[BUG] Following Symbols doesn't work with Pluralization lookup

[movermeyer]

What I tried to do

I'm creating an issue to capture the problem laid out in #480 by @Tietew

Using links/Symbols/aliases (whatever you call 'em) and pluralization (i.e., count) doesn't work.

autos: :cars
cars: 
  porsche:
    one: "I have %{count} Porsche 🚗"
    other: "I have %{count} Porsches 🚗"

I18n.t(:'autos.porsche', count: 1)

What I expected to happen

"I have 1 Porsche 🚗"

What actually happened

I18n::InvalidPluralizationData: translation data {:porsche=>{:one=>"I have %{count} Porsche 🚗", :other=>"I have %{count} Porsches 🚗"}} can not be used with :count => 1. key 'one' is missing.

What is happening?

The issue is occurring in the lookup of the linked data.

Since all the original parameters are passed into the sublookup, including count, it is attempting to perform the pluralization lookup prior to returning the resolved value and failing.

History of this issue

• This was noticed by Exclude :count option on retrieve link #480 and the fix was to exclude the count parameter within the Symbol resolving code.
• This broke Symbol resolving within default and an attempt was made to fix it in a different way (Preserve count option #503)
• Preserve count option #503 implicitly changed the definition of what was permitted in a translation file (namely, it broke nested keys within pluralization contexts), causing and Some translations are returned as hash after 1.7.1 #510 and Translations returned as hash when model-inflection has nested keys #514.
• Preserve count option #503 was partially reverted in 1b5e345, leaving us in the state we have today:
  • Using Backend::Pluralization (with a i18n.plural.rule) fails to resolve pluralization values after following Symbols (this issue, originally from Exclude :count option on retrieve link #480)
  • Using Backend::Base fails to handle nested keys within a pluralization context (Translations returned as hash when model-inflection has nested keys #514)

Q&A

Q: Why wasn't this noticed by the tests when the changes to Backend::Pluralization were reverted?

A: There was no test with both symbol resolution and pluralization in the Backend::Pluralization tests.

The #480 changes were reverted, but not the #503 changes, so the test added in #480 for Backend::Base didn't start failing.

Versions of i18n, rails, and anything else you think is necessary

i18n 1.12.0 (and many prior versions)