SmartApps Data handlers

Data handlers: the app's connection to rows, in the system that owns them.

A data handler provides data to an app as rows with columns, and takes rows back.

SCREENSDATA HANDLERSSOURCE OBJECTS, READ IN PLACEScreens and componentsscreenExpressions and criteriascreenDecision points in the flowscreenVariablesscreenTable handlerfederated table, e.g. EKPOQuery handlersaved query across tablesPipeline handlerBackbone result setAPI / file / feed handlerREST, CSV, IoTSAP · OracleERP objects, in placeSnowflake · Databrickshistory and aggregatesSalesforce · ServiceNow · Workdayapplication objectsFiles, APIs, devicesthe restrows stay in the source · access rules apply at read time · writes go through the source's released interface and are confirmed
Everything that consumes data in an app goes through a handler, and every handler reads in place. The four handler kinds cover tables, queries, prepared results and everything else.
5
kinds of handler: table, query, pipeline, API or file, device feed
product
4
operations: read, update, save, delete, each through the source's released interface
product
60
connectors in the library, across ERPs, planning, lakes, databases, files, streams and applications
connector library
2.1 s
for a question across SAP, Oracle, the WMS and a 3PL API on the showcase estate
showcase estate
0
rows copied to build an app
product
The five kinds

Choose by where the rows come from

HandlerDefined onReadsWritesTypical use
Table handlerA federated table registered in the catalog: an ERP table, a lake table, an application objectThe rows that match the filter, with the columns the screen needs, in placeThrough the source's released interface: BAPI, REST, MERGEOrder lines, vendors, stock, claims, tickets
Query handlerA saved query across one or several sources, with parametersThe query result, computed in the sources with pushdownUsually read-only; writes go to the underlying tablesStock by plant, consumption against contract, history by month
Pipeline handlerThe result of a Backbone pipeline or a data-science pipelineA prepared, checked result set, refreshed on the pipeline's scheduleResults can be written to a target the pipeline ownsProjected stock, supplier scores, forecast error, a model's prediction
API or file handlerA REST endpoint, a CSV or Excel file, a documentRows parsed from the answer or the fileCalls the endpoint; writes a fileScreening services, carrier APIs, uploads from suppliers
Device feed handlerAn IoT or event streamThe latest values or a windowCommands where the device allowsTemperature on a batch, a scan at a dock, a counter on a line

A table handler is the workhorse: it is the federated table itself, seen through a filter and a column list, so a screen on it shows live rows and a write on it goes to the source. A query handler saves a question, for example stock by plant and material with its cover in weeks, and lets screens and criteria reuse it with parameters. A pipeline handler is for results that take preparation: the Backbone lands them, checks them and refreshes them on a schedule, and the app reads the prepared rows.

API, file and device handlers cover the rest of the world: a sanctions screening service, a carrier's tracking API, a spreadsheet a supplier uploads, the temperature feed on a cold room. Their rows look like any other rows to the screens and the expressions.

Handlers are registered once per app, with their operations and their audit information, and listed on a screen that shows where each one is used. A handler can be shared between apps through the catalog, so a vendor handler defined for onboarding serves the scorecard and the invoice apps too, with its lineage recorded.

The catalog is the same one the Data Fabric and the Backbone use, so a table registered for a pipeline is available to an app the same day, and the lineage of an app's screen shows the source columns it came from.

Read in place

What happens when a screen opens

The screen asks the handlerFor the rows matching its filter (the plant, the buyer, the status) and the columns its components need.

Because the handler reads in place, an app is possible on the day the source is connected, and it shows what the source shows. The Data Fabric page explains federated queries and pushdown; the point for an app builder is simpler: a screen is a filter on a handler, and the source does the work.

Where a screen needs history at scale, a year of stock positions or five years of journals, the pipeline handler reads it from the lake where the Backbone landed it. The catalogue's pair pages say, per app and per system, which objects are read live and which come prepared.

Write back

What happens when a person decides

A write on a handler goes through the source's released interface: a BAPI or IDoc for SAP, a REST resource for Oracle Fusion, Salesforce and ServiceNow, a web-service operation for Workday, a MERGE for a lake table. The handler maps the screen's values to the interface's fields, calls it under the acting person's identity and the app's grants, waits for the acknowledgement, and stores the document number or the error message beside the decision in the log.

Write-back is off by default per application and object. It is switched on after a confirmed test write in the target, and each object can carry an approval rule: a price change above a band waits for a second person before the interface is called. The Backbone write-back page has the ledger, the reconciliation and the SAP and Oracle interfaces in detail.

SourceWrite pathAcknowledgement
SAP S/4HANA and ECCReleased BAPIs and IDocsDocument number and return messages
Oracle Fusion CloudREST resources (POST, PATCH)Resource ID and status
Oracle E-Business SuiteOpen interface tables and public APIs, run by a concurrent programRequest ID and interface status
Snowflake · DatabricksStage and COPY, MERGE INTO a Lima-owned tableRows merged
Salesforce · ServiceNowREST API under the integration userRecord ID
WorkdayWeb-service operationsWorkday ID
In expressions and decisions

The same handler decides the path

A handler is available wherever an expression is: a start criterion selects rows from it; a validation compares a field to it; a decision node asks it whether the new date covers demand; an action reads a row from it or writes one to it. Variables can be instantiated from a handler at load, so an app variable holds the customer's credit limit from the first screen onward.

This is what makes an app one thing: the rows a person sees, the rule that routes their decision and the write that records it all speak to the same handler.

# decision node: does the supplier's new date cover demand?
let need = dh_planned_orders.min(start) where material = app.material and plant = app.plant
if app.new_date <= need then 'covered' else 'gap'

# action outcome: write the confirmed date and store the acknowledgement
dh_po_lines.update(number = app.po, line = app.line, confirmed_date = app.new_date)
→ BAPI_PO_CHANGE · 4500012345 · RETURN: S · logged 14:06:12
Terms on this page

Data handlerA named connection from an app to rows in a system: a federated table, a saved query, a pipeline result, an API or a file. Screens, expressions and decision points use it to read and write.
Read in placeThe rows stay in the system that owns them. The app sends the filter and the columns it needs, and the system returns the matching rows.
Write-backA change made by the app in the system of record through that system's released interface, with the acknowledgement stored beside the decision.
Federated tableA table of a source system registered in the catalog and readable in place through the Data Fabric.
PushdownSending each source only the filter and the columns a query needs, so the source does the work and returns matching rows.
Pipeline resultRows prepared by a Backbone or data-science pipeline, checked and refreshed on a schedule.
Questions

Is a handler a copy of the table?

A table handler is the federated table seen through a filter and a column list. The rows stay in the source and are read when a screen needs them. A pipeline handler is the exception by design: a prepared result the Backbone lands and refreshes.

How fast is a screen on a large table?

As fast as the source answers a filtered query on its indexed columns, usually within seconds for operational screens. The showcase estate answers a question across four systems in 2.1 seconds. History at scale goes through a pipeline handler on the lake.

Who defines handlers?

Business configurators define table and query handlers from the catalog by picking objects and columns. Pipeline, API and device handlers are usually set up once by an application team and then reused.

What about a system that lacks a connector?

Sixty connectors exist and new ones are built from the same operator library in weeks. An API or file handler covers many cases in the meantime.

See a SmartApp built on your data

Bring one process and the systems it touches. In the first session we register the data handlers, draw the levels or the lanes and run the first screen on your live rows.