Use UUIDs as indentifiers for sessions in public #176
Conversation
|
@yshmarov Do you have an opinion on this? Also very okay if you don't want to bother, just let me know 😅 |
Heyyy @mikker, I'm loving this PR! 🤗 Having to add postgres plugins as dependencies is not cool very; it's much better if the gem is database-agnostic 👍 I am happy with adding an additional |
| @@ -18,9 +18,11 @@ | |||
| t.datetime "expires_at", precision: nil, null: false | |||
| t.datetime "claimed_at", precision: nil | |||
| t.string "token_digest", null: false | |||
| t.string "identifier", null: false | |||
There was a problem hiding this comment.
hmm, with null: false it might be worth considering column "identifier" of relation "passwordless_sessions" contains null values
There was a problem hiding this comment.
I think we should go with null: false as the new default in new installations but null: true in the upgrade instructions? That way it'll handle legacy tables and Rails can still validates :presence
|
looks lovely |
mikker
force-pushed
the
use-uuids-in-public
branch
from
f2d888e to
967562a
Compare
As per discussion with @yshmarov in #169, a proposal to use UUIDs as Session identifiers in public.
It might be a bit more space and cpu optimized to use UUIDs as the primary keys directly but AFAIK that involves having to turn on a Postgres plugin
pgcryptoguides ref.I feel like that puts unnecessary hassle on the user's end re: using DBs you don't control, etc. I'm also hesitant to add another required migration upgrade.
This way we can do the switch with only a vaguely breaking change (lol, if that's a thing)
@yshmarov what do you think?