Skip to content

Latest commit

 

History

History
258 lines (201 loc) · 7.7 KB

concepts.md

File metadata and controls

258 lines (201 loc) · 7.7 KB

Server side extensions

Qlik's Server Side Extension feature enables you to extend the engine with custom calculations.

There are two ways you can do this; by defining custom functions, and by enabling runtime script evaluation.

Functions

Intro

Functions are defined based on the input type you expect them to have and could be one of scalar, aggregation or tensor.

To explain them, we're going to assume we have the following data model:

Product group Product Sales
Snack Chips 19
Snack Popcorn 11
Snack Cookies 18
Dairy Cheese 12 
Prods:
LOAD * Inline [
Product group,Product,Sales
Snack, Chips, 19
Snack, Popcorn, 11
Snack, Cookies, 18
Dairy, Cheese, 12
];

Data in SSE is received in the shape of bundles:

const bundle = {
  rows: [{
    duals: [{ // columns
      numData, // numerical value of the dual
      strData, // string value of the dual
    }, /* ... */]
  }, /* ... */]
];

Tensor (and Scalar)

Tensor is used when there is a one-to-one relationship between the fields in the cube and the expressions you provide to the SSE function.

If we use Product group as dimension and pass in sum(Sales) as a parameter to the function, we would receive one bundle with one row per Product group:

  • [48, 12]

Example

The following example returns the value 'green' if the value of the expression sum(Sales) is > 20, otherwise it returns 'red'.

The expression:

sse.color(sum(Sales));

The SSE function:

function greenOrRed(request) {
  request.on('data', (bundle) => {
    const rows = [];
    bundle.rows.forEach((row) => {
      rows.push({ // for each row in the input bundle
        duals: [{ // we add one row to the output, containing only one dual
          // that dual must contain 'strData' because the return type of this function is STRING
          strData: row.duals[0].numData > 20 ? 'green' : 'red' // numData is expected to be valid since the dataType of the first param is NUMERIC
        ],
      });
    });
    request.write({ rows }); // send data back to Qlik engine
  }
}

// register the function
server.addFunction(greenOrRed, {
  functionType: q.sse.FunctionType.TENSOR,
  returnType: q.sse.DataType.STRING,
  name: 'color', // optional, if not specified the name will be the function itself ('greenOrRed')
  params: [{
    name: 'first',
    dataType: q.sse.DataType.NUMERIC
  }]
})

Aggregation

Aggregation is used when there is a one-to-many relationship between the fields in the cube and the expressions you provide to the SSE function. Common use cases are calculating sum, avg, concatenation etc.

If we use Product group as dimension and pass in Sales as a parameter to the function, we would receive three bundles:

  • For Product group 'Snacks': [19, 11, 18]
  • For Product group 'Dairy': [12]
  • For total: [19, 11, 18, 12]

Example

If we want to calculate the sum of Sales for each Product group, we simply aggregate the rows in each bundle:

(bundle) => {
  let v = 0;
  bundle.rows.forEach((row) => {
    row.duals.forEach((dual) => {
      v += dual.numData;
    });
  });
  request.write({
    rows: [{ // return only one row, containing one dual
      duals: [{ numData: v }],
    }],
  });
}

Script evaluation

Note! Enabling script evaluation is dangerous!

Intro

Script evaluation makes it easy to run some basic arithmetic and string manipulation.

To simplify the access to values inside scripts:

  • bundles are first remapped into an args parameter. The values inside args are:
    • strings when parameter type is STRING
    • numbers when parameter type is NUMERIC
    • dual when parameter type is DUAL
  • the script is evaluated by wrapping it inside a function: 
    function(args, params, common) {
  // script evaluated here
}
    • args may be one of
      • (string | number | dual)[][] - when any of the *Aggr functions are called
      • (string | number | dual)[]
    • params <Parameter[]>
    • common <CommonRequestHeader>
  • values are then remapped back to appropriate bundles

Example

Calculate the sum of Sales per Product Group:

HelloSSE.ScriptAggr('return args.reduce((prev, curr) => (prev + curr[0]), 0)', 'Sales')

In this example we're using one of the *Aggr functions, the value of args will then be:

  • For Product group 'Snacks': [[19], [11], [18]]
  • For Product group 'Dairy': [[12]]
  • For total: [[19], [11], [18], [12]

Example

Return 'green' if the value of the expression sum(Sales) is > 20, otherwise return 'red'.

sse.ScriptEvalExStr('N', 'return args[0] > 20 ? "green" : "red"', sum(Sales));

args will in this case be:

  • For Product group 'Snacks': [48]
  • For Product group 'Dairy': [12]

In this example, we use *Ex because we want to get the numeric value of sum(Sales) by specifying N as the first parameter. The values inside args will therefore be the numData part of the dual. We use *Str because we want a string as output.

Table load

Intro

It's possible to load entire tables and augment script data through a table load defined in the script.

Example

Let's first define a function that has a table description and returns additional columns of data:

const NUTRITIONAL_FACTS = {
  // protein, fat, carbs (in grams)
  Cheese: [20, 25, 4],
  Popcorn: [12, 4, 78],
  Chips: [7, 35, 53],
  Cookies: [6, 24, 65],
};

function nutrition(request) {
  request.on('data', (bundle) => {
    const rows = [];
    bundle.rows.forEach((row) => {
      const c = NUTRITIONAL_FACTS[row.duals[0].strData];
      rows.push({
        duals: [
          { strData: row.duals[0].strData }, // first column - product
          { numData: c ? c[0] : '-' }, // second column - protein
          { numData: c ? c[1] : '-' }, // third column - fat
          { numData: c ? c[2] : '-' }, // fourth column - carbs
        ],
      });
    });
    request.write({ rows });
  });
}

s.addFunction(nutrition, {
  functionType: q.sse.FunctionType.TENSOR,
  // returnType: q.sse.DataType.NUMERIC, // returnType doesn't matter when called with the 'extension' clause
  params: [{
    name: 'does not matter',
    dataType: q.sse.DataType.STRING,
  }],
  tableDescription: { // describe the table this function is expected to return
    fields: [
      { dataType: q.sse.DataType.STRING, name: 'Product'},
      { dataType: q.sse.DataType.NUMERIC, name: 'protein'},
      { dataType: q.sse.DataType.NUMERIC, name: 'fat'},
      { dataType: q.sse.DataType.NUMERIC, name: 'carbs'}
    ],
  },
});

In the load script we can then load that additional data using the Load ... Extension syntax:

Prods:
LOAD * Inline [
Product group,Product,Sales
Snack, Chips, 19
Snack, Popcorn, 11
Snack, Cookies, 18
Dairy, Cheese, 12

Load * Extension sse.nutrition( Prods{Product} );
];

which will result in the following table being added to the data model:

Product fat protein carbs
Chips 20 25 4
Popcorn 12 4 78
Cookies 7 35 53
Cheese 6 24 65