Inconsistent ::bit(N) and get_bit()? - Mailing list pgsql-hackers

From Boszormenyi Zoltan
Subject Inconsistent ::bit(N) and get_bit()?
Date
Msg-id 4C62AA44.6000503@cybertec.at
Whole thread Raw
Responses Re: Inconsistent ::bit(N) and get_bit()?
List pgsql-hackers
Hi,

I just came across the following confusing thing.

zozo=# create table bit_test(i integer);
CREATE TABLE
zozo=# insert into bit_test values (1), (2), (3);
INSERT 0 3
zozo=# select i, i::bit(2), get_bit(i::bit(2), 1) as bit1,
get_bit(i::bit(2), 0) as bit0 from bit_test;i | i  | bit1 | bit0
---+----+------+------1 | 01 |    1 |    02 | 10 |    0 |    13 | 11 |    1 |    1
(3 rows)

So, conversion from int to bitstring creates a readable bitstring, i.e.
the least significant bit is the rightmost one. But get_bit() on the same
bit string works in the opposite order. The only description about get_bit
I found in the 9.0beta docs are in
http://www.postgresql.org/docs/9.0/static/functions-binarystring.html#FUNCTIONS-BINARYSTRING-OTHER

-------------------------------------
Function     Return Type     Description     Example     Result


||
|get_bit|(string, offset)     int     Extract bit from string
get_bit(E'Th\\000omas'::bytea, 45)     1

-------------------------------------
||
and in
http://www.postgresql.org/docs/9.0/static/functions-bitstring.html

-------------------------------------
||The following functions work on bit strings as well as binary strings:
|get_bit|, |set_bit|.
-------------------------------------

Shouldn't it at least be documented in more depth? Say, get_bit(, N)
provides the Nth bit (0-based) counting from the leftmost bit?
I would certainly appreciate a warning spelled out about this
so if you convert a number to bitstring of length N and you want the
Mth bit (according to any programming language) then you need to use
get_bit(..., N-1-M).

Best regards,
Zoltán Böszörményi



pgsql-hackers by date:

Previous
From: Robert Haas
Date:
Subject: Re: RecordTransactionCommit() and SharedInvalidationMessages
Next
From: Tom Lane
Date:
Subject: Re: Regression tests versus the buildfarm environment