Обсуждение: Re: BUG #4083: Return type of MAX and MIN of a VARCHAR column is TEXT

Peter Eisentraut wrote:

> Tom Lane wrote:
> > I think you need to file a bug with Zeos to fix whatever their
> > problem is with text.
>
> Well, to be fair, type text is not SQL standard.  But surely Zeos
> won't get very far with this superstrict typing approach.  At least a
> little bit more context might help us understand.

Sure. Zeos allows connection to multiple servers, not just PostgreSQL;
for example, Firebird, MySQL and MSSQL.

Zeos (as well as the Borland Database Engine, which it just mimics in
this sense) assumes that VARCHAR fields in general (not just
PostgreSQL's) are textual fields with a limit of 255 characters for all
supported servers, which are graphically associated with single-line
edit controls.

On the other side, Zeos (as BDE did before as well as others) recognizes
that many servers have mechanisms for allowing fields with texts longer
than 255 characters and adds support for them; for example in Firebird
the type is BLOB SUB_TYPE TEXT and in MySQL it's TEXT. It assumes that
PostgreSQL's TEXT type falls into that category and associates it with a
"large text" type rather than "short text". The support for them is
quite limited, as they are a kind of "largest common divisor" of the
capabilities of the supported servers.

I think, but I'm not sure, that even the ODBC standard makes such
distinction between short and large text fields.

BTW, of course Tom is right in that most operators return TEXT and not
VARCHAR; I just stumbled upon this problem in a query that used MAX().

-- Pedro Gimeno

Pedro Gimeno escribió:

> Zeos (as well as the Borland Database Engine, which it just mimics in
> this sense) assumes that VARCHAR fields in general (not just
> PostgreSQL's) are textual fields with a limit of 255 characters for all
> supported servers, which are graphically associated with single-line
> edit controls.

Oh, so if you have a column of type VARCHAR(1000) or anything longer
than 255 chars it fails too?

And I'd guess that if you have a varchar(255) column and store a value
longer than 255 bytes (because there are many wide chars) it will fail
as well.

Both cases should really be tested and taken to Zeos if they fail.

--
Alvaro Herrera                                http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Alvaro Herrera wrote:

> Pedro Gimeno escribió:
>
>> Zeos (as well as the Borland Database Engine, which it just mimics in
>> this sense) assumes that VARCHAR fields in general (not just
>> PostgreSQL's) are textual fields with a limit of 255 characters for all
>> supported servers, which are graphically associated with single-line
>> edit controls.
>
> Oh, so if you have a column of type VARCHAR(1000) or anything longer
> than 255 chars it fails too?
>
> And I'd guess that if you have a varchar(255) column and store a value
> longer than 255 bytes (because there are many wide chars) it will fail
> as well.
>
> Both cases should really be tested and taken to Zeos if they fail.

I haven't tried. The 255 character limit may be a wrong assumption from
my side and not an actual Zeos limitation, but still VARCHAR fields are
treated as single-line edits and TEXT fields are treated as blob-like,
multiline texts. Anyway, in general Delphi does not do very well with
multibyte character sets, which by the way I'm not using. If it's
relevant for this bug report I can find out if Zeos has trouble with
>255 character varchar fields. Otherwise you may want to make the test
yourself and report to http://zeosbugs.firmos.at/ which is Zeos' bug
tracker. I don't use >255 VARCHAR fields for compatibility with other
servers.

-- Pedro Gimeno