Creating APIs‎ > ‎Logic‎ > ‎

Logic Patterns

Explore the design patterns of reactive logic in the context of running examples, such as those provided in the reactive logic tutorial. Live API Creator also provides examples, so you can experiment with real transactions, and study the logs, for example.

You can use the following table of contents as a checklist of the design patterns.

Concept: Think Spreadsheet

You can write JavaScript event handlers to handle table updates. Experienced application developers will understand them instantly. The key to getting value out of API Creator is to use events with spreadsheet-like rules. This concept is called "think spreadsheet." A good way to understand it is by way of the reactive logic tutorial examples, which illustrate the design patterns for reactive logic. The basic idea is that derivations are like spreadsheet cell formulas in that they set up reactive logic.

Watch, React, and Chain: Reactive logic watches for changes to referenced data and adjusts the derived data accordingly. This can chain when the adjusted data is itself referenced by other derivations.

For more information:

Access Related Data

Access other values in the same row using formulas. Often useful, but most interesting transactions involve data from a related table. Consider the following example:

Multi-table derivations are automated by rules based on relationships (either foreign keys that you define in your database or virtual foreign keys):

  • Parent access to child data. When you define aggregates (sum, count, min, max), the child watches for changes to referenced child values, and, when necessary, adjusts the parent.
  • Child access to parent data. When a child formula or validation references parent data (for example, parent.attributeName), the parent watches for changes to attributeName and propagates changes to all the child rows. This propagation does not occur for parent copy derivations.
For more information:

Forward Chaining: Constraining Chained Derivations Pattern

Probably the most common pattern is defining a validation that requires a series dependent derivations to produce the data used in the validation. The most familiar example is Check Credit, where the Customer's balance is derived by a series of derivations over three domain objects.

Access data from related domain objects using business logic services:
  • You can enable parent objects to aggregate related child data using sum and count rules. Automatic adjustment processing occurs when child rows are changed in such a manner.
  • You can create parent references using formula rules. Automatic cascade processing occurs when the parent row's reference data is changed.
For more information:

Counts for Existence Checks

The count derivation can produce a count of related child rows, optionally qualified by a condition. In many cases, the objective is to know whether the count is 0 or non-zero - that is, are there any [qualified] children. For example:
  • The No Empty Orders requirement is implemented by a countItems, which is then checked in a validation.
For more information about this requirement, see No Empty Orders Example.
  • The Bill of Materials Explosion (see rules 1 and 2) requires identifying whether a Product is a kit and has any components.
For more information about this example, see the Reactive Logic Tutorial.

Replicate Junction Pattern

The relational model supports many-to-many relationships by introducing junction tables: a table with foreign keys to both end-points. For example, in the sample database, orders can have many Products, and a Product can be ordered on many Orders. The Junction Lineitem has foreign keys to both, as well as additional attributes (such as QuantityOrdered).

You can introduce junction tables for many-to-many relationships between two endpoint objects. It is often a requirement that an endpoint object needs to sum/count an attribute from the other endpoint. Since sums/count operate on 1:n relationships and not n:m relationships, replicate the summed attribute into the junction.

The replicate can take the following forms, depending on whether the sum should reflect changes in the summed attribute. The Reactive Logic Tutorial illustrates both.

  • Reflect summed changes into sum: use Parent Reference

The Bill of Materials Price Rollup example illustrates is a many-to-many relationship between Product objects, informally referred to as kit and components. ProductBillofmaterials is the Junction Entity implementing the many-to-many relationship. A kit needs to sum its components price to determine its own price.

Since the Business Requirement is that Component price changes be reflected in the Kit Price. We define ProductBillofmaterials.value as kitNumberRequired * product.price. We can then sum this result into the kit. This formula contains a Parent Reference to the product.price. Changes to parent references are cascaded to child rows, unlike @copy logic as shown in the next example.

  • Do not reflect summed changes into sum: use @ParentCopy

The Place Order example illustrates a many-to-many relationship between Orders and Products, where Lineitem is the Junction entity. Orders.amountTotal is the sum(Lineitems.amount).

Our business requirement is that we do not wish to change Purchaseorder amountTotals for subsequent Product price changes. We therefore define Lineitem.amountTotal as an @Copy of Product.price.

For more information:

Request Objects

The classic request pattern, or command pattern, invokes behavior from the creation of command objects. These are commonly used in text editors, for example to turn a set of characters bold or italic. You can maintain these objects in a list and use them for undo, for example.

You can use this same pattern in data processing applications. You insert the request object, which represents a transaction requesting some behavior, such as a Credit Request, or give an Employee a raise, call a series of external services, or run scheduled tasks in background. Instead of code, you implement the desired behavior by declaring logic.

You can extend the request pattern to non-database use cases, for example commands.

For more information:

Form Flow based on Derivation Results

It is a common requirement to drive form flow based on derivation results. For example, you might define a wizard for document processing, with tab sheets enabled/activated on the basis of whether restricted Products are ordered, or the customers credit history.

You can compute the data used to drive these conditional transitions using derivation mechanisms. Derivations fire only when you save a transaction. We recommend using the Ready Pattern.

Ready Pattern

Imagine an interactive transaction where the end user proceeds through a set of screens. When screen-1 is complete, business logic derives various data. The presentation layer uses these derivations for display and for conditional processing (for example, a conditional transition, or to cause hiding of not-relevant portions of the screen).

It is undesirable to hold an open database transaction (and blocking locks) over screen transitions. Given that you need the business logic derivation results, this suggests that you should submit and commit changes at the completion of screen-1. But this presents a problem. It might not be appropriate to "post" this incomplete transaction (for example, adjust General Ledger, the Customer Balance, Product stock) until you have completed all of the screens. This is a common scenario in online shopping. You can fill your cart and see how much things cost (a derivation) as you shop. It is only when you checkout that your account is charged.

Best Practices:

  • Define an isReady attribute (for example, in Purchaseorder). Its value is false until the checkout screen, at which point it is set to true.
  • When defining rules, predicate the rules on isReady == true as in the "Place Order" business transaction:
Derive Customer.balance as Sum (purchaseorders.amountUnPaid where isReady=true)
  • Provide a Make Ready (aka "checkout") transaction to activate the remaining logic.

For more information about how to provide a Make Ready transaction, see the Reactive Logic Tutorial.

Commit Time Logic: Validations, Actions

Logic processing begins with the logic phase (initial row processing) and ends with the commit phase (after all rows are processed). Counts of inserted child rows are always zero (0) during the logic phase. Requirements such as No Empty Order (which depend on the count value) need to use commit validations.

For more information:

Parameterized Formulas

You can enable business users to alter data used in formulas. For example, you might apply a discountAmount to a purchase order.

  1. Create a "scalar" domain object (for example, Parameters), with attributes you want authorized business users to change, such as discountAmount.
  2. Provide access to this object in your applications' user interface.
  3. Extend LogicBase (see BusLogicIntro) by defining your logic classes, where LogicBase implements the method to return the single Parameters object.
    You can use getParameters() in formulas (just as you can parent references), for example, getParameters().getDiscountAmount().
  4. Create a one-row domain object (for example, Parameters).
  5. Provide getParameters() in superclass of Logic Classes.
  6. Use LogicBase in formulas.

Data Integration

Resources to define Partner object mappings

You can interact with other systems, both as a receiver and as a sender, using the RESTful API mechanism. You employ the JavaScript aspects of reactive logic.

For more information about how to employ the JavaScript aspects of reactive logic, see Reactive Logic Tutorial.

Metadata tags for Matching

You can match incoming data to target data using metadata action tags, for example, to locate parent data or insert/update master data.

For more information about how to match incoming data, see Complex Transaction Processing.

Group By

You can store subtotals that are incrementally maintained as updates occur, using the Group By Pattern. For example, you might wish to maintain total sales for each sales rep for each month.

For more information:

Extensibility

The following patterns are common patterns implemented by the extension package. They illustrate how to provide logic extensions.

Allocation

The allocation pattern can be summarized as follows:

Allocates a provider amount to designated recipientscreating allocation objects (a Provider/Recipient Junction) for each such allocation. 

For more information about the allocateFromTo system method, see allocateFromTo System Method.

Insert From (Copy)

It is a common requirement to copy one or more source rows to target rows. A corollary requirement is that the created target rows are usually children of the source rows, and require proper initialization of their attributes. 

For more information about the InsertIntoFrom system method, including the auditing, deep copy, and bill of materials explosion services, see InsertIntoFrom System Method.

Inclusion/Exclusion Logic

You can search through a list to find rows that match a specified criteria using the FindWhere system method. You can search for the first row that matches or the only row (which throws an exception if there are multiple matches).

For more information about the FindWhere system method, see findWhere System Method.

Subpages (1): Command Pattern