Remove ingest conditionals _type deprecation warning

[joegallo]

_type isn't special in ingest conditional scripts anymore, and we shouldn't warn about it as if it is.

With this pipeline:

PUT _ingest/pipeline/underscore_type { "processors": [ { "script": { "source": "ctx._foo = ctx._bar" # this is a field } }, { "script": { "source": "ctx._baz = ctx._type" # no warning from this, it's not a _conditional_ } }, { "set": { "if": "ctx._type == null", # this warns on `main` but won't anymore after this PR "field": "message_1", "value": "test" } }, { "set": { "if": "ctx._missing == null", # this is just another field "field": "message_2", "value": "some other test" } } ] } 

This _simulate call:

POST _ingest/pipeline/underscore_type/_simulate { "docs": [ { "_index": "index", "_id": "id", "_source": { "hello": "world", "_bar": "quux" } } ] } 

gives this result on main:

// Warning: 299 Elasticsearch-9.2.0-494dc7075a299cb5edb06c36681601550c4fcdae "[types removal] Looking up doc types [_type] in scripts is deprecated." { "docs" : [ { "doc" : { "_index" : "index", "_version" : "-3", "_id" : "id", "_source" : { "_baz" : null, "message_2" : "some other test", "message_1" : "test", "_bar" : "quux", "hello" : "world", "_foo" : "quux" }, "_ingest" : { "timestamp" : "2025-09-16T21:36:01.143913Z" } } } ] } 

The result would be the same after this PR, there just wouldn't be a warning anymore -- that is, _type is just a regular old non-special field, like _foo or _bar or whatever (edit: that is, at least in terms of reads).

Related to #134816, in that I'm pulling a part of that out as its own PR for searchability and accounting purposes.

[elasticsearchmachine]

Pinging @elastic/es-data-management (Team:Data Management)

[elasticsearchmachine]

Hi @joegallo, I've created a changelog YAML for you.

[joegallo]

@masseyke for the code review (maybe I need more tests?), and @dakrone for the process review (that is, "yes, this is a good change to make").

[masseyke]

LGTM

[dakrone]

Is there a reason we wouldn't do:

FUNCTIONS = Map.of("_type", ignored -> "_doc");

and just hardcode it to return _doc?

[dakrone]

My guess would be because it already does that in Painless?

[joegallo]

That'd be a change in the behavior. The behavior has been (for a while, I could track down the PR that changed it, but I'm 95% sure I didn't write it so I don't know what I'm looking for off the top of my head) that the _type is null (in both the conditional and the ordinary script context):

POST _ingest/pipeline/_simulate { "pipeline": { "processors": [ { "script": { "source": "ctx.type_from_script = ctx._type" } }, { "script": { "if": "ctx._type == null", "source": "ctx.type_from_conditional_was_null = true" } } ] }, "docs": [ { "_index": "index", "_id": "id", "_source": { "hello": "world" } } ] } 

results in:

{ "docs" : [ { "doc" : { "_index" : "index", "_version" : "-3", "_id" : "id", "_source" : { "type_from_script" : null, "type_from_conditional_was_null" : true, "hello" : "world" }, "_ingest" : { "timestamp" : "2025-09-17T11:33:43.056173Z" } } } ] } 

Before and after this PR the above script works the same (that is, _type is null), it's just that prior to this PR a deprecation would be generated only for the conditional part of the one script. All this PR does is remove the deprecation warning, other than that doesn't change the behavior of the scripts in question.

[joegallo]

I haven't found a PR that clearly is the one that makes the _type be null for these, but I have confirmed that if you run 8.0.0 the behavior is the same as I show above. So this has been the behavior since at least February 2022.

[dakrone]

Ahh okay, I hadn't realized it was null, so removing it makes sense. Thanks for the clarification.

[dakrone]

LGTM