Adds subproperty as new literal type

[isaiahvita]

To support subproperties in the aws config file, it was decided to extend the ini lexer to treat subproperties as literals. Then parse these literals in the config parser.

It was specifically avoided to add this support in the INI parser. This is because the current state of the INI parser:

• does not follow traditional language parser construction from a CFG (or BNF) into a parse table.
• parser logic from a parse table is obscure and is non-standard
• all institutional knowledge of this implementation is lost

[lucix-aws]

Is there going to be followup to bubble this further up to the actual parsed Sections? We have test data right now that still expects nested fields to be skipped:

[foo]                            
aws_access_key_id =              
    aws_secret_access_key = valid
aws_session_token = valid        
[bar]                            
aws_access_key_id = valid        
aws_secret_access_key = valid    
aws_session_token = valid        
[baz]                            
aws_access_key_id = valid        
aws_secret_access_key = valid    
aws_session_token = valid        

expects

{
    "foo": {
        "aws_session_token": "valid"
        // doesn't expect {"aws_access_key_id": { "aws_secret_access_key": "valid"} }
    },
    "bar": {
        "aws_access_key_id": "valid",
        "aws_secret_access_key": "valid",
        "aws_session_token": "valid"
    },
    "baz": {
        "aws_access_key_id": "valid",
        "aws_secret_access_key": "valid",
        "aws_session_token": "valid"
    }
}

I would think that test case should fail as written and be updated with this change.

• https://github.com/aws/aws-sdk-go-v2/blob/main/internal/ini/testdata/valid/nested_fields
• https://github.com/aws/aws-sdk-go-v2/blob/main/internal/ini/testdata/valid/nested_fields_expected

[isaiahvita]

@lucix-aws

Is there going to be followup to bubble this further up to the actual parsed Sections? We have test data right now that still expects nested fields to be skipped:

with the last commit, the data is now bubbled up in Sections.container["section_name"].values["item_name"].mp
additionally, the referenced test data has been amended

[lucix-aws]

So given that this new token is "greedy" (like, it's made up of smaller tokens that could be recognized on their own) we probably want to extend the test data (can do nested_fields) to cover a few more things

• multiple nested fields
• nested fields at the end of a profile where the next line is a [profile] declaration
• nested fields at the end of a file
• nested fields at the end of a file + newlines

LGTM otherwise, fixandship.

[lucix-aws]

Why no MapType? I don't fully remember what all uses this, though.

[lucix-aws]

Do you have gofmt running? It should clean up extra newlines like this.