[formrecognizer] Fix dictionary methods on DocumentField model (#23673)

* add fix for dict and list methods

* improve testing

* add invoice tests

* update changelog

* remove extra invoice test

* update tests to use AzureJSONEncoder

* move currency value in if

* pylint fix

Author: catalinaperalta

sdk/formrecognizer/azure-ai-formrecognizer/CHANGELOG.md

@@ -10,6 +10,7 @@
 - Renamed `model_count` and `model_limit` on `AccountInfo` to `document_model_count` and `document_model_limit`.
 
 ### Bugs Fixed
+- Fixed `to_dict()` and `from_dict()` methods on `DocumentField` to support converting lists, dictionaries, and CurrenyValue field types to and from a dictionary.
 
 ### Other Changes
 - Renamed `sample_copy_model.py` and `sample_copy_model_async.py` to `sample_copy_model_to.py` and `sample_copy_model_to_async.py` under the `3.2-beta` samples folder. Updated the samples to use renamed copy model method.

sdk/formrecognizer/azure-ai-formrecognizer/azure/ai/formrecognizer/_models.py

@@ -2598,9 +2598,18 @@ def to_dict(self):
         :return: dict
         :rtype: dict
         """
+        value = self.value
+        # CurrencyValue objects are interpreted as dict, therefore need to be processed first
+        # to call the proper to_dict() method.
+        if self.value_type == "currency":
+            value = self.value.to_dict()
+        elif isinstance(self.value, dict):
+            value = {k: v.to_dict() for k, v in self.value.items()}
+        elif isinstance(self.value, list):
+            value = [v.to_dict() for v in self.value]
         return {
             "value_type": self.value_type,
-            "value": self.value,
+            "value": value,
             "content": self.content,
             "bounding_regions": [f.to_dict() for f in self.bounding_regions]
             if self.bounding_regions
@@ -2620,9 +2629,20 @@ def from_dict(cls, data):
         :return: DocumentField
         :rtype: DocumentField
         """
+
+        value = data.get("value", None)
+        # CurrencyValue objects are interpreted as dict, therefore need to be processed first
+        # to call the proper from_dict() method.
+        if data.get("value_type", None) == "currency":
+            value = CurrencyValue.from_dict(data.get("value"))  #type: ignore
+        elif isinstance(data.get("value"), dict):
+            value = {k: DocumentField.from_dict(v) for k, v in data.get("value").items()}  # type: ignore
+        elif isinstance(data.get("value"), list):
+            value = [DocumentField.from_dict(v) for v in data.get("value")]  # type: ignore
+
         return cls(
             value_type=data.get("value_type", None),
-            value=data.get("value", None),
+            value=value,
             content=data.get("content", None),
             bounding_regions=[BoundingRegion.from_dict(v) for v in data.get("bounding_regions")]  # type: ignore
             if len(data.get("bounding_regions", [])) > 0
@@ -2880,7 +2900,7 @@ class DocumentPage(object):
     :vartype width: float
     :ivar height: The height of the image/PDF in pixels/inches, respectively.
     :vartype height: float
-    :ivar unit: The unit used by the width, height, and boundingBox properties. For
+    :ivar unit: The unit used by the width, height, and bounding box properties. For
      images, the unit is "pixel". For PDF, the unit is "inch". Possible values include: "pixel",
      "inch".
     :vartype unit: str

sdk/formrecognizer/azure-ai-formrecognizer/tests/test_dac_analyze_prebuilts.py

@@ -5,12 +5,14 @@
 # ------------------------------------
 
 import pytest
+import json
 import functools
 from datetime import date, time
 from devtools_testutils import recorded_by_proxy
 from io import BytesIO
 from azure.core.exceptions import ClientAuthenticationError, ServiceRequestError, HttpResponseError
 from azure.core.credentials import AzureKeyCredential
+from azure.core.serialization import AzureJSONEncoder
 from azure.ai.formrecognizer._generated.v2022_01_30_preview.models import AnalyzeResultOperation
 from azure.ai.formrecognizer import DocumentAnalysisClient, AnalyzeResult, FormContentType
 from testcase import FormRecognizerTest
@@ -182,6 +184,9 @@ def test_invoice_jpg(self, client, **kwargs):
         poller = client.begin_analyze_document("prebuilt-invoice", invoice)
 
         result = poller.result()
+        d = result.to_dict()
+        json.dumps(d, cls=AzureJSONEncoder)
+        result = AnalyzeResult.from_dict(d)
         assert len(result.documents) == 1
         invoice = result.documents[0]
 
@@ -419,6 +424,8 @@ def test_receipt_multipage(self, client):
         result = poller.result()
 
         d = result.to_dict()
+        # this is simply checking that the dict is JSON serializable
+        json.dumps(d, cls=AzureJSONEncoder)
         result = AnalyzeResult.from_dict(d)
 
         assert len(result.documents) == 2

sdk/formrecognizer/azure-ai-formrecognizer/tests/test_dac_analyze_prebuilts_async.py

@@ -5,11 +5,13 @@
 # ------------------------------------
 
 import pytest
+import json
 import functools
 from io import BytesIO
 from devtools_testutils.aio import recorded_by_proxy_async
 from datetime import date, time
 from azure.core.exceptions import HttpResponseError
+from azure.core.serialization import AzureJSONEncoder
 from azure.ai.formrecognizer.aio import DocumentAnalysisClient
 from azure.ai.formrecognizer import AnalyzeResult
 from azure.ai.formrecognizer._generated.v2022_01_30_preview.models import AnalyzeResultOperation
@@ -210,6 +212,8 @@ async def test_receipt_multipage(self, client):
             result = await poller.result()
 
         d = result.to_dict()
+        # this is simply checking that the dict is JSON serializable
+        json.dumps(d, cls=AzureJSONEncoder)
         result = AnalyzeResult.from_dict(d)
 
         assert len(result.documents) == 2
@@ -491,6 +495,9 @@ async def test_invoice_jpg(self, client, **kwargs):
             poller = await client.begin_analyze_document("prebuilt-invoice", invoice)
 
             result = await poller.result()
+        d = result.to_dict()
+        json.dumps(d, cls=AzureJSONEncoder)
+        result = AnalyzeResult.from_dict(d)
         assert len(result.documents) == 1
         invoice = result.documents[0]