[TA] Changes based on V3.0 swagger, etc

[mssfang]

• Use V3.0 swagger code base, fixes: [TA] Regenerate code base by using V3.0 swagger #10585, fixes: [TA] update TextAnalyticsServiceVersion when service GA. #8480
• Hide public constructor of analyzed result, which is no longer an issue. We will keep constructor design, fixes: [TA] Remove output type public constructors #10346
• Renaming:
  • DetectedLanguage: getScore() to getConfidenceScore()
  • CategorizedEntity: getSubCategory() to getSubcategory()
  • LinkedEntity: getLinkedEntityMatches() to getMatches()
  • DocumentResult to TextAnalyticsResult
  • TextAnalyticsException and TextAnalyticsError: getCode() to getErrorCode()
  • TextAnalyticsWarning: getCode() to getWarningCode()
• Internal transfer TextAnalyticsErrorException to HttpResponseException and added tests
• Support not throws exception directly in asycn client but return it as monoError or fluxError. Added class DocumentInputAsyncTest to support it
• Update sync client to make sure the client throws exception.
• Added one new method getText() to SentenceSentiment
• Added a new class, TextAnalyticsWarning and one expandatable enum WarningCode
• Deprecated TextDocumentInput(String id, String text, String language) constructor, but added setLanguage() setter since language is optional.
• Remove graphme properties: removed getGraphemeLength() and getGraphemeOffset() from CategorizedEntity, SentenceSentiment, and LinkedEntityMatch. getGraphemeCount() in TextDocumentStatistics has been renamed to getCharacterCount(). fixes: [TA] Encoding and removing grapheme prefix #10587
• Warningsproperty added to each document-level response object returned from the endpoints. It is a list ofTextAnalyticsWarnings`. fixes: [TA] Warnings is enable in V3.0. Add it to single and batch APIs #10586
• Added CategorizedEntityCollection, KeyPhrasesCollection, LinkedEntityCollection for having getWarnings() to retrieve warnings.
• Added a new enum value ADDRESS to EntityCategory.
• Replaced all single input asynchronous APIs, e.x.,
  • TextAnalyticsPagedFlux<CategorizedEntity> recognizeEntities(String document) to Mono<CategorizedEntityCollection> recognizeEntities(String document).
  • TextAnalyticsPagedFlux<LinkedEntity> recognizeLinkedEntities(String document) to Mono<LinkedEntityCollection> recognizeLinkedEntities(String document).
  • TextAnalyticsPagedFlux<String> extractKeyPhrases(String document) to Mono<KeyPhrasesCollection> extractKeyPhrases(String document).
• Replaced all single input synchronous APIs, e.x.,
  • TextAnalyticsPagedIterable<CategorizedEntity> recognizeEntities(String document) to CategorizedEntityCollection recognizeEntities(String document).
  • TextAnalyticsPagedIterable<LinkedEntity> recognizeLinkedEntities(String document) to LinkedEntityCollection recognizeLinkedEntities(String document).
  • TextAnalyticsPagedIterable<String> extractKeyPhrases(String document) to KeyPhrasesCollection extractKeyPhrases(String document).

[samvaity]

Left a few comments, but rest LGTM.

[samvaity]

Is it expected to only return the status code in the response?

[maririos]

The exception the user will get should be the same as if the service returns a status = 400 and all the information with it.

For .net, for example, we have status, error code, and error message and we throw a RequestFailedException

[mssfang]

This is HttpResponse implementation class. Unless we know what is the expected header values and etc. We can make it up. But the for error message, it exist in the HttpResponseException level.

[samvaity]

This is still not updated.

[samvaity]

Any particular reason we changed this from AAD to AzureKeyCredential?

[mssfang]

the new endpoint, is only support Key Credential, it is owned by the service team. We are just using it temporately.

[samvaity]

extra / .
Also, do we need to update all the recordings?

[mssfang]

extra / should be fine.

There is a lot of place involving renaming, such as, score to confidenceScore

[iscai-msft]

Sorry I was wrong when I suggested this, I didn't realize that this TextDocumentInput constructor didn't take language. I'm confused though: it looks like in master you had two constructors: one that took language, one that did not.

I think you should still have two constructors here like in master. For the one that doesn't take language as a parameter, it should set this.language = "en", since that is the default. Otherwise, this.language = language

[mssfang]

if in the future, the default lanauge is change to 'es. we need to update it to es`. Since service team take default value in service side, we should not add what they are maintaining here. So keep no-assigned in SDK should be an better idea.

[iscai-msft]

i'm not sure i understand the context on just making setLanguage a setter and not adding it to a separate constructor like you used to have it. Is there a reason why you can't add this constructor back like it is in master:

azure-sdk-for-java/sdk/textanalytics/azure-ai-textanalytics/src/main/java/com/azure/ai/textanalytics/models/TextDocumentInput.java

Line 50 in fcdaff1

public TextDocumentInput(String id, String text, String language) {

I really don't think it's a good idea to force users to call TextDocumentInput.setLanguage("en") for the most basic version of TextDocumentInput (you're doing this in your tests). What I think you should do is have two constructors exactly how it is in master. Is there a reason why you're changing this code?

[mssfang]

the only time constructor needed is when the parameter is required. But since language is optional. it should not go to constructor. Let's say maybe in the future, more parameter will be added to this class, we will not add another constructor if the value is not required.