Skip to content

Latest commit

 

History

History
375 lines (318 loc) · 10.6 KB

functional-model.md

File metadata and controls

375 lines (318 loc) · 10.6 KB
Raw

This document will help you understand how ViewDefinitions work "under the hood" using a functional paradigm.

The application of a ViewDefinition is an algorithm that describes the transformation of FHIR resources and is composed of combinations of a small set of functions:

  • column({name:column_name,path: fhirpath},...) - The main workhorse of transformation. This function will extract elements by FHIRPath expressions and put the result into columns
  • where(fhirpath) - This function filters resources by FHIRPath expression. For example, you may want to transform only specific profiles like blood pressure into a simple table
  • forEach(expr, transform) - This function unnests collection elements into separate rows
  • select(rows1, rows2) - This function cross-joins rows1 and rows2, and is mostly used to join results of forEach with top-level columns
  • union(rows, rows) - This function concatenates sets of rows. It's main use case is combining rows from different branches of a resource (for example, telecom and contact.telecom)

A ViewDefinition is represented as a FHIR logical model (in this case as a JSON document) where the keywords of the ViewDefinition correspond to the functions described above.

An Example ViewDefinition

{
    "resourceType": "ViewDefinition",
    "resource": "Patient",
    // Step 1
    "where": [{"path": "active = true"}],
    // Step 6
    "select": [
        {
          // Step 5
          "column": [
             {"path": "getResourceKey()", "name": "id"},
             {"path": "identifier.where(system='ssn')", "name": "ssn"},
          ]
        },
        {
          // Step 4
          "unionAll": [
            {
              // Step 2
              "forEach": "telecom.where(system='phone')",
              "column": [{"path": "value", "name": "phone"}]
            },
            {
              // Step 3
              "forEach": "contact.telecom.where(system='phone')",
              "column": [{"path": "value",  "name": "phone"}]
            }
       ]}
    ]
}

Applying the ViewDefinition

The application of this ViewDefinition produces a table of contacts with one row per telecom from two different locations in a FHIR Patient resource.

Algorithmically the application can be though of in the following steps:

  1. "where": filter data to only the active patients
  2. "forEach": unnest Patient.telecom and select "phone"
  3. "forEach": unnest Patient.contact.telecom and select "phone"
  4. "unionAll": concatenate the results of 2 and 3
  5. "column": extracts "id" and "ssn"
  6. "select": cross-join "id" and "ssn" with telecom phones from step 4

Here is example input and output for this ViewDefinition:

[
    {
        "resourceType": "Patient",
        "id": "pt1",
        "identifier": [{ "system": "ssn", "value": "s1" }],
        "telecom": [{ "system": "phone", "value": "tt1" }],
        "contact": [
            { "telecom": [{ "system": "phone", "value": "t12" }] },
            { "telecom": [{ "system": "phone", "value": "t13" }] }
        ]
    },
    {
        "resourceType": "Patient",
        "id": "pt2",
        "identifier": [{ "system": "ssn", "value": "s2" }],
        "telecom": [{ "system": "phone", "value": "t21" }],
        "contact": [
            { "telecom": [{ "system": "phone", "value": "t22" }] },
            { "telecom": [{ "system": "phone", "value": "t23" }] }
        ]
    }
]

The resulting output:

id ssn phone
pt1 s1 t11
pt1 s1 t12
pt1 s1 t13
pt2 s1 t21
pt2 s1 t22
pt2 s1 t23

The FHIRPath Subset

ViewDefinitions use minimal subset of FHIRPath to make implementation as simple as possible.

The specification also introduces two special functions:

  • getResourceKey - Indirectly get a resource's id. Since it my not be straightforward to access a resource's id, this layer of indirection is useful
  • getReferenceKey(resourceType) - A similar function to get an id from a reference

The Functions in Detail

Let’s walk through every function in detail with examples:

The column Function

column extracts elements into tabular columns using FHIRPath expressions. The algorithm for this function begins by receiving a list of {name, path} pairs. For each record in the given context, it evaluates the path expression to extract the desired elements. The resulting values are then added as columns to the output row.

{
    "column": [
        { "name": "id", "path": "getResourceKey()" },
        { "name": "bod", "path": "birthDate" },
        { "name": "first_name", "path": "name.first().given.join(' ')" },
        { "name": "last_name", "path": "name.first().family" },
        {
            "name": "ssn",
            "path": "identifier.where(system='ssn').value.first()"
        },
        {
            "name": "phone",
            "path": "telecom.where(system='phone').value.first()"
        }
    ]
}

Here is the naive JavaScript implementation:

function column(cols, rows) {
    return rows.map((row) => {
        return cols.reduce((res, col) => {
            res[col.name] = fhirpath(col.path, row);
            return res;
        }, {});
    });
}

The where Function

where retains only those records for which it FHIRPath expression returns true.

{
    "resourceType": "ViewDefinition",
    "resource": "Patient",
    "where": [
        { "path": "meta.profile.where($this = 'myprofile').exists()" },
        { "path": "active = 'true'" }
    ]
}

Basic JavaScript implementation:

function where(exprs, rows) {
    return rows.filter((row) => {
        return exprs.every((expr) => {
            return fhirpath(expr, row) == true;
        });
    });
}

The forEach & forEachOrNull Functions

forEach and forEachOrNull are intended for flattening nested collections by applying a transformation to each element. It consists of FHIRPath expression for collection to iterate and transformation to apply to each item. This function is akin to flatMap or mapcat in other programming languages.

{
    "resourceType": "ViewDefinition",
    "resource": "Patient",
    "select": [
        {
            "forEach": "name",
            "column": [
                { "path": "given.join(' ')", "name": "first_name" },
                { "path": "family", "name": "last_name" }
            ]
        }
    ]
}

There are two versions of this function: forEach and forEachOrNull. The primary difference is that forEach removes records where the FHIRPath expression returns no results, whereas forEachOrNull keeps an empty record in such cases.

Basic JavaScript implementation:

function forEach(path, expr, rows) {
    return rows.flatMap((row) => {
        return fhirpath(expr, row).map((item) => {
            // evalKeyword will call column, select or other functions
            return evalKeyword(expr, item);
        });
    });
}

The select Function

select is used in combination with forEach or forEachOrNull to cross-join parent elements with unnested collection elements. For example, Patient.id with an unnested collection such as Patient.name.

Put another way, this function merges columns from each row set, resulting in a comprehensive combination of data from the input collections.

{
    "resourceType": "ViewDefinition",
    "resource": "Patient",
    "select": [
        {
            "column": [{ "path": "getResourceKey()", "name": "id" }]
        },
        {
            "forEach": "name",
            "column": [
                { "path": "given.join(' ')", "name": "first_name" },
                { "path": "family", "name": "last_name" }
            ]
        }
    ]
}

The naive JavaScript implementation:

function select(rows1, rows2) {
    return rows1.flatMap((r1) => {
        return rows2.map((r2) => {
            // merge r1 and r2
            return { ...r1, ...r2 };
        });
    });
}

select([{ a: 1 }, { a: 2 }], [{ b: 1 }, { b: 2 }])[
    //=>
    ({ a: 1, b: 1 }, { a: 1, b: 2 }, { a: 2, b: 1 }, { a: 2, b: 2 })
];

The unionAll Function

unionAll combines rows from different branches of a resource tree by concatenating multiple record sets. Essentially a concatenation of several collections of records into a single, unified collection while preserving all rows from the input sets.

{
    "resourceType": "ViewDefinition",
    "resource": "Patient",
    "select": [
        {
            "column": [{ "path": "getResourceKey()", "name": "id" }]
        },
        {
            "unionAll": [
                {
                    "forEach": "telecom.where(system='phone')",
                    "column": [{ "path": "value", "name": "phone" }]
                },
                {
                    "forEach": "contact.telecom.where(system='phone')",
                    "column": [{ "path": "value", "name": "phone" }]
                }
            ]
        }
    ]
}

In JavaScript this is just a simple concatenation:

function unionAll(rowSets) {
    return rowSet.flatMap((rows) => {
        return rows;
    });
}

unionAll([1, 2, 3], [3, 4, 5])[
    //=>
    (1, 2, 3, 3, 4, 5)
];

Function Precedence

To interpret such nodes, you have to re-order keywords (functions) according to precedence (higher bubble up):

In a ViewDefinition, different keywords/functions can appear at the same level. To interpret such node in a ViewDefinition, an implementation must reorder the keywords/functions according to the following precedence rule.

Keyword/function reordering rule (highest to lowest precedence):

  • forEach or forEachOrNull
  • select
  • unionAll
  • column

For example given the following ViewDefinition snippet with all keywords/functions at the top level, the keywords/functions should be reordered as shown in the output.

{
    "forEach":   FOREACH,
    "column":    [COLUMNS], // reorder to be nested under select
    "unionALL":  [UNIONS],  // reorder to be nested under select
    "select":    [SELECTS]
}
//=>
{
    "forEach": FOREACH
    "select": [
       {"column":   [COLUMNS]},
       {"unionAll": [UNIONS]},
       SELECTS...
    ]
}

Reference Implementation

The JavaScript reference implementation implements the functional model described here. It is very concise at roughly 400 lines of code and reading it is a good way to understand the model in detail.