Skip to content

beda-software/fhir-py

Repository files navigation

build status codecov pypi Checked with mypy Supported Python version

fhir-py

async/sync FHIR client for python3. This package provides an API for CRUD operations over FHIR resources

pip install fhirpy

or to install the latest dev version:

pip install git+https://github.com/beda-software/fhir-py.git

You can test this library by interactive FHIR course in the repository Aidbox/jupyter-course.

Getting started

Async example

import asyncio
from fhirpy import AsyncFHIRClient


async def main():
    # Create an instance
    client = AsyncFHIRClient(
        'http://fhir-server/',
        authorization='Bearer TOKEN',
    )

    # Search for patients
    resources = client.resources('Patient')  # Return lazy search set
    resources = resources.search(name='John').limit(10).sort('name')
    patients = await resources.fetch()  # Returns list of AsyncFHIRResource

    # Create Organization resource
    organization = client.resource(
        'Organization',
        name='beda.software',
        active=False
    )
    await organization.save()

    # Update (PATCH) organization. Resource support accessing its elements
    # both as attribute and as a dictionary keys
    if organization['active'] is False:
        organization.active = True
    await organization.save(fields=['active'])
    # `await organization.patch(active=True)` would do the same PATCH operation

    # Get patient resource by reference and delete
    patient_ref = client.reference('Patient', 'new_patient')
    # Get resource from this reference
    # (throw ResourceNotFound if no resource was found)
    patient_res = await patient_ref.to_resource()
    await patient_res.delete()

    # Iterate over search set
    org_resources = client.resources('Organization')
    # Lazy loading resources page by page with page count = 100
    async for org_resource in org_resources.limit(100):
        print(org_resource.serialize())


if __name__ == '__main__':
    loop = asyncio.get_event_loop()
    loop.run_until_complete(main())

Searchset examples

patients = client.resources('Patient')

patients.search(birthdate__gt='1944', birthdate__lt='1964')
# /Patient?birthdate=gt1944&birthdate=lt1964

patients.search(name__contains='John')
# /Patient?name:contains=John

patients.search(name=['John', 'Rivera'])
# /Patient?name=John&name=Rivera

patients.search(name='John,Eva')
# /Patient?name=John,Eva

patients.search(family__exact='Moore')
# /Patient?family:exact=Moore

patients.search(address_state='TX')
# /Patient?address-state=TX

patients.search(active=True, _id='id')
# /Patient?active=true&_id=id

patients.search(gender__not=['male', 'female'])
# /Patient?gender:not=male&gender:not=female

Chained parameters

patients.search(general_practitioner__Organization__name='Hospital')
# /Patient?general-practitioner:Organization.name=Hospital
patients.search(general_practitioner__name='Hospital')
# /Patient?general-practitioner.name=Hospital

Reference

practitioner = client.resources('Practitioner').search(_id='john-smith').first()
patients.search(general_practitioner=practitioner)
# /Patient?general-practitioner=Practitioner/john-smith

Date

import pytz
import datetime


patients.search(birthdate__lt=datetime.datetime.now(pytz.utc))
# /Patient?birthdate=lt2019-11-19T20:16:08Z

patients.search(birthdate__gt=datetime.datetime(2013, 10, 27, tzinfo=pytz.utc))
# /Patient?birthdate=gt2013-10-27T00:00:00Z

Modifiers

conditions = client.resources('Condition')

conditions.search(code__text='headache')
# /Condition?code:text=headache

conditions.search(code__in='http://acme.org/fhir/ValueSet/cardiac-conditions')
# /Condition?code:in=http://acme.org/fhir/ValueSet/cardiac-conditions

conditions.search(code__not_in='http://acme.org/fhir/ValueSet/cardiac-conditions')
# /Condition?code:not-in=http://acme.org/fhir/ValueSet/cardiac-conditions

conditions.search(code__below='126851005')
# /Condition?code:below=126851005

conditions.search(code__above='126851005')
# /Condition?code:above=126851005

Raw parameters

Sometimes you can find that fhir-py does not implement some search parameters from the FHIR specification. In this case, you can use Raw() wrapper without any transformations

from fhirpy.base.searchset import Raw

patients = client.resources('Patient')
patients.search(Raw(**{'general-practitioner.name': 'Hospital'}))
# /Patient?general-practitioner.name=Hospital

Get resource by id

Use reference to get resource by id

patient = await client.reference('Patient', '1').to_resource()
# /Patient/1

Or use FHIR search API with .first() or .get() as described below.

Get exactly one resource

practitioners = client.resources('Practitioner')

try:
    await practitioners.search(active=True, _id='id').get()
    # /Practitioner?active=true&_id=id
except ResourceNotFound:
    pass
except MultipleResourcesFound:
    pass

Get first result

await practitioners.search(name='Jack').first()
# /Practitioner?name=Jack&_count=1

await patients.sort('active', '-birthdate').first()
# /Patient?_sort=active,-birthdate&_count=1

Get total count

await practitioners.search(active=True).count()

await patients.count()

Fetch one page

await practitioners.fetch()
# /Practitioner

await patients.elements('name', 'telecom').fetch()
# /Patient?_elements=resourceType,name,id,telecom

Fetch all resources on all pages

Keep in mind that this method as well as .fetch() doesn't return any included resources. Use fetch_raw() if you want to get all included resources.

# Returns a list of `Practitioner` resources
await practitioners.search(address_city='Krasnoyarsk').fetch_all()

await patients.fetch_all()

Page count (_count)

# Get 100 resources
await practitioners.limit(100).fetch()

Sort (_sort)

observations = client.resources('Observation')

observations.sort('status', '-date', 'category')
# /Observation?_sort=status,-date,category

Elements (_elements)

# Get only specified set of elements for each resource
patients.elements('identifier', 'active', 'link')
# /Patient?_elements=identifier,active,link

# Get all elements except specified set
practitioners.elements('address', 'telecom', exclude=True)

Include

result = await client.resources('EpisodeOfCare') \
    .include('EpisodeOfCare', 'patient').fetch_raw()
# /EpisodeOfCare?_include=EpisodeOfCare:patient
for entry in result.entry:
    print(entry.resource)

await client.resources('MedicationRequest') \
    .include('MedicationRequest', 'patient', target_resource_type='Patient') \
    .fetch_raw()
# /MedicationRequest?_include=MedicationRequest:patient:Patient

Modifier :iterate (or :recurse in some previous versions of FHIR)

# For FHIR version >= 3.5 we can also use modifier :iterate
await client.resources('MedicationRequest') \
    .include('MedicationDispense', 'prescription') \
    .include('MedicationRequest', 'performer', iterate=True) \
    .fetch_raw()
# /MedicationRequest?_include=MedicationDispense:prescription
#    &_include:iterate=MedicationRequest:performer

# For FHIR version 3.0-3.3 use modifier :recurse
await client.resources('MedicationDispense') \
    .include('MedicationRequest', 'prescriber', recursive=True) \
    .fetch_raw()
# /MedicationDispense?_include:recurse=MedicationRequest:prescriber

Wild card (any search parameter of type=reference be included)

await client.resources('Encounter').include('*') \
    .fetch_raw()
# /Encounter?_include=*

Revinclude

await practitioners.revinclude('Group', 'member').fetch_raw()
# /Practitioner?_revinclude=Group:member

or

await practitioners.include('Group', 'member', reverse=True).fetch_raw()
# /Practitioner?_revinclude=Group:member

Wild card (any search parameter of type=reference be included)

await client.resources('EpisodeOfCare').revinclude('*') \
    .fetch_raw()
# /EpisodeOfCare?_revinclude=*

Conditional operations

Conditional create

FHIR spec: Conditional create

For resource

# resource.create(search_params)
# executes POST /Patient?identifier=fhirpy

patient = client.resource("Patient",
    identifier=[{"system": "http://example.com/env", "value": "fhirpy"}],
    name=[{"text": "Mr. Smith"}],
)
await patient.create(identifier="other")

For SearchSet

# searchset.get_or_create(resource)
# executes POST /Patient?identifier=fhirpy

patient, created = await client.resources("Patient").search(identifier="fhirpy").get_or_create(patient_to_save)

# no match -> created is True
# one match -> created is False, return existing resource
# multiple matches -> 412 'MultipleResourcesFound'

Conditional update

FHIR spec: Conditional update

# resource, created: bool = searchset.patch(resource)
# executes PUT /Patient?identifier=fhirpy

patient_to_update = client.resource("Patient", 
                                    identifier=[{"system": "http://example.com/env", "value": "fhirpy"}],
                                    active=False)
new_patient, created = await client.resources("Patient").search(identifier="fhirpy").update(patient_to_update)

# no match -> created is True
# one match -> created is False, the matched resource is overwritten
# multiple matches -> 412 'MultipleResourcesFound'

Conditional patch

FHIR spec: Conditional patch

# patched_resource = searchset.patch(resource)
# executes PATCH /Patient?identifier=fhirpy

patient_to_patch = client.resource("Patient", 
                                    identifier=[{"system": "http://example.com/env", "value": "fhirpy"}], 
                                    name=[{"text": "Mr. Smith"}])
patched_patient = await client.resources("Patient").search(identifier="fhirpy").patch(patient_to_patch)

# no match -> 404 'ResourceNotFound'
# multiple matches -> 412 'MultipleResourcesFound'

Conditional delete

FHIR spec: Conditional delete

response_data, status_code = await self.client.resources("Patient").search(identifier="abc").delete()

# no match -> status_code = 204 'No Content'
# one match -> status_code = 200 'OK'
# multiple matches -> status_code = 412 'MultipleResourcesFound' (implementation specific)

Data models

Third party typing data models might be used along with fhir-py. The typing data models should match ResourceProtocol, e.g. have resourceType attribute, optional id and be iterable for serialization. There's a third party repository fhir-py-types that is written on top of pydantic models is fully compatible with fhir-py.

Static typechecking

fhir-py uses typehints in the codebase and it statically checked by mypy. Some interfaces that are described below designed in the way to properly infer the return type based on the model class or instance.

Resource instantiation

To instantiate a resource, simply use type model constructor, e.g.

patient = Patient(name=[HumanName(text='Patient')])

CRUD client methods

Client class provides CRUD methods that designed to work with typed models.

Create

await client.create(patient) # returns Patient

Update

await client.create(patient) # returns Patient

Save

Smart helper that creates or updated the resource based on having id

await client.save(patient) # returns Patient

Also it supports overriding specific fields using patch:

await client.save(patient, fields=['identifier']) # returns Patient

Patch

Patch accepts different syntaxes for patching, there're two syntaxes for general usage, without inferred types:

  • Patch using reference defined by separate resource type and id:
await client.patch('Patient', 'id', name=[HumanName(text='Patient')]) # returns Any
  • Patch using reference string:
await client.patch('Patient/id', name=[HumanName(text='Patient')]) # returns Any

And two types that infers type:

  • Patch using model class and id
await client.patch(Patient, 'id', name=[HumanName(text='Patient')]) # returns Patient
  • Patch using model instance
await client.patch(patient, name=[HumanName(text='Patient')]) # returns Patient

Delete

Delete accepts different syntaxes for resource deletion, there're also syntaxes similar to patch, but without output type because delete usually returns nothing.

  • Delete using reference defined by separate resource type and id:
await client.delete('Patient', 'id') 
  • Delete using reference string:
await client.delete('Patient/id')
  • Delete using model class and id
await client.delete(Patient, 'id')
  • Delete using model instance
await client.delete(patient)

Read

For fetching single resource by resourceType and id:

ss = await client.get(Patient, 'id') # returns Patient

For fetching multiple resources, SearchSet needs to be instantiated using the model class as the first argument

ss = client.resources(Patient) # returns AsyncFHIRSearchSet[Patient]
await ss.fetch_all() # returns list[Patient]

In that case search set infers model type for all methods that described above in the sections about search sets, including data fetching and conditional CRUD.

Resource and helper methods

Validate resource using operation $validate

try:
    await client.resource('Patient', birthDate='date', custom_prop='123', telecom=True) \
        .is_valid(raise_exception=True)
except OperationOutcome as e:
    print('Error: {}'.format(e))

patient = client.resource('Patient', birthDate='1998-01-01')
if (await patient.is_valid()):
    pass

Accessing resource attributes

patient = await client.resources('Patient').first()

# Work with the resource as a dictionary
patient_family = patient['name'][0]['family']

# Or access value by an attribute
patient_given_name = patient.name[0].given[0]

get_by_path(path, default=None)

patient_postal = patient.get_by_path(['resource', 'address', 0, 'postalCode'])

# get_by_path can be also used on any nested attribute
patient_name = patient.name[0]
patient_fullname = '{} {}'.format(
    patient_name.get_by_path(['given', 0]),
    patient_name.get_by_path(['family'])
)

# Get identifier value by specified system or empty string
uid = patient.get_by_path([
        'resource', 'identifier',
        {'system':'http://example.com/identifier/uid'},
        'value'
    ], '')

# Get base value amount or 0
invoice = await client.resources('Invoice').first()
base_value = invoice.get_by_path([
    'totalPriceComponent',
    {'type': 'base'},
    'amount', 'value'], 0)

set_by_path(obj, path, value)

resource = {
    "name": [{"given": ["Firstname"], "family": "Lastname"}],
}

set_by_path(resource, ["name", 0, "given", 0], "FirstnameUpdated")

# resource
# {"name": [{"given": ["FirstnameUpdated"], "family": "Lastname"}]}

serialize()

# Returns resources as dict
patient = await client.reference('Patient', '1').to_resource()
patient.serialize()
# Or 
await client.reference('Patient', '1').to_resource().serialize()
# {'resourceType': 'Patient', 'id': '1', 'meta': {'versionId': '1', 'lastUpdated': '2021-11-13T11:50:24.685719Z'}, ...}

Reference

Main class structure

Both async and sync clients have identical sets of classes and methods.

Sync Async
Client SyncFHIRClient AsyncFHIRClient
SearchSet SyncFHIRSearchSet AsyncFHIRSearchSet
Resource SyncFHIRResource AsyncFHIRResource
Reference SyncFHIRReference AsyncFHIRReference

Async client (based on aiohttp) – AsyncFHIRClient

Import library:

from fhirpy import AsyncFHIRClient

To create AsyncFHIRClient instance use:

AsyncFHIRClient(url, authorization='', extra_headers={})

Returns an instance of the connection to the server which provides:

  • .reference(resource_type, id, reference, **kwargs) - returns AsyncFHIRReference to the resource
  • .resource(resource_type, **kwargs) - returns AsyncFHIRResource which described below
  • .resources(resource_type) - returns AsyncFHIRSearchSet
  • .resources(resource_class: T) - returns AsyncFHIRSearchSet[T]
  • async .execute(path, method='post', data=None, params=None) - returns a result of FHIR operation

data model methods:

  • async .get(resource_type: type[T], id) - returns T instance by resourceType/id
  • async .get(resource_type: type[T], reference) - returns T instance by reference
  • async .get(resource_type: str, id) - gets instance by resourceType/id
  • async .get(reference: str) - gets instance by reference
  • async .save(resource: T, fields=[]) - creates or updates or patches (with fields=[...]) T instance
  • async .create(resource: T) - creates T instance
  • async .update(resource: T) - updates T instance
  • async .patch(resource: T, **kwargs) - patches T instance
  • async .patch(resource_type: type[T], id, **kwargs) - patches instance by resourceType/id and returns T instance
  • async .patch(resource_type: type[T], reference, **kwargs) - patches instance by reference and returns T instance
  • async .patch(resource_type: str, id, **kwargs) - patches instance by resourceType/id
  • async .patch(reference: str, **kwargs) - patches instance by reference
  • async .delete(resource: T) - deletes T instance
  • async .delete(resource_type: type[T], id) - deletes resource by resourceType/id
  • async .delete(resource_type: type[T], reference) - deletes resource by reference
  • async .delete(resource_type: str, id) - deletes instance by resourceType/id
  • async .delete(reference: str) - deletes instance by reference

Aiohttp request parameters

Sometimes you need more control over the way http request is made and provide additional aiohttp session's request parameters like ssl, proxy, cookies, timeout etc. It's possible by providing aiohttp_config dict for AsyncFHIRClient:

client = AsyncFHIRClient(
    FHIR_SERVER_URL,
    aiohttp_config={
        "ssl": ssl.create_default_context(),
        "timeout": aiohttp.ClientTimeout(total=100),
    }
)

Be careful and don't override other request values like params, json, data, auth, because it'll interfere with the way fhir-py works and lead to an incorrect behavior.

AsyncFHIRResource

provides:

  • .serialize() - serializes resource
  • .get_by_path(path, default=None) – gets the value at path of resource
  • async .save(fields=[]) - creates or updates or patches (with fields=[...]) resource instance
  • async .create() - creates resource instance
  • async .update() - updates resource instance
  • async .patch(**kwargs) - patches resource instance
  • async .delete() - deletes resource instance
  • async .refresh() - reloads resource from a server
  • async .to_reference(**kwargs) - returns AsyncFHIRReference for this resource
  • async .execute(operation, method='post', data=None, params=None) - returns a result of FHIR operation on the resource

AsyncFHIRReference

provides:

  • async .to_resource() - returns AsyncFHIRResource for this reference
  • async .execute(operation, method='post', data=None, params=None) - returns a result of FHIR operation on the resource
  • async .patch(**kwargs) - patches resource instance
  • async .delete() - deletes resource instance

AsyncFHIRSearchSet

provides:

  • .search(param=value)
  • .limit(count)
  • .sort(*args)
  • .elements(*args, exclude=False)
  • .include(resource_type, attr=None, recursive=False, iterate=False)
  • .revinclude(resource_type, attr=None, recursive=False, iterate=False)
  • .has(*args, **kwargs)
  • async .fetch() - makes query to the server and returns a list of Resource filtered by resource type
  • async .fetch_all() - makes query to the server and returns a full list of Resource filtered by resource type
  • async .fetch_raw() - makes query to the server and returns a raw Bundle Resource
  • async .first() - returns Resource or None
  • async .get() - returns Resource or raises ResourceNotFound when no resource found or MultipleResourcesFound when more than one resource found (parameter 'id' is deprecated)
  • async .count() - makes query to the server and returns the total number of resources that match the SearchSet
  • async .get_or_create(resource) - conditional create
  • async .update(resource) - conditional update
  • async .patch(**kwargs) - conditional patch

Sync client (based on requests) – SyncFHIRClient

Import library:

from fhirpy import SyncFHIRClient

To create SyncFHIRClient instance use:

SyncFHIRClient(url, authorization='', extra_headers={})

Returns an instance of the connection to the server which provides:

  • .reference(resource_type, id, reference, **kwargs) - returns SyncFHIRReference to the resource
  • .resource(resource_type, **kwargs) - returns SyncFHIRResource which described below
  • .resources(resource_type) - returns SyncFHIRSearchSet

Requests request parameters

Pass requests_config parameter to SyncFHIRClient if you want to provide additional parameters for a request like verify, cert, timeout etc.

client = SyncFHIRClient(
    FHIR_SERVER_URL,
    requests_config={
        "verify": False,
        "allow_redirects": True,
        "timeout": 60,
    }
)

Be careful and don't override other request values like params, json, data, headers, which may interfere with the way fhir-py works and lead to an incorrect behavior.

SyncFHIRResource

The same as AsyncFHIRResource but with sync methods

SyncFHIRReference

provides: The same as AsyncFHIRReference but with sync methods

SyncFHIRSearchSet

The same as AsyncFHIRSearchSet but with sync methods

Run integration tests

(need some test FHIR server to run with, e.g. https://docs.aidbox.app/installation/setup-aidbox.dev)

  1. Clone this repository: https://github.com/beda-software/fhir-py.git

  2. Go to fhir-py folder and install dev dependencies:

cd fhir-py
pip install -r requirements.txt

If you've already installed fhir-py library and want to test the last changes, reinstall it by running python setup.py install (or uninstall pip uninstall fhirpy)

  1. Provide ENV variables FHIR_SERVER_URL and FHIR_SERVER_AUTHORIZATION, or edit tests/config.py

  2. Run pytest

If you've found any bugs or think that some part of fhir-py is not compatible with FHIR spec, feel free to create an issue/pull request.