Auto detection of Data type

Elastic Stack Elasticsearch
1

This is the mapping initially which I created, considering that Roll number will be limit to only numbers. But eventually as they are growing in the roll numbers are Alphanumeric and the data type is changed.
PUT student_info
{
"mappings":
{
"properties":
{
"Roll_no":{"type":"integer"},
"Percentage": {"type": "float"},
"Rank": {"type": "integer"},
"Name":{"type":"text"}
}
}
}

This the 1st record I want to insert to the index student_info
This is getting indexed correctly
PUT student_info/_doc/1
{
"Roll_no":123,
"Percentage": 81.0,
"Rank":4,
"Name":"Sam"
}

This is giving error
PUT student_info/_doc/2
{
"Roll_no":"A123",
"Percentage": 83.0,
"Rank":9,
"Name":"David"
}

Error "reason": "failed to parse field [Roll_no] of type [integer] in document with id '2'. Preview of field's value: 'A123'"

2

What is the question?
If the mapping is number, other values will fail.
You need to change the mapping to type text.

3

If you are using dynamic mapping the first occurrence of a field will determine the mapping. As you have the Roll_no set to a number this is how it is getting mapped. As mentioned this will prevent documents with non-numeric values in that field from getting indexed as mappings can not be changed for existing indices and each field has to have a single mapping. If you know a field can have non-numeric values, format it as a string in the JSON document and it will get mapped as a string. You can also specify the mapping ahead of time through an index template.