[FormRecognizer] Parse operation IDs with regex

[kinelski]

Summary

The implementation of the Operation constructors uses some magic numbers when parsing the operation ID.

Operations to address

• CopyModelOperation
• RecognizeContentOperation
• RecognizeCustomFormsOperation
• RecognizeReceiptsOperation
• RecognizeIdDocumentsOperation
• TrainingOperation
• AnalyzeDocumentOperation
• ClassifyDocumentOperation

Internal constructor

azure-sdk-for-net/sdk/formrecognizer/Azure.AI.FormRecognizer/src/RecognizeCustomFormsOperation.cs

Lines 78 to 83 in e32b2e7

	string[] substrs = operationLocation.Split('/');
	_resultId = substrs[substrs.Length - 1];
	_modelId = substrs[substrs.Length - 3];
	Id = string.Join("/", substrs, substrs.Length - 3, 3);

The internal constructor takes a parameter called operationLocation, which is returned from the service and is expected to look like:

{endpoint}/formrecognizer/{version}/custom/models/{modelId}/analyzeresults/{resultId}

Public constructor

azure-sdk-for-net/sdk/formrecognizer/Azure.AI.FormRecognizer/src/RecognizeCustomFormsOperation.cs

Lines 102 to 112 in e32b2e7

	string[] substrs = operationId.Split('/');
	if (substrs.Length < 3)
	{
	throw new ArgumentException($"Invalid {operationId}. It should be formatted as: '{{modelId}}/analyzeresults/{{resultId}}'.", operationId);
	}
	_resultId = substrs.Last();
	_modelId = substrs.First();
	Id = operationId;

The public constructor, on the other hand, takes a parameter called operationId, which is given by the developer and is expected to be a substring of the aforementioned operationLocation:

{modelId}/analyzeresults/{resultId}

Goals

• Stop using magic numbers and correctly parse the arguments using regex, handling format errors.
• Make sure we support parsing model IDs with characters _, ~, and .. These characters are allowed when naming a custom model.
• Include new tests to verify proper behavior.
• Make sure existing tests still pass reliably.

The following file can be used as a reference:

azure-sdk-for-net/sdk/formrecognizer/Azure.AI.FormRecognizer/src/FormField.cs

Lines 90 to 91 in 80a6324

	private static Regex _wordRegex = new Regex(@"/readResults/(?<pageIndex>\d*)/lines/(?<lineIndex>\d*)/words/(?<wordIndex>\d*)$", RegexOptions.Compiled, TimeSpan.FromSeconds(2));
	private static Regex _lineRegex = new Regex(@"/readResults/(?<pageIndex>\d*)/lines/(?<lineIndex>\d*)$", RegexOptions.Compiled, TimeSpan.FromSeconds(2));

Related: #10385

[maririos]

@kinelski is this for P3? or should we put it in the backlog epic

[kinelski]

@maririos I wouldn't say it's a high-priority issue, since it just changes the behavior internally. @annelo-msft is keeping track of what issues we want closed for P3.

[maririos]

Thanks Caio. I put the issue in the backlog epic. In general, it is better if we don't have issues floating around without an epic.

[annelo-msft]

Closed #10385 and added full list of operations to this issue.