[data/values] Make '---' in values.yaml optional

[jessehu]

Is the '---' in values.yaml required? We hit an issue related to '---' and yq (see mikefarah/yq#351).

#@data/values
---
key: value

Without '---', ytt gives this error. But '---' is optional per yaml spec.

ytt: Error: Expected YAML document to be annotated with data/values but was *yamlmeta.MapItem

[cppforlife]

Is the '---' in values.yaml required?

it's required because @data/values annotation is associated with a document. if it was removed there would be no way to add @data/values to the document.

[jessehu]

So can ytt adds '---' implicitly for a values.yaml if it's not there?

[cppforlife]

ytt knows that yaml document is data values is by finding documents with that annotation. file name itself (values.yml) isnt indicative that it contains data values.

[jessehu]

Hi @cppforlife, I meant when ytt reads a file my-data-values.yaml and finds #@data/values at the beginning, can ytt implictly adds --- if there is no such line --- after #@data/values? yaml spec doesn't require --- at the beginning of a file. It's interesting that current ytt requires --- exists in the data values yaml , but doesn't require --- exists in other ytt templates yaml file.

[cppforlife]

can ytt implictly adds --- if there is no such line --- after #@data/values

something like this?

#@data/values
foo: blah
other: another

we could make it wrong however from more of a purist ytt perspective this isnt "conceptually valid" since @data/values annotation is only apply to foo: blah instead of both of the kvs. this would be pretty significant step away from an understanding that a user might have about how annotations work.

[jessehu]

I see what you meant. Does foo: blah with the annotation #@data/values make any sense to ytt ? If not, there is no ambiguity. Then why not ytt interprets the following valid yaml file

#@data/values
foo: blah
other: another

as

#@data/values
---
foo: blah
other: another

[cppforlife]

If not, there is no ambiguity. Then why not ytt interprets the following valid yaml file

there is no ambiguity in this case, but it goes against all other cases how annotations are set up in ytt (annotation never is applied to a parent node. only to self or children).

it also seems that this is driven by a deficiencies in other tools. mikefarah/yq#351 (comment) seems to indicate that yq doesnt understand documents at all.

[jessehu]

We can put aside yq. One of the principle of ytt is that Works directly with YAML structures, not text, but it dosen't work with this valid yaml file:

#@data/values
foo: blah
other: another

[ewrenn8]

ytt assigns the annotations to the following yaml node, which in this case is the map entry foo. Your example above would be the same as saying:

---
#@data/values
foo:blah
other: another

To apply the annotation to the implied document start would not make much sense. For example, users would expect:

#@yaml/map-key-override
foo: bar

to translate to

---
#@yaml/map-key-override
foo: bar

not

#@yaml/map-key-override
---
foo: bar

[cppforlife]

given above explanation from eli, ill close this issue.