Page 13 of 35

Re: 7.2-alfa testers-thread

Posted: 10.01.2014, 19:53
by miab3
@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R3062:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64 on Windows7-64(CodeTyphon ver 4.60 with i386_cgcpu.pas.patch: http://bugs.freepascal.org/view.php?id=25269) (W7-64)
CodeTyphon ver 4.60 (with FastCode) rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.34, MariaDB 5.5.32, PostgreSQL 9.3.0.1, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Michael,
100 times faster!!!
After setting:
PRAGMA synchronous = 0
PRAGMA locking_mode=EXCLUSIVE
My update Test runs 100 times faster (regardless of the type of the disk).
SQLite is now the fastest :D

Michal

Re: 7.2-alfa testers-thread

Posted: 12.01.2014, 00:02
by miab3
@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R3064:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64 on Windows7-64(CodeTyphon ver 4.60 with i386_cgcpu.pas.patch: http://bugs.freepascal.org/view.php?id=25269) (W7-64)
CodeTyphon ver 4.60 (with FastCode) rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.34, MariaDB 5.5.32, PostgreSQL 9.3.0.1, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Approximate(different location of servers) order in my Test on update(Time estimated):
1.SQLite(10sec), 2.Firebird(12sec), 3.MariaDB(15sec), 4.MSSQL(FreeTDS)(18sec), 5.ORACLE(40sec), 6.MSSQL(ADO)(50sec), 7.PostgreSQL(100sec).
PostgreSQL is known to slow update.


Michal

Re: 7.2-alfa testers-thread

Posted: 12.01.2014, 16:10
by miab3
@EgonHugeist,

Michael,
I noticed that my quick update Test2 SQLite is 2 to 2.5 times slower on the DXE2 compared to D2006.

Code: Select all

i7 
  D2006 - 1.65 sec.  
  DXE2 - 4.1 sec.
C2D 
  D2006 - 2.69 sec. 
  DXE2 - 5.81 sec.
Is that the overhead of Unicode version of Delphi!?
Do you gave would do something about it?

Michal

Re: 7.2-alfa testers-thread

Posted: 14.01.2014, 13:15
by miab3
@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R3066:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64 on Windows7-64(CodeTyphon ver 4.60 with i386_cgcpu.pas.patch: http://bugs.freepascal.org/view.php?id=25269) (W7-64)
CodeTyphon ver 4.60 (with FastCode) rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.35, MariaDB 5.5.32, PostgreSQL 9.3.0.1, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Michal

Re: 7.2-alfa testers-thread

Posted: 15.01.2014, 09:57
by miab3
@mdaems,@EgonHugeist,

ZEOS 7.1 (7.1.3)
Zeos7.1-patched_R3068:
http://svn.code.sf.net/p/zeoslib/code-0 ... 1-patches/
compiles and runs on (I have tested):

- D2007 -> C++, (W8.1-64)
- DXE2 32/64 -> C++(32), (W8.1-64)
- Lazarus Windows7-64 /fpc 2.7.1 32/64 (CodeTyphon ver 4.60 with i386_cgcpu.pas.patch: http://bugs.freepascal.org/view.php?id=25269) (W7-64)
CodeTyphon ver 4.60 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.35, MariaDB 5.5.32, PostgreSQL 9.3.0.1, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Michal

Re: 7.2-alfa testers-thread

Posted: 15.01.2014, 23:10
by miab3
@mdaems,@EgonHugeist,

ZEOS 7.1 (7.1.3)
Zeos7.1-patched_R3068:
http://svn.code.sf.net/p/zeoslib/code-0 ... 1-patches/
compiles and runs on Lazarus 32/64(fpc 2.7.1) on Windows7-64(CodeTyphon ver 4.70)
CodeTyphon ver 4.70 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.
It also works with Sternas ADO mod for Windows.

ZEOS 7.2
Zeos7.2-alpha branches_ testing_R3066:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on Lazarus 32/64(fpc 2.7.1) on Windows7-64(CodeTyphon ver 4.70)
CodeTyphon ver 4.70 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Michal

Re: 7.2-alfa testers-thread

Posted: 17.01.2014, 10:49
by miab3
@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R3070:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64(fpc 2.7.1) on Windows7-64(CodeTyphon ver 4.70)
CodeTyphon ver 4.70 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.35, MariaDB 5.5.34, PostgreSQL 9.3.0.1, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Michal

Re: 7.2-alfa testers-thread

Posted: 20.01.2014, 12:08
by miab3
@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R3072:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64(fpc 2.7.1) on Windows7-64(CodeTyphon ver 4.70)
CodeTyphon ver 4.70 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.35, MariaDB 5.5.34, PostgreSQL 9.3.0, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Michal

Re: 7.2-alfa testers-thread

Posted: 24.01.2014, 11:59
by miab3
@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R3074:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- D2007 -> C++???, (W8.1-64)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64(fpc 2.7.1) on Windows7-64(CodeTyphon ver 4.70)
CodeTyphon ver 4.70 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

When using the C++ 2007 error pops up(In C++ XE2 does not have this error):

Code: Select all

		struct 
		{
			System::TDateTime VDateTime;
			
		};

[BCC32 Error] Zvariant.hpp(81): E2019 'TZVariant:: :: :: ()' cannot be declared in an anonymous union
Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.35, MariaDB 5.5.34, PostgreSQL 9.3.0, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.


Add:
Michael,
After fix R3075 the mistake disappeared in C++2007.


Michal

Re: 7.2-alfa testers-thread

Posted: 25.01.2014, 11:44
by miab3
@EgonHugeist, @mdaems,

Zeos7.2-alpha branches_ testing_R3076:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- D2007 -> C++, (W8.1-64)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64(fpc 2.7.1) on Windows7-64(CodeTyphon ver 4.70)
CodeTyphon ver 4.70 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.35, MariaDB 5.5.34, PostgreSQL 9.3.0, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

Michal

Re: 7.2-alfa testers-thread

Posted: 25.01.2014, 21:40
by miab3
@EgonHugeist, @mdaems,

Michael,
I've been thinking that maybe those of your TFields (and maybe DataSets) should be harnessed to mapping and maybe even verification and conversion fields data.

Michal

Re: 7.2-alfa testers-thread

Posted: 26.01.2014, 17:13
by miab3
@mdaems, @EgonHugeist,

Zeos7.2-alpha branches_ testing_R3077:
http://svn.code.sf.net/p/zeoslib/code-0 ... sting-7.2/
compiles and runs on (I have tested):

- D2006, (W7-32)
- D2007 -> C++, (W8.1-64)
- DXE2 32/64 -> C++(32), (W7-64)
- Lazarus 32/64(fpc 2.7.1) on Windows7-64(CodeTyphon ver 4.70)
CodeTyphon ver 4.70 rebuilt without a problem. Current LazSqlX 4.0.3.1286-Win-32/64-bit build with success.

Test databases and 32/64-bits clients:
Firebird 2.5.3, MySQL 5.5.35, MariaDB 5.5.34, PostgreSQL 9.3.0, Oracle XE 11.2g, MSSQL 2008 R2 Express(mssql, FreeTDS, ado-sqlncli), *.mdb-ado-Jet.OLEDB.4.0, SQLite 3.7.17.

[Pascal Hint] ZDbcGenericResolver.pas(828): H2164 Variable 'lValidateUpdateCount' is declared but never used in 'TZGenericCachedResolver.PostUpdates'

Michal

Re: 7.2-alfa testers-thread

Posted: 27.01.2014, 21:25
by miab3
@EgonHugeist,

Michael,
In ZEOS 7.2 also something wrong with the space in aliases (FB).

Code: Select all

select ID, NAM as "NAM 1" from TAB1 
can not be modified (in dataset)

Code: Select all

select ID, NAM as "NAM_1" from TAB1 
can be modified.

Similar to the:
http://zeoslib.sourceforge.net/viewtopi ... 38&t=10781

Michal

Re: 7.2-alfa testers-thread

Posted: 28.01.2014, 23:03
by EgonHugeist
@miab3
Michael,
I've been thinking that maybe those of your TFields (and maybe DataSets) should be harnessed to mapping and maybe even verification and conversion fields data.
Yes this will be done. BUT actually i'm working out a DML Array binding way for Oracle. Main propblem isn't the API, the more it exports loads of code and we're discussiong about best practice. On my local working copy i can success fully updete/delete/insert an huge array of values with one network rundtrip inbetween. To be honest, the current implementation is for oracle only and leads to pain for all current other plains. The cool trip: It's really lighthing fast.

Also do we discuss about advantages/disadvantages to use such an implementation. As you know Synopse uses it. But it doesn't make realy sence if we can't provide the ZDBC api to the TDataSet-API. So i'm planning to make such DML execution possible for TZDataSet.CachedUpdates = True.

To be clear: I'm trying to optimize the current ZDBC-layer(and below) as best as i can and without breaking the current architecture. Than TComponent-Layer comes next. Logical, since TZDataSet accesses ZDBC, right?

My actual code adds more than 2000 lines again, where i'm not really happy about. Always think about maintaining such code... And it should be suitable for all other protocols too. Actually i see only BULK insertiation for DBLib(which doesn't support prepared statements btw) or (untested but possible) ADO+Vararrays can push the performance too.
select ID, NAM as "NAM_1" from TAB1
i've noticed this thread. But i don't think it's related to the Alias. The more it's a common issue with the Field names since identifier-quoting is missing for these only. See: http://zeosbugs.firmos.at/view.php?id=189
So i'ts a known bug AFAKS. Had not the time to build a testcase with our suites, yet. If i'm ready with OCI DML than i'll check this issue too. The user should just avoid spaces in the field and it's alias names . An simple "_" would avoid it. I think our tokenizer splits the query wrong and returns two ttWord tokens instead of one Identifier... (very deeply in ZCore... Btw.)

Re: 7.2-alfa testers-thread

Posted: 29.01.2014, 01:32
by miab3
@EgonHugeist

Unless something is wrong with the aliases, because such thing to me works:

Code: Select all

select LP, "NAZWA 1", "NAZWA 2", NAZWA  from A0

UPDATE A0 SET
  LP = :LP,
  "NAZWA 1" = :"NAZWA 1",
  "NAZWA 2" = :"NAZWA 2",
  NAZWA = :NAZWA
WHERE
  A0.LP = :OLD_LP
(can be edit and update.)

And for this Query:

Code: Select all

select LP, NAZWA as "NAZWA 3"  from A0
field "NAZWA 3" can not be edited.

Michal