Currently, the sanitize of tck column name to be compliant with Avro & Studio do:

• If 1st chars is not avro compliant it is deleted : it raises issues !
• ASCII chars non-avro-compliant are transformed to '_'
• Non-ASCII chars, like jap ones, are transformed to '_'
• '+' and '/' from base64 are transformed to _'
  https://github.com/Talend/component-runtime/blob/a2d9069e2c5b820630e5d535d9b04b077e37fae1/component-api/src/main/java/org/talend/sdk/component/api/record/Schema.java#L298

In some case, it can generate column name collision.
For instance:

final Record record_c = this.factory.newRecordBuilder()
        .withString("1name_b", "a_value") 
        .withString("2name_b", "b_value")
        .build();

The first character is a number and so not Avro compatible. The columns name will be sanitized and it will generate a collision.

The proposal to fix that collision is to add a suffix "_${index}" in that way:

"shema":[
	{"name": "name_b", "rawname": "1name_b"},
	{"name": "name_b_1", "rawname": "2name_b"},
]

Their should be a fall back to rawname when we want to retrieve the value from a record:

record.getString("2name_b")

should return "b_value".

Scope of collision support
(a) Should we limite this fix only on collision generated after a sanitize ? (b) Or, we allow to add two column with same name ?
=> emmanuel_g / clesaec : if we allow (b) it may change the current behavior, so we limit to (a) : we only support collision after a sanitize.

Need to validate

• A given column has its name equals to the sanitized one of another column:

  final Record record_c = this.factory.newRecordBuilder()
          .withString("1name_b", "a_value") 
          .withString("2name_b", "b_value")
          .withString("name_b", "c_value")
          .build();
  

  Should generate:

  "shema":[
  	{"name": "name_b_1", "rawname": "1name_b"}, // a_value
  	{"name": "name_b_2", "rawname": "2name_b"}, // b_value
  	{"name": "name_b", "rawname": ""},          // c_value
  ]
  

• A given column has its name equals to the sanitized one of another column:

  final Record record_c = this.factory.newRecordBuilder()
          .withString("1name_b", "a_value") 
          .withString("2name_b", "b_value")
          .withString("name_b_1", "c_value")
          .build();
  

  May generate:

  "shema":[
  	{"name": "name_b", "rawname": "1name_b"}, // a_value
  	{"name": "name_b_2", "rawname": "2name_b"}, // b_value
  	{"name": "name_b_1", "rawname": ""},        // c_value
  ]
  

  If we generalize:

  final Record record_c = this.factory.newRecordBuilder()
          .withString("1name_b", "a_value")    // name_b
          .withString("2name_b", "b_value")    // name_b_1
          ...
          .withString("{n}name_b", "b_value")  // name_b_{n-1}
          .withString("name_b_x", "c_value")   // if 1 <= x < n
          .build();
  

  Should we support case ?