Re: Double newline bug with DBD::Pg: Where best to fix?

From: ivan <ivan-pgsql-interfaces(at)420(dot)am>
To: pgsql-interfaces(at)postgresql(dot)org
Cc: Dan Lyke <danlyke(at)flutterby(dot)com>
Subject: Re: Double newline bug with DBD::Pg: Where best to fix?
Date: 2001-02-27 20:47:14
Message-ID: 20010227124713.A3629@cleanwhisker.420.am
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-interfaces

On Tue, Feb 27, 2001 at 03:21:58PM -0500, Tom Lane wrote:
> Dan Lyke <danlyke(at)flutterby(dot)com> writes:
> > If, using DBI with DBD::Pg, you call
>
> > $dbh->do('INSERT INTO xyz(stuff) VALUES('
> > $dbh->quote("a string with \n\n a two line break").')');
>
> > You'll get a string with a single line break back, because DBD::Pg
> > doesn't escape newlines into '\n'.
>
> There shouldn't be any need to convert newlines into \n in quoted
> strings; either way should work.
>
> > This is consistent with the interface via psql:
>
> > test=# insert into xyz(stuff) values('a string with
> > test'#
> > test'# a two line break');
> > test=# select stuff from xyz;
> > stuff
> > ----------------------------------
> > a string with
> > a two line break
> > (1 row)
>
> > (As you can see there are also some interesting issues with leading
> > spaces going on here).
>
> I would call this a clear bug in psql. I cannot replicate the loss of
> leading whitespace in current sources, but I do see the loss of the
> empty line. Turning on debug logging shows that the empty line has
> been stripped by psql, not the backend --- it's not there in the query
> that arrives at the backend.
>
> I tried a similar experiment in pgtclsh and didn't see any loss of
> blank lines, so the problem seems to be specific to psql. Your
> observation in DBI must be an independent bug in either DBD::Pg or
> DBI itself.
>
> > I don't have a copy of the SQL standard in front of me, so I'm not
> > sure whether this needs to be fixed in the PostgreSQL parser (seems
> > like the logical place, but we're into minutae of compatibility) or
> > DBD::Pg (the easy place, but maybe this is a hack that's going to have
> > to go in every interface).
>
> The Postgres parser is not broken, AFAICT. If you can find where this
> is happening in the DBI/DBD path, please submit a patch.

One of my cow orkers sent me a patch for this; it sounds like we tickled
the same bug you did. See below.

The other (unrelated) problem I'm having with DBD::Pg is that you can't
insert more than 64k of data using placeholders, i.e.

$sth = $dbh->prepare("INSERT INTO table VALUES ( ?, ?, ? )");
$sth->execute($a, $b, $c);

will *segfault* if $a $b or $c is larger than 64k.

--
meow
_ivan

--- /tmp/dbd-pg/DBD-Pg-0.95/dbdimp.c Mon Jul 10 10:47:51 2000
+++ ./dbdimp.c Fri Feb 23 15:17:37 2001
@@ -634,9 +634,22 @@

if (in_literal) {
/* check if literal ends but keep quotes in literal */
+ if (dbis->debug) {
+ fprintf(DBILOGFP, "*src = %c, in_literal = %c, *(src+1) = %c\n", *src, in_literal, *(src+1));
+ }
+#if 1
+ if (*src == in_literal) {
+ if (*(src+1) == in_literal) {
+ *dest++ = *src++;
+ } else {
+ in_literal = 0;
+ }
+ }
+#else
if (*src == in_literal && *(src-1) != '\\') {
in_literal = 0;
- }
+ }
+#endif
*dest++ = *src++;
continue;
}
@@ -1022,9 +1035,21 @@

if (in_literal) {
/* check if literal ends but keep quotes in literal */
+ if (dbis->debug) {
+ fprintf(DBILOGFP, "*src = %c, in_literal = %c, *(src+1) = %c\n", *src, in_literal, *(src+1));
+ }
+ if (*src == in_literal) {
+ if (*(src+1) == in_literal) {
+ *dest++ = *src++;
+ } else {
+ in_literal = 0;
+ }
+ }
+#if 0
if (*src == in_literal && *(src-1) != '\\') {
in_literal = 0;
- }
+ }
+#endif
*dest++ = *src++;
continue;
}

In response to

Browse pgsql-interfaces by date

  From Date Subject
Next Message J. T. Vermeulen 2001-02-28 12:55:00 libpq++: suggested patches for PgTransaction
Previous Message Dan Lyke 2001-02-27 20:39:47 Re: Double newline bug with DBD::Pg: Where best to fix?