float8 transmitted in ascii

From: Timo Nentwig <postgres(at)nentwig(dot)biz>
To: pgsql-jdbc(at)postgresql(dot)org
Subject: float8 transmitted in ascii
Date: 2015-04-29 15:40:44
Message-ID: alpine.LNX.2.03.1504291738060.41809@nentwig.biz
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-jdbc

Hi!

9.1-901-1.jdbc4 (and probably also the latest 9.4 driver, did most testing with
the 9.1 one, though [on pg 9.4.1])

postgres=# select 47.88::DOUBLE PRECISION,
(47.88::DOUBLE PRECISION)::text,
(47.88::DOUBLE PRECISION)::numeric,
replace((47.88::DOUBLE PRECISION)::text, '.', ',');

float8 | text | numeric | replace
--------+-------+---------+---------
47.88 | 47.88 | 47.88 | 47,88

Good. What I actually get with JDBC doesn't look quite as good, though:

47.88
47.8800000000000026
47.88
47,8800000000000026

So, I looked into the driver and the first surprise was that for the first 2
columns the driver gets an identical 19-byte array; despite one is a float8 (and
8!=19) and the other is supposedly Ba string (with 5 characters).

Here it is (AbstractJdbcResultSet.this_row):
52,55,46,56,56,48,48,48,48,48,48,48,48,48,48,48,48,50,54

Well, that appears to be the 19-char *string* '47.8800000000000026' in ascii.

And indeed:

return toDouble( getFixedString(columnIndex) );
return Double.parseDouble(s);

Srsly?! Doubles are sent over the wire in "some" text representation and then
parsed in the JDBC driver? Really?

What I still don't get, though: if this is what postgres sends over the wire,
why does psql/pgadmin print the 4th (and 2nd) column correctly?

thx
tcn

P.S.:

if ("hstore".equals(getPGType(columnIndex))) {
return HStoreConverter.toString((Map) obj);
}

So, internally we already have a java.util.Map and there are a few 'solutions'
out there that *parse* the string representation back into java.util.Map? Srsly?

Responses

Browse pgsql-jdbc by date

  From Date Subject
Next Message Dave Cramer 2015-04-29 22:51:34 Re: float8 transmitted in ascii
Previous Message Gonzalo Ortiz Jaureguizar 2015-04-27 12:20:55 TIMESTAMP value binary transfer problem