Table test subtests with non-string names are not recognized correctly

[oleynikandrey]

When writing table-driven tests in Go, neotest-golang does not recognize subtests correctly if the name of a subtest is not a simple string literal, but an expression (for example, using fmt.Sprintf or concatenation). In these cases, the whole table test is shown as a single test, and individual subtests are not detected or displayed.

Steps to reproduce:

1. Write a table-driven test where the subtest name is generated by an expression:

   tests := []struct {
       name string
       input int
       want  int
   }{
       {name: "case1", input: 1, want: 2},
       {name: "case2", input: 2, want: 4},
   }
   for _, tt := range tests {
       t.Run(fmt.Sprintf("%s-%d", tt.name, tt.input), func(t *testing.T) {
           // test logic
       })
   }

2. Run neotest-golang.

Expected behavior:

• Each subtest should be recognized and displayed individually, showing their dynamic names.

Actual behavior:

• Only one test is detected, and subtests are not recognized.

I understand that these values are calculated at runtime, but I wonder if support for this could be added.

[fredrikaverpil]

Hey @oleynikandrey, I don't think it will be easy to support this.
Somewhat related: #281

I've investigated whether Go itself can be used during the test detection phase, in addition (or replacing) AST parsing in lua. But I haven't found any way for Go to return all test found during runtime without also executing them. If you find a way, I'm all ears.

With #373 I'm adding streaming support. This means I will be able to detect tests not in the Neotest tree. Theoretically it could (I'm not sure) be possible to inject new nodes into the tree, but they would disappear if you save the file as Neotest would re-run the AST parsing and re-draw the tree.