Skip to content

Latest commit

 

History

History
181 lines (118 loc) · 6.22 KB

advanced.rst

File metadata and controls

181 lines (118 loc) · 6.22 KB

Advanced usage

Batch requests

Example of batch request that contains 3 simple entity queries:

batch = northwind.create_batch()

batch.add_request(northwind.entity_sets.Employees.get_entity(108))
batch.add_request(northwind.entity_sets.Employees.get_entity(234))
batch.add_request(northwind.entity_sets.Employees.get_entity(23))

response = batch.execute()

print(response[0].EmployeeID, response[0].LastName)
print(response[1].EmployeeID, response[1].LastName)
print(response[2].EmployeeID, response[2].LastName)

Example of batch request that contains simple entity query as well as changest consisting of two requests for entity modification:

batch = northwind.create_batch()

batch.add_request(northwind.entity_sets.Employees.get_entity(108))

changeset = northwind.create_changeset()

changeset.add_request(northwind.entity_sets.Employees.update_entity(45).set(LastName='Douglas'))

batch.add_request(changeset)

response = batch.execute()

print(response[0].EmployeeID, response[0].LastName)

Error handling

PyOData returns HttpError when the response code does not match the expected code. Basically the exception is raised for all status codes >= 400 and its instances have the property response which holds return value of the library making HTTP requests.

For example, the following code show how to print out error details if Python Requests is used as the HTTP communication library.

try:
    new_data = create_request.execute()
except pyodata.exceptions.HttpError as ex:
    print(ex.response.text)

In the case you know the implementation of back-end part and you want to show accurate errors reported by your service, you can replace HttpError by your sub-class HttpError by assigning your type into the class member VendorType of the class HttpError.

from pyodata.exceptions import HttpError

class MyHttpError(HttpError):

    def __init__(self, message, response):
        super(MyHttpError, self).__init__('Better message', response)


HttpError.VendorType = MyHttpError

The class pyodata.vendor.SAP.BusinessGatewayError is an example of such an HTTP error handling.

Enable Logging

The library uses Python logging without own handler, so to enable logging it is enough to set log level to the desired value.

import logging

logging.basicConfig()
root_logger = logging.getLogger()
root_logger.setLevel(logging.DEBUG)

Dynamically referenced EntitySet

If you need to work with many Entity Sets the same way or if you just need to pick up the used Entity Set name in run-time, you may find out the ability to get an instance of Entity Set proxy dynamically. Here is an example of how you can print a count of all employees:

count = getattr(northwind.entity_sets, 'Employees').get_entities().count().execute()
print(count)

Inspecting responses with a hook

Some OData services communicate domain errors via HTTP response headers on otherwise-200 responses. Because pyodata discards headers before returning the domain result, callers cannot detect these errors through the normal return value.

Client (and Service directly) accept an optional response_hook parameter — a Callable[[response], None] that fires before the domain handler runs, for every request type (including async_execute()). The hook receives the raw response object. Raising an exception from the hook propagates to the caller and prevents the domain handler from running.

import pyodata
import requests

SERVICE_URL = 'https://odata.example.com/MyService.svc'

def my_hook(response):
    if response.headers.get('x-custom-error'):
        raise RuntimeError(f"Service signalled error: {response.headers['x-custom-error']}")

service = pyodata.Client(SERVICE_URL, requests.Session(), response_hook=my_hook)

The hook must be stateless to be safe under concurrent and async use. If you need to handle SAP-specific header errors, use the ready-made hook in pyodata.vendor.SAP — see :doc:`vendors`.

Out-of-range JSON dates

Some services return JSON /Date(...)/ values outside the range Python's datetime can represent. By default, PyOData raises pyodata.exceptions.PyODataModelError rather than silently replacing them. The supported millisecond range is -62135596800000 through 253402300799999, inclusive.

If your service deliberately uses such values as minimum/maximum date sentinels, you can enable either workaround independently before reading entity responses:

import pyodata.v2.model as model

model.FIX_SCREWED_UP_MINIMAL_DATETIME_VALUE = True
model.FIX_SCREWED_UP_MAXIMUM_DATETIME_VALUE = True

Both flags default to False. They apply to JSON decoding of Edm.DateTime and Edm.DateTimeOffset values, not to malformed date strings or date literals in URLs. Values already within the supported range are not changed.

  • FIX_SCREWED_UP_MINIMAL_DATETIME_VALUE replaces values below the minimum with January 1, year 1 at midnight.
  • FIX_SCREWED_UP_MAXIMUM_DATETIME_VALUE replaces values above the maximum with December 31, year 9999 at midnight. This is not datetime.max (the end of that day).

For example, with both flags enabled:

dates = model.Types.from_name('Edm.DateTime').traits
print(dates.from_json('/Date(-62135596800001)/'))
# 0001-01-01 00:00:00+00:00
print(dates.from_json('/Date(253402300800000)/'))
# 9999-12-31 00:00:00+00:00

The timezone supplied by the decoder is retained, including the offset for Edm.DateTimeOffset. These are module-level settings shared by all clients in the process, not per-client options. Configure only the flags needed by your service when your application starts; avoid toggling them around concurrent requests. Set a flag back to False to restore strict range checking.