Namespace stripping for functions/actions.

[andrueastman]

Related to

• Incorrect namespace assigned to 'UrlTemplate' in the Security RunHuntingQuery builder microsoftgraph/msgraph-sdk-dotnet#1691
  RunningHuntingQuery is an action and only works with the full namespace or the entire namespace stripped. not security.RunningHuntingQuery
• [Client bug]: Microsoft.Graph 5.4.0 generates invalid request Uris for GetDirectRoutingCallsWithFromDateTimeWithToDateTime  microsoftgraph/msgraph-sdk-dotnet#1787
  getDirectRoutingCalls is an action and only works with the full namespace or the entire namespace stripped. not callRecords.getDirectRoutingCalls

Stripping a specific prefix from namespace from Functions and Action can lead to invalid paths.

Following the discussion in #348 (comment)

Does it make sense to only use aliasing for TypeSegments and strip the namespace for Functions/Actions? I'm not sure if there are cases where a function/action fails if the entire namespace is stripped off?

cc @baywet @irvinesunday

[baywet]

I provided more context here but I don't believe we should change any of the conversion logic at this point. OData semantics clearly indicate that the method should be able to be called with the fully qualified name, the aliased name and the short method name. AGS seems to be messing up the aliased scenario and we should open an ICM on them for that reason.

[baywet]

After further discussion here is an updated reference

flowchart TD
     A[generate path segment] ---> B{EnableUnqualifiedCall?};
     B -- True --> C[use type name as segment];
     B -- False --> E{Is namespace aliased?};
     E -- True --> F[project alias to segment];
     E -- False --> G[project namespace to segment];
     F ----> H[Add type name to segment];
     G ----> H[Add type name to segment];
     H ----> I{Is namespacePrefixToStripForMethodPaths empty?};
     I --  True --> D[Path segment generated];
     I -- False --> J{Does the segment contain more than one dot?};
     J --  True --> D[Path segment generated];
     J -- False --> K[Strip prefix from generated segment];
     K ----> D[Path segment generated];
     C ----> D[Path segment generated];

Loading

[baywet]

See the additional "Does the segment contain more than one dot?" question, which could be roughly translated to "is the method under a sub namespace?"

[baywet]

For anyone else reading this thread, I retract my comment about AGS and ICMs. Since those namespaces are not being aliased, the functions can only be used with their short name or their fully qualified name, not the aliased name. This effectively means there's a shortcoming in the metadata conversion process (this lib) which the SDKs rely on for their generation. We'll address the conversion fix and things should roll out to the service libs shortly after.