Hello!
For example i have an schema Photo.

{
  "filename": "photo123.jpg",
  "content_type": "image/jpeg",
  "versions": {
     "original": {...},
     "thumbnail": {...}
     "facebook_version": {...}, 
     # ...
     # and some more versions
     # ...
  }
}

And there is a list of versions that should be present, but some other versions are optional.

After i took a look at this video - https://www.youtube.com/watch?v=sPmHP9ZPOWc, i see 2 cases to define data structure:

1. Schema-less data (Map → JSON → Map)

   • Set versions field to :map type in Ecto and that will be interpreted as JSONB in Postgres;
   • Implement nested versions data validations on Photo changeset level.

2. Embedded schema (Schema → JSON → schema)

   • define embedded Versions schema with full list of versions
   • all data validations functions are invoking at Versions changeset level

Objective
I am developing API application. And there is what i need from my data structure:

• secure and trusted data saving
• Convenient and understandable structure for querying and sending to frontend client

The second way looks more secure. But embedded schema it’s not dynamic data. I don’t want to store empty version fields in my DB, if some optional versions are absent(actually I do not know if this is a good idea). As i understood, the dynamic fields can’t be in Ecto embedded schema. Am i wrong?