You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 21 Next »

Status

Current state: Voting

Discussion thread: here

JIRA: here

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

Motivation

Single Message Transforms (SMT), KIP-66, have greatly improved Connector's usability by enabling transforming input/output data format and content without the need for additional streaming applications. 

Though, these benefits have been limited by SMTs constraint to only lookup for fields available on the root of the data structure:

This KIP is aimed to include support for nested structures on the existing SMTs.

Proposed Changes

Nested notation

Using dots tends to be the most intuitive way to describe the path to nested fields in a record structures, e.g. jq already uses it[1], and will cover most of the scenarios.

However, dots are already allowed as part of element names on JSON (i.e. Schemaless) records (e.g. {'nested.key': {'val':42}}).

Instead of escaping them with backslashes — which in JSON configurations will lead to unfriendly configurations — it's proposed to follow a similar approach as the CSV format[2] to escape double-quotes by preceding it with the same character (double quotes in this case).

Then, for transform configurations, double-dots can be used to escape existing dots that are part of the field name.

[1] https://stedolan.github.io/jq/manual/#Basicfilters

[2] https://datatracker.ietf.org/doc/html/rfc4180 2.7

> If double quotes are used to enclose fields, then a double-quote appearing inside a field must be escaped by preceding it with another double quote.

Accessing values by deep-scan

There are scenarios to target multiple fields with the same name at different levels, e.g. dynamic/unknown structures, in a recursive manner.

For these cases, an asterisk can be used to search all elements within a path (e.g. similar to JsonPath[3], but using asterisk instead of .. ).

  • a.*.b  will access a  and then search all child objects for the field b, including arrays.

Deep scans are expected to return multiple paths. If only one element is found, a list of one path is returned. If no path is found, an empty list is returned.

SMTs have to define how to process paths lists.

scenariodatapathaccessfields
  1. a. Deep-scan as prefix to a field
{
  "k1": { "b": "b1" },
  "k2": { "b": "b2" }, 
  "k3": { "b": "b3" },
  "b": "b4"
}
*.b All the data structure is scanned to find a b  field.
  • k1.b
  • k2.b
  • k3.b
  • b
  1. b. Deep-scan as prefix to a nested field
{
  "k1": { "b": { "c": "c1" } },
  "k2": { "b": { "c": "c2" } }, 
  "k3": { "b": { "c": "c3" } },
  "b": { "c": "b4" } 
}
*.b.c All the data structure is scanned to find a b  field.
  • k1.b.c
  • k2.b.c
  • k3.b.c
  • b.c
2. Deep-scan in the mid of a path expression
{ "a": {
  "k1": { "b": { "c": "c1" } },
  "k2": { "b": { "c": "c2" } }
}, "a2": {
  "k3": { "b": { "c": "c3" } }
}}
a.*.b 

Find the prefix element (a).

If found, scan the nested structure as in the scenarios 1.a and 1.b.

  • a.k1.b
  • a.k2.b
3. a. Deep-scan as suffix to an array


{ "a": [ "v1", "v2", "v3"] }
a.*If a deep-scan is the last path expression under an array, all the array items are returned
  • a.0
  • a.1
  • a.2
3. b. Deep-scan as suffix to a struct/map
{ "a": {
"k1": "v1",
"k2": "v1",
"k3": { "b": "c" }
}
a.*If a deep-scan is the last path expression under a struct/map, all the child fields are returned
  • a.k1
  • a.k2
  • a.k3
3.c. Deep-scan as suffix to a field
{ "a": {
"k1": "v1",
"k2": "v1"
}
a.k1.*If a deep-scan is the last path expression under a field, an empty list is returned.<empty>
4. a. Escaping asterisks
{ "*": {
"k1": "v1",
"k2": "v1"
}
**.k1

If an asterisk is already being used as a field name, the to target that field, a double-asterisk is used.

This is only needed when the field's full name is only asterisks.

  • *.k1
4. b. Escaping multiple asterisks
{ "**": {
"k1": "v1",
"k2": "v1"
}
***.k1If more than one asterisk is used, then the number of asterisk to escape the deep-scan expression should include an additional
  • **.k1


Accessing Arrays

Arrays can be accessed in different ways and at different levels.


scenariodatapathfields
Accessing array as a single element
{ "a": [ "a1", "a2", "a3"]
a
  • a
Accessing all items
{ "a": [ "a1", "a2", "a3"]
a.*
  • a.0
  • a.1
  • a.2
Accessing an item by index
{ "a": [ "a1", "a2", "a3"]

a.<index>

a.0

  • a.0
Accessing elements within objects
{ "a": [ { "b": "b1" }, { "b": "b2" } ]
a.*.b
  • a.0.b
  • a.1.b
Accessing an item by index, and its elements within an object
{ "a": [ { "b": "b1" }, { "b": "b2" } ]
a.0.b
  • a.0.b

Public Interfaces

From the existing list of the SMTs, there are the following to be impacted by this change:

New configuration flags

NameTypeDefaultImportanceDocumentation
field.syntax.version STRING v1HIGH 

Permitted values: v1 , v2 . Defines the version of the syntax to access fields. If set to "v1", then the field paths are limited to access the elements at the root level of the struct or map. If set to "v2", the syntax will support accessing nested elements. o access nested elements, dotted notation is used. If dots are already included in the field name, then dots themselves can be used to represent dots part of the field name. e.g. to access elements from a struct/map named "same.field", the following format can be used to access its elements: "same..field.element".

This configuration will affect all the field paths used by the transform.

These flags will be added conditionally to some SMTs, as described below.

Affected SMTs

Cast

Changes:

  • Extend spec to support nested notation.
  • Supports arrays and deep-scan to access multiple fields.
    • If the paths returned do not match a supported type to be converted by spec, then ignores.

Examples:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.Cast$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.spec": "k1:string,parent.child.k2:int64"
}
{
  "k1": "123",
  "parent": {
    "child": {
      "k2": 123    
    }
  }
}
2. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.Cast$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.spec": "k1:string,parent..child.k2:int64"
}
{
  "k1": "123",
  "parent.child": {
    "k2": 123
  }
}
3. Multiple paths found
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": "123"    
    }
  },
  "parent2": {
    "child": {
      "k2": "123"    
    }
  } 
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.Cast$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.spec": "k1:string,*.child.k2:int64"
}
{
  "k1": "123",
  "parent1": {
    "child": {
      "k2": 123    
    }
  },
  "parent2": {
    "child": {
      "k2": 123    
    }
  }
}
4. Multiple paths found, but some types do not match and are ignored
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": "123"    
    }
  },
  "parent2": {
    "child": {
      "k2": {}    
    }
  } 
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.Cast$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.spec": "k1:string,*.child.k2:int64"
}
{
  "k1": "123",
  "parent1": {
    "child": {
      "k2": 123    
    }
  },
  "parent2": {
    "child": {
      "k2": {}    
    }
  }
}

ExtractField

Changes:

  • Extend field to support nested notation.
  • Supports arrays and deep-scan to access multiple fields.
    • The result produces an array, even for one or no field is found.

Example:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ExtractField$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.field": "parent.child.k2"
}
"123"
2. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ExtractField$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.field": "parent..child.k2"
}
"123"
3. Nested field, an object returned.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ExtractField$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.field": "parent.child"
}
{ "k2": "123" }
3. Nested field, an array returned.
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": "123"    
    }
  },
  "parent2": {
    "child": {
      "k2": "234"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ExtractField$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.field": "*.child.k2"
}
[ "123", "234" ]

HeaderFrom

Changes:

  • Extend fields to support nested notation.
  • As this SMT affects only existing fields, additional configurations will not be required.
  • Does not support multiple values (e.g. deep scan or array), if multiple paths are found, only the first one is used.

Example:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.HeaderFrom$Value",
"transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "k1,parent.child.k2",
"transforms.smt1.headers": "k1,k2"
}
headers:
- k1=123
- k2="123"
2. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.HeaderFrom$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "k1,parent..child.k2",
"transforms.smt1.headers": "k1,k2"
}
headers:
- k1=123
- k2="123"
3. Nested field, an array returned.
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": "123"    
    }
  },
  "parent2": {
    "child": {
      "k2": "234"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ExtractField$Value",
"transforms.smt1.fields": "k1,*.child.k2",
"transforms.smt1.headers": "k1,k2"
}
headers:
- k1=123
- k2="123"

MaskField

Changes:

  • Extend fields to support nested notation.
  • Supports arrays and deep-scan to access multiple fields.

Example:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.MaskField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "parent.child.k2"
}
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": ""    
    }
  }
}
2. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.MaskField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "parent..child.k2"
}
{
  "k1": 123,
  "parent.child": {
    "k2": ""
  }
}
1. Nested field.
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": "123"    
    }
  },
  "parent2": {
    "child": {
      "k2": "234"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.MaskField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "*.child.k2"
}
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": ""    
    }
  },
  "parent2": {
    "child": {
      "k2": ""    
    }
  }
}

ReplaceField

Changes:

  • Extend the include and exclude lists
  • Supports arrays and deep-scan to access multiple fields.

Example:

scenarioinputsmtoutput
1. Nested field. Drop field
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ReplaceField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.exclude": "parent.child.k2"
}
{
  "k1": 123,
  "parent": {
    "child": {
    }
  }
}
2. Nested field. Drop struct
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ReplaceField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.exclude": "parent.child"
}
{
  "k1": 123,
  "parent": {
  }
}
3. Nested field. Include field
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123",
      "k3": "234"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ReplaceField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.include": "parent.child.k2"
}
{
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
4. Nested field. Include struct
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123",
      "k3": "234"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ReplaceField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.include": "parent.child"
}
{
  "parent": {
    "child": {
      "k2": "123",
      "k3": "234"    
    }
  }
}
5. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ReplaceField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.renames": "parent..child.k2:field2"
}
{
  "k1": 123,
  "parent.child": {
   "field2": "123"
  }
}
6. Multiple fields
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": "123"    
    }
  },
  "parent2": {
    "child": {
      "k2": "234"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ReplaceField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.renames": "*.child.k2:field2"
}
{
  "k1": 123,
  "parent1": {
    "child": {
      "field2": "123"    
    }
  },
  "parent2": {
    "child": {
      "field2": "234"    
    }
  }
}

TimestampConverter

Changes:

  • Extend fields to support nested notation.
  • Supports arrays and deep-scan to access multiple fields.

Example:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": 1556204536000         }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.TimestampConverter$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.field": "parent.child.k2",
"transforms.smt1.format": "yyyy-MM-dd",
"transforms.smt1.target.type": "string"
}
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "2014-04-25"         }
  }
}
2. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
      "k2": 1556204536000         }
  }
}


{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.TimestampConverter$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.field": "parent..child.k2",
"transforms.smt1.format": "yyyy-MM-dd",
"transforms.smt1.target.type": "string"
}



{
  "k1": 123,
  "parent.child": {      "k2": "2014-04-25"   }
}

ValueToKey

Changes:

  • Extend fields to support nested notation.
  • Supports arrays and deep-scan to access multiple fields.
    • The result produces an array, even for one or no field is found.

Example:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ValueToKey", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "parent.child.k2"
}
"123"
2. Nested struct to Key.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ValueToKey", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "parent.child"
}
{
  "k2": "123"    
}
3. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ValueToKey", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "parent..child.k2"
}
"123"
4. Multiple values to key
{
  "k1": 123,
  "parent1": {
    "child": {
      "k2": "123"    
    }
  },
  "parent2": {
    "child": {
      "k2": "234"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.ValueToKey", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.fields": "*.child.k2"
}
[ "123", "234 ]

InsertField

Changes:

  • Extend *.field to support nested notation.
  • Does not support multiple values (e.g. deep scan or array)

New configurations (additional to field.syntax.version described above):

Name

TypeDefaultImportanceDocumentation
field.on.missing.parentSTRINGcreateMEDIUMPermitted values: create, ignore. Defines how to react when the field to act on does not have a parent and "field.style" is "nested". If set to "create", then the SMT will create the parent struct/map when it does not exist. If set to "ignore", then it will SMT have no effect.
field.on.existing.fieldSTRINGoverwriteMEDIUMPermitted values: overwrite, ignore. Defines how to react when the field to act on already exists. If set to "overwrite", then the SMT will be applied to the existing field. If set to "ignore", then it will SMT have no effect.

Example:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.InsertField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.static.field": "parent.child.k3"
"transforms.smt1.static.value": "v3" 
}
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123",
      "k3": "v3"   
    }
  }
}
2. Nested field, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.InsertField$Value",  "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.static.field": "parent..child.k3"
"transforms.smt1.static.value": "v3" 
}
{
  "k1": 123,
  "parent.child": {
    "k2": "123",
    "k3": "v3"
  }
}
3. Nested field with the parent missing
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.InsertField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.static.field": "parent.other.k3"
"transforms.smt1.static.value": "v3" 
}
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"  
    },
    "other": {
      "k3": "v3"  
    }
  }
}
4. Nested field with the parent missing, and ignore is set
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.InsertField$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.static.field": "parent.other.k3"
"transforms.smt1.static.value": "v3",
"transforms.smt1.field.on.missing.parent": "ignore"
}
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"  
    }
  }
}
5. Nested field with the parent missing
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.InsertField$Value",  "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.static.field": "parent.child.k2"
"transforms.smt1.static.value": "456"
}
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "456"  
    }
  }
}
6. Nested field with the parent missing, and ignore is set
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.InsertField$Value",  "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.static.field": "parent.child.k2"
"transforms.smt1.static.value": "456",
"transforms.smt1.field.on.existing.field": "ignore"
}
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"  
    }
  }
}

HoistField

Changes:

  • Add a hoisted config to point to a specific path to hoist.

  • Does not support multiple values (e.g. deep scan or array)

New configurations:

Name

TypeDefaultImportanceDocumentation
hoisted STRING <empty>MEDIUM Path to the element to be hoisted. If empty, the root struct/map is hoisted.

Examples:

scenarioinputsmtoutput
1. Nested field.
{
  "k1": 123,
  "parent": {
    "child": {
      "k2": "123"    
    }
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.HoistFIeld$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.hoisted": "parent.child.k2",
"transforms.smt1.field": "other"
}
{
  "k1": 123,
  "parent": {
    "child": {
      "other": {
        "k2": "123"
      }    
    }
  }
}
2. Nested struct, when field names include dots
{
  "k1": 123,
  "parent.child": {
    "k2": "123"
  }
}
{
"transforms": "smt1",
"transforms.smt1.type": "org.apache.kafka.connect.transforms.HoistFIeld$Value", "transforms.smt1.field.syntax.version": "v2",
"transforms.smt1.hoisted": "parent..child",
"transforms.smt1.field": "other"
}

{
  "k1": 123,
  "other": {
    "parent.child": {
      "k2": "123"
    }
  }
}

Non-affected SMTs

These SMT do not require nested structure support:

  • DropHeaders: Drop one or multiple headers.
  • Filter: Drops the whole message based on a predicate.
  • InsertHeader: Insert a specific message to the header.
  • RegexRouter: Acts on the topic name.
  • SetSchemaMetadata: Acts on root schema.
  • TimestampRouter: Acts on timestamp.
  • Flatten: Acts on the whole key or message. 

Compatibility, Deprecation, and Migration Plan

Existing SMT configurations will not be affected by these changes as the default field.style  is plain, which represents the current behavior.

Rejected Alternatives

Keep ExtractField as it is and use it multiple times until reaching nested fields

This KIP proposes simplifying this configuration by replacing multiple invocations with only one nested one.

Use dots as the only separator and escape with backslashes when collides

Trying to keep only one separator, one of the alternatives is to use dots to separate; if it collides with the existing field names use backslashes "\" to represent dots that are part of the name e.g.  "this.field" (which would refer to the nested field "field" under the top-level "this" field), and "this\.field" (which would refer to the field named "this.field").

However, backslashes are also used by JSON. This could lead unfriendly configurations like "this\\\\.is\\\\.not\\\\.very\\\\.readable"

Use custom separators for edge cases

Using double dots to escape separators is another alternative to try sticking to using only dots as a field separator.

Comparing:

With double dotsWith separator
{
  "transforms": "cast",
  "transforms.cast.field.syntax.version": "v2",          
  "transforms.cast.type": "..."
  "transforms.cast.spec": "address..personal.country:string"
}
{   
  "transforms": "cast",
  "transforms.cast.field.syntax.version": "v2",
  "transforms.cast.field.separator": "/", 
  "transforms.cast.type": "..."
  "transforms.cast.spec": "address.personal/country:string",
}

Even if using custom separators represent a more explicit configuration, there is always the possibility that all the separators are already included as part of the field name, leading to issues and request for changes.

To avoid this, this KIP proposes using the approach to precede dots with another to escape itself.

Use JSONPath notation to access nested elements

JSONPath[1] was a proposed alternative to the nested notation. A drafted version of the KIP with examples using the proposed notation is outlined here: [DRAFT] KIP-821: Connect Transforms support for nested structures (JsonPath-based draft)

The following limitations were found:

  • The JSONPath spec is too extensive for the use-cases included in this KIP.
  • A sub-set of JSONPath was proposed, but the custom spec ends up being more complex than the notation proposed here.
    • A sub-set will imply not using existing dependencies. Though adding an existing dependency would also reduce the chance of the KIP being accepted as the risk for external vulnerabilities will increase.
    • The sub-set will require users to learn JSONPath, and then what's covered and what's not by the custom implementation.

Given these cons, the KIP is preferring the dotted notation.

[1] https://github.com/json-path/JsonPath

Use named styles instead of syntax versions

Was considered to use a configuration to name the styles to target fields:

  • field.style  with valid values: "plain", "nested".

Even though this configuration is self-describing, it limits the semantics of the values.

Instead, the KIP is considering a versioned configuration "field.syntax.version" to avoid affecting current behavior and make it easier to extend by including compatible changes on the same version.

Use configuration flag per SMT instead of per-field configuration

Instead of adding a configuration under each field config, e.g. include.syntax.version , the KIP proposed to have a single configuration per SMT, to affect all the input fields.

  • No labels