Can Active Record stop treating blank? as nil in

[SamSaffron]

I have been looking at some very minor perf issues that have been bugging me in Active Record.

In particular this revert by @carlosantoniodasilva :

111611b#commitcomment-2894962

At Discourse I notices blank? on string is being called a lot:

This is mostly originating from the reverted change in AR, once you replace apply the changes in the reverted commit the cost drops way down:

Let's pretend String#blank? was empty?

---

The revert bothers me cause its pointlessly slower and dictates a rather odd semantics:

Any unicode space is always treated as nil for boolean columns.

[2] pry(main)> u.moderator = "  \r\n\v\f\r\s\u00A0\u2000\u2001\u2002\u2003\u2004\u2005\u2006\u2007\u2008\u2009\u200A\u202F\u205F\u3000"
=> "  \r\n\v\f\r               　"
[3] pry(main)> u.save
   (0.4ms)  BEGIN
  User Exists (3.7ms)  SELECT 1 AS one FROM "users" WHERE ("users"."email" = 'bla@gmail.com' AND "users"."id" != 3650) LIMIT 1
   (0.7ms)  UPDATE "users" SET "moderator" = NULL, "updated_at" = '2013-03-29 20:57:34.968383' WHERE "users"."id" = 3650
   (0.6ms)  COMMIT

A single char that is not a unicode space is treated as false

 u.moderator = "a"
=> "a"
[16] pry(main)> u.save
   (0.3ms)  BEGIN
  User Exists (0.5ms)  SELECT 1 AS one FROM "users" WHERE ("users"."email" = 'bla@gmail.com' AND "users"."id" != 3650) LIMIT 1
   (0.3ms)  UPDATE "users" SET "moderator" = 'f', "updated_at" = '2013-03-29 21:02:18.428719' WHERE "users"."id" = 3650
   (0.5ms)  COMMIT
=> true

---

@tenderlove agrees that treating non empty strings as null is odd and a db adapter issue as opposed to something that should be handled in core.

I can go ahead and put a PR in place that reverts this, and nukes some of the odd tests for this behavior like.

7f160b0

introduced with the resolution of:

#6045

---

For Discourse I am working around this by implementing blank? natively, https://github.com/SamSaffron/fast_blank ... and staying consistent with .strip.length == 0 ... treating odd unicode spaces as text.

However, this odd AR behaviour kind of bothers me, and I am not sure why:

"\r\n\v\f\r\s\u00A0\u2000\u2001\u2002\u2003\u2004\u2005\u2006\u2007\u2008\u2009\u200A\u202F\u205F\u3000" really means NULL and "\u1999" means not NULL

Can we revert connection_adapters/column.rb to do empty? checks instead, reverting all the problem tests?

[tenderlove]

Can we revert connection_adapters/column.rb to do empty? checks instead, reverting all the problem tests?

👍 send a PR please.

[SamSaffron]

Cool! will do, first thing next week.

[SamSaffron]

@tenderlove @carlosantoniodasilva , sent through a PR, also corrected the test suite so this is not reverted again.

Only breaking change is on booleans, time and data are functionally equivalent.