Keep the entries of a map keyed by ints when it is read as an object

[danielcweber]

A g:Map whose keys are not strings, such as ints, now keeps its entries when it is requested as an object. It used to come back as an empty dynamic object with every entry gone. It now comes back as a Dictionary<object, object> whose keys are still the ints they were sent as, the same entries a request for Dictionary<int, string> already got. Maps keyed by strings are unchanged and still become dynamic objects, and so are element maps whose id and label arrive under g:T keys.

The route to object copied the map into a JObject, and a JObject can only hold keys that are names, so any other key was dropped on the way.

Changes

• Tests.Infrastructure: Map_of_typed_int_keys_as_object reads Map_of_Typed_Int_Keys_Typed_String_Values as an object. It records the result's type and its keys' types next to the value, because a snapshot writes the int key 1 and the string key "1" the same way, and a map that kept its entries by turning one into the other would not have kept them.
• Tests.Infrastructure: Constructor_arguments_from_map_with_int_key pins the other half. A map with an int-keyed entry among its constructor arguments, read into a ClassWithFieldsAndConstructor, gives { StringArg: stringValue, IntArg: 42 }: the entry no member can be named after is left out, and what follows it is still read. Both implementations already answered this.
• Support.NewtonsoftJson: when the target is object and some key is neither a string nor a g:T, MapDeferralConverterFactory asks for a Dictionary<object, object> instead of building a JObject. MapToDictionaryConverter already builds that type from a g:Map, converting each key like any other token. Its answer is taken as it is: it checks for the same g:Map this converter has just found, so there is no fallback to the JObject road.
• Providers.GremlinServer.Tests: Group and Group_with_key_identity in ObjectQueryIntegrationTests update their snapshots. group() keys its map by the element itself, so these results have non-string keys too. Both had pinned [ {} ], with every group dropped. They now record the groups: each vertex keys the list of vertices it grouped.

Notes

• Contains Say that an element map with an unknown label is still an element #2458, which it is branched from. Merge that first. Both are fast-forwards of 14.x.
• Behaviour change. A caller that requested object for such a map used to get an empty object and now gets a dictionary with the entries. This includes real server output: a group() requested as object now comes back with its groups. Targets other than object are unaffected, since a POCO has no member a key of 1 could fill. No public API change, so version.json is left alone.
• Measured before the fix: the test failed here with an empty DynamicDictionary and no keys at all. The System.Text.Json implementation in Gremlinq.Extensions gave the same answer. Its fix is Gremlinq.Extensions#207.
• The contract test and the fix are separate commits, test first.
• Support.NewtonsoftJson.Tests on net10.0: 273, up from 271, all green. Coverlet shows every changed line fully covered, branches included. dotnet build ExRam.Gremlinq.slnx succeeds with 0 warnings.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Syxmt77AL1HAfGM6p3LseP

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.20%. Comparing base (531e3cc) to head (66ad4da).

Additional details and impacted files

@@           Coverage Diff           @@
##             14.x    #2459   +/-   ##
=======================================
  Coverage   93.19%   93.20%           
=======================================
  Files         279      279           
  Lines        7803     7809    +6     
  Branches      874      877    +3     
=======================================
+ Hits         7272     7278    +6     
  Misses        334      334           
  Partials      197      197           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.