Re: PATCH: ordering of hash table scans

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Feb 6,  9:42pm, Peter Stephenson wrote:
}
} I have a vague memory that it's deliberate that we don't use strcoll()
} here, which would make the sorting locale dependent and in particular
} possibly case-insensitive.

I have the same (probably equally vague) memory, for what that's worth.

} I then noticed that it's actually thoroughly inconsistent which objects
} are sorted before printing out and which are printed out any old how as
} they emerge from the hash table. [...]
} 
} I've tidied this up so that in any place where the result is being
} directly printed sorting is applied.  Maybe someone can think of cases
} where this isn't appropriate, but given the current half-baked state
} it's hard to see.

The only thing I could imagine is if someone were relying on use of a
hashtable to "randomize" the order of some list of items; but that
seems unlikely, and certainly there was no promise it would work.