Network optimization V4.50

[Antonio Wiesner]

We have installed an application on a network (3-user pack) and the shadow directory has been created in a worksation using network optimize command. From what I see this shadow directory includes "shadow" copies of each table, while the "real" table reside in the server. On the other hand, complete library files -operations, forms, Scripts and so on- exist in the workstation and can be edited locally without changing the libraries in the server.

One apparent problem is that many saved operations -copy, summarize and maybe other- were specified to create tables in a default location, so when the operation is run from the workstation, the new table is saved where the database resides, that is the shadow directory. I can edit the saved operations in the workstation, setting the path of the table as "server _applicationdir_tablename", and the result table will be be saved in the server overwriting any previous file. This is Okay. But now the shadow database will include the new server table alongside the shadow copy. I guess I should Drop the server table from the shadow database and keep the shadow copy. Is this correct?

Another issue is that the paths of many scripts were originally written as C:_ApplicationDir_tablename, (for example) and now I don't know if the shadow database recognizes the path to the server files. From my initial trials it appears that the scripts work well, performing all actions on the server files. Is this correct?

I will appreciate all comments about these network issues. Thanks regards Antonio Wiesner

[JerryBrightbill]

RE: Network optimization V4.50

Antonio

I think you sould read a newsletter article about network optimization. There are many misconceptions about what actually is happening. The article was in issue 13

One of the biggest issues concerns path issues. If you have defined fixed paths in a database, the terminal will not be able to find the path. The best way is to either keep eveything in the server database directory, or a relative path to the server database directory. If a path is not specifed in an Alpha operation, the program always assumes the database directory.

For instance, you can create sub folders under the database directory. On the server, the result path would be a5.get_path()+ the sub directory path. On the terminal, this would give the same path back to the server - a5.get_master_path()+ sub directory path.

Many users define their own function or code to determine if a script is running on the server or a shadow.

if a5.get_master_path() = ""
data_path = a5.get_path()
else
data_path = a5.get_master_path()
end if

If you use something like this, data_path will always point back to the database directory on the server.

Jerry

[Antonio Wiesner]

RE: Network optimization V4.50

Jerry, thanks for your assistance.

Yes I read your article that explains very well what happens with network optimization. Our application works fine optimized, except for the saved operations that CREATE (or overwrite) tables: copy, summarize and crosstab. The problem is, the saved operation has a default path, inherited from the server database, and the new data will be stored at the terminal hard drive, which of course is not adequate.

I found that if I edit the saved operation IN THE TERMINAL I can set the path for the result table as Server_ ServApliPath_resultable.dbf, and the physical table in the server will be created /overwritten, (not surprisingly). But now the server table will be included in the terminal database, together with the shadow version. If I DROP the server table at the terminal, the shadow version will still be correct with the server table data, so this would solve the problem.

I wonder if dropping this resulting server table (unoptimized) from the terminal database is necessary at all. I fear that having two tables with the same name at the terminal, one from the server and one in the shadow directory, could create problems, so I would prefer to Drop the server table.
I hope I was able to explain these issues, and I will appreciate your valuable comments

Antonio

[JerryBrightbill]

RE: Network optimization V4.50

Antonio

Have you tried removing all path information from the operations? This should create the new file on the server in the server database directory, since the operation should inherit that path.

Jerry

[Antonio Wiesner]

RE: Network optimization V4.50

Thanks, Jerry. The operations indeed had no path, only the default which is "resultable.dbf" only. Operations in the terminal will specify exactly the same, the local database default directory, and the result table will be stored in the shadow directory. The only way to create /overwrite the server table is that the operation specifies a network path to the server, which is entirely possible, and then the only problem is that the terminal database shows two versions of the same table, the server and the shadow.

I have come to think that this duplication is really no problem, as the shadow table correctly addresses the hard server data. I am testing wether there is any confusion at the terminal between the "server" and "shadow" concerning Forms, other operations and so on.

On the other hand, all scripts and the operations that do not create tables work fine. It is a pleasant surprise that a script line that specifies a hard drive path in the server works smoothly from the terminal, thanks to the Network Optimize development.
I will appreciate further comments. Regards
Antonio

[JerryBrightbill]

RE: Network optimization V4.50

Antonio

If you have the shadow up to date, all forms, etc on the terminal are exactly the same as on the server. Since the shadow uses the local tables that are only pointers to the server, everything should work just fine.

If you really want to write a table back to the server with an operation, you can convert saved operations to low level xbasic and then specify the path. I rarely use saved operations, but instead use xbasic since I can use variables in the code to specify filters, relative paths, etc.

Jerry

[Antonio Wiesner]

RE: Network optimization V4.50

Thanks again. It is a good idea to write XBasic operations and I will try it. Still I fear that the result table in the server will show up in the terminal database, let me try and I will let you know. This very useful chat we have had confirms my belief that the Alpha community is a wonderful user environment.
Antonio

[JerryBrightbill]

RE: Network optimization V4.50

Antonio

Here is a sample of code that will write a table back to the server

' These lines find the path for the data

If a5.master_path() = ""
Data_path = a5.get_path()
Else
Data_path = a5.get_master_path()
End if

' these are the first 2 lines of actual subtract xbasic code
' This code will create an output file "hist_orphan.dbf" on the server

subtract.t_db = "sale_det.dbf"
subtract.o_file = data_path+"\hist_orphan.dbf"

In many cases, tables created by joins, subtract, or copies are used just for reports or to create temporary tables that can then append data back to an existing table. It is a rare operation that has to create an actual table on the server. Usually the table already exists. You can always zap the existing table and then append the data from the temp file to the existing shadow table. That will send the data back to the "real" table on the server. In those cases, the output table is frequently deleted when the process is complete.

Jerry

[Antonio Wiesner]

RE: Network optimization V4.50

Jerry, all your suggestions have worked really fine. I now have the data_path variable which I use in Xbasic operations both in the server and the terminal. Sometimes the operations from the terminal -copy, summarize- execute very slowly, and maybe the next time they run faster. Why would that be? In my code I have used the server addresses for all table paths such as in a summarize:

c_tbl = table.open(data_path+"_procompo.dbf")
sum.db = data_path+"_sulicompo.dbf"
sum.db_dd_delete = .F.
sum.order = "COMPONEN"
sum.filter = ""
sum.options = "I"
sum.fields = 3
sum.field1 = "Componen"
sum.code1 = 0
sum.field2 = "Tipo"
sum.code2 = 6
sum.field3 = "Campo1"
sum.code3 = 6
c_tbl.summarize()
c_tbl.close()

Should I use a default "procompo.dbf" file name in the first line, or keep the full path as above?

Another question: it appears that Xbasic crosstab operations form the terminal have some kind of problem, as they seem to stop at 98-99 percent records and restart, then run to the end with good results. Have you seen this type of problem?

Last question. "Refresh shadow" from the control panel at the terminal doesn't work and hangs Alpha Five.

Thaks in advance for your valuable comments

[JerryBrightbill]

RE: Network optimization V4.50

Antonio

Using table.open, you shouldn't need any path. The shadowed table will find the data on the server. Accessing the table directly from the server could slow down the process. The difference in speed could be to file locking issues if users have any of the tables open during the process.

I don't use crosstab, so I can't answer your question there.

Refresh shadow in versions 4.5 and older is somewhat unreliable. That was one of the many improvements in version 5. There have been a number of threads about how to work around the problem. I gave up on the built-in refresh and created my own. I think Cal Locklin posted some code in the archive that should help.

Jerry