYMENT_STATUS paymentStatus;
    uuid paymentId;
    
    // Stock details
    string stockName;
    string stockType;
    
    // Farmer details
    string farmerNames;
    string farmerMobileNumber;
    
    // Shop details
    string shopCode;
    string shopName;
}

SQL Script

The following script is used to drop the existing table or materialised view and recreate it. Note the use of the __he_delete_table_or_view__ function. This function is provided by Helium as part of the app schema in order to assist with use cases such as this. This is required for the following reason:

When the script is first run, the existing table needs to be dropped. Using "drop table if exists" will achieve this. After the table is dropped, the materialised view is created. For subsequent deployments, however, the script will again attempt to drop the table even though the table does not exists, while a view with the same name does exist. This will result is a SQL error and the schema upgrade and thus deployment will fail.

DO $$ BEGIN
    PERFORM __he_delete_table_or_view__('farmerpurchasesummary');
END $$;
CREATE MATERIALIZED VIEW farmerpurchasesummary AS 
SELECT 
	farmerpurchase._id_, 
	current_date, 
	purchasedon, 
	quantity, 
	unitprice, 
	goodscost, 
	discount, 
	finalcost, 
	paymentstatusupdatedon, 
	paymentstatus, 
	paymentid, 
	stock.name AS stockname, 
	stock.stocktype AS stocktype, 
	farmer.firstname || ' ' || farmer.lastname AS farmernames, 
	farmer.mobilenumber AS farmermobilenumber, 
	shop.shopcode AS shopcode, 
	shop.name AS shopname 
FROM farmerpurchase 
LEFT JOIN stock on farmerpurchase.stock_fk = stock._id_ 
LEFT JOIN farmer on farmerpurchase.farmer_fk = farmer._id_
LEFT JOIN shop on farmerpurchase.shop_fk = shop._id_;

View & Presenter Additions

In addition to the above, we also add a new view with a data table and a backing presenter to display the data from our view. See the lessons source code for the complete example.

Lesson Outcomes