Does the spec leave implementation of indexes on the substr() function
less than one undefined? By mistake, I had a substr() invocation with
an initial index of zero (would be nice if all the computer languages of
the world could agree to a single definition.) Oracle silently treats
this the same as 1 (one), while I just learned that PG treats indexes
less than 1 as a sliding window off the left of the string. Oracle
defines negative indexes to mean "from the end of the string."
So, given a string "abcdefg":
Oracle substr('abcdefg', 1, 4) = 'abcd'
Oracle substr('abcdefg', 0, 4) = 'abcd'
Oracle substr('abcdefg', -1, 4) = 'd'
PG substr('abcdefg', 1, 4) = 'abcd'
PG substr('abcdefg', 0, 4) = 'abc'
PG substr('abcdefg', -1, 4) = 'ab'
--
Guy Rouillier