Gregory Stark wrote:
"Ken Johanson" <[EMAIL PROTECTED]> writes:
Tom Lane wrote:
SQL92 section 6.1 <data type> quoth
<character string type> ::=
CHARACTER [ <left paren> <length> <right paren> ]
| CHAR [ <left paren> <length> <right paren> ]
...
4) If <length> is omitted, then a <length> of 1 is implicit.
Therefore, writing just "char" is defined as equivalent to "char(1)".
However when length is not defined I think it will generally be safe(r) to
auto-size. In the grand scheme auto-size creates much more sensible output than
a 1-char wide one (even if right-padded to max char-length of the type).
Sure, but you're a prime candidate for understanding the value of following
the spec if you're trying to write software that works with multiple
databases.
The spec has diminished in this (CAST without length) context:
a) following it produces an output which has no usefulness whatsoever
(123 != 1)
b) all the other databases chose to not follow the spec in the context
of cast and char with implicit length.
When the length is unqualified, a cast to char should one of:
1) failfast
2) auto-size to char-count (de facto)
3) pad to the max-length
It's a bit crazy to be using CHAR and then complaining about padding...
I did say earlier that I could at least accept padding to the max-char
length, even though in my use-case it wont work.
---------------------------(end of broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?
http://www.postgresql.org/docs/faq
