Full Unicode/Ansi-Support in /testing branch

[EgonHugeist]

miab3,

woooops. Sorry Patch done Rev.1483

Thank you Michal.

Michael

[EgonHugeist]

@miab3

Michal i did prepare the testsuites and project-files for XE2 64Bit. Now they do compile fine.

Rev. 1497

Michael

[miab3]

@EgonHugeist

Yes I've seen.

Zeos7eh1499 compiles and runs on D2006, DXE2 32/64, LazarusWin1.1.0/fpc 2.7.1 32/64.

Still considering the problem of nested and array types in Oracle and PostgreSQL, I mentioned earlier.
Here, interestingly discussed a similar problem with Access 2007 format:
http://www.da-soft.com/blogs/anydac-for ... ields.html

Michal

[EgonHugeist]

miab3,

Yes i know we do not support the NestedDataSets actually. All i can say: It is possible. Actually we do some backgroundwork. Ludob is closing the remaining MemoryLeaks which Z7 had . And we're fixing the Bugreports of Mantis to restart a cleared list of issues IF Zeos7eh becomes the offical trunk. (i hope)

Michael

Btw. Zeos7eh Rev.1515.....

[miab3]

@EgonHugeist

Zeos7eh1515 compiles and runs on D2006, DXE2 32/64, LazarusWin1.1.0/fpc 2.7.1 32/64.

But to not get AV for PostgreSQL array types in ZdbcPostgreSQLUtils.pas had to change from line 225

else if (TypeName = 'int2vector') or (TypeName = 'oidvector')
{or (TypeName = '_aclitem') M.A.} then
Result := stAsciiStream
else if (TypeName <> '') and (TypeName[1] = '_') then // ARRAY TYPES
Result := stUnknown// M.A. was stAsciiStream
else
Result := stUnknown;

I see that this is an old issue:
http://zeosbugs.firmos.at/view.php?id=148

Michal

[EgonHugeist]

@miab3,

Patch applyed. Rev1517

Hmpf Michal, wasn't able to reproduce the issue. Added a dbc test to the Bugreports without effect. Can you use the bug-template app from Andre and make me a small bugreport? This is going very quickly. Just look to you Zeos root folder..

Michael

[miab3]

@EgonHugeist

Michael,
I see that for now only turning off the aclitem should be enough:
line 228: else if (TypeName <> '') and (TypeName[1] = '_') and not (TypeName = '_aclitem') then // ARRAY TYPES M.A. no '_aclitem'

Again from line 225:

Code: Select all

  else if (TypeName = 'int2vector') or (TypeName = 'oidvector')
    {or (TypeName = '_aclitem') M.A.} then
    Result := stAsciiStream
  else if (TypeName <> '') and (TypeName[1] = '_') and not (TypeName = '_aclitem')  then // ARRAY TYPES M.A. no '_aclitem'
    Result := stAsciiStream
  else
    Result := stUnknown;

Zeos7eh1520 compiles and runs on D2006, DXE2 32/64, LazarusWin1.1.0/fpc 2.7.1 32/64.
(With my fix better without AV in PostgreSQL)


Michal

[miab3]

@EgonHugeist

Michael,
Even better:

Line 283:
1034: Result := stAsciiStream; {aclitem[]} // M.A. was //28: Result := stAsciiStream; {aclitem[]}

and and you can undo the lock '_aclitem'

From line 225:
else if (TypeName = 'int2vector') or (TypeName = 'oidvector')
then
Result := stAsciiStream
else if (TypeName <> '') and (TypeName[1] = '_') then // ARRAY TYPES
Result := stAsciiStream
else
Result := stUnknown;

Michal

[EgonHugeist]

miab3,

Michal patch applyed Rev1522 thnx. Did you check it? Can we read that field now? Can you eventually make a testcase for our testsuites ready to check the type support for future?

Michael

[miab3]

@EgonHugeist

So I checked, they can be read now. And their content is seen in dbmemo.
for Zeos7eh1524 on D2006, DXE2 32/64, LazarusWin1.1.0/fpc 2.7.1 32/64.

Michal

[EgonHugeist]

[b@]miab3[/b],

Trank you for your konfirmations that everything works

Michal, if you've got additional issues like the nested fieds, i propose you make a testcase for our testsuites. Each failing test is a reminder for us. Actually on eh are no fails and no exceptions left! ((((:

Michael

[miab3]

@EgonHugeist

I will think about the sample test for nested tables in Oracle.

But I have a different issue.
Why limiting FetchRow works well only for Firebird
worse for Oracle
very bad for MySQL, MSSQL and PostgreSQL
and for SQLite does not work at all.
(It's about the speed for small number of first downloaded records from big table.)

Michal

[EgonHugeist]

miab3,

I wish i could give you an answer right now. I think that issue is related to the ReadOnly issue, which you've already reported. I know there is a bugreport on mantis which points to MySQL. Had not the time to check that issue. But knowing that it works well with FB can help to solve that issue...

Really Michal, i had not the time to check this and such things are hard to debug. It might be possible that it can also be engiene related...

Michael

[miab3]

@EgonHugeist

Michael,
Why in Olacle for table

CREATE TABLE "Batch_test"
(
INT1 NUMBER(10,0),
STRING1 VARCHAR2(20),
BLOB1 BLOB
)

the query

select * from "Batch_test"
or
select INT1, STRING1, BLOB1 from "Batch_test"

is is tens of times slower than

select INT1, STRING1 from "Batch_test"

(for empty BLOB's)

Michal

[EgonHugeist]

miab3,

A good question again. It might be possible that the collumn isNull byte isn't set corectly. Also is it possible that either we or the OCIis trying to load an empty lob permanently. Hmm is it possible to create a littlle testcase with a tickcount or something like this?

Michael