Symmetry of maps in class_table_inheritance plugin

[adam12]

(I don't remember raising this before but having deja-vu so maybe I have raised it before)

I've used the class_table_inheritance plugin a few times, but this one generally catches me quite frequently.

plugin :class_table_inheritance, key: kind,
  model_map: {
    "executive" => :Executive, # Symbol for class
    "staff" => "Staff" # String for class
  },

  table_map: {
    :Executive => "executive_employees", # Symbol for class (this works)
    "Staff" => "staff_employees" # String for class (this silently fails but has symmetry with model_map)
  }

When using String keys for table_map, I've found the queries to be incorrect, with the INNER JOIN never being used. I'm presuming the lookup misses.

Is it worth updating the plugin to automatically convert the keys to Symbols in the table_map or using a second fetch for the String version?

[jeremyevans]

model_map converts database values (strings by default, but could be other values) into model class names, and both strings and symbols are supported as values (just as const_get supports both).

table_map maps class name symbols to subclass table names. Both keys and values should be symbols, as specified by the documentation. String values work because Sequel will treat a string table name as an identifier, as you cannot SELECT FROM a string in SQL (that's only due to backwards compatibility, I would drop that behavior otherwise).

Using symbols for model_map keys wouldn't work. So supporting strings for table_map keys is not something I would generally be in favor of. In your example, all strings other than the model_map keys should ideally be symbols.

However, Sequel does already have access to the string, since Class#name returns a string and not a symbol. So I guess there is no harm in using a string we already have for lookup, if the lookup by symbol fails. Can you submit a pull request for that?

[adam12]

Can you submit a pull request for that?

Let me look into this.