Обсуждение: modifying the table function

"Islam Hegazy" <islheg@gmail.com> writes:
> My question is how to inform the client that there is a tuple to display =
> and return back to the backend to continue the query execution?

I'd suggest you start by reading
http://developer.postgresql.org/pgdocs/postgres/protocol.html
and then develop a clear specification at that level of what you
think should happen.  Perhaps after that exercise it will be clearer
how to change the code.  "Think first, program later."
        regards, tom lane

Thanks for the documentation link. It helped me to understand how data are 
passed back to the client.

I figured out that data is sent back to the client using the 'printtup' 
function. It is called by ExecSelect, called by ExecutorRun, etc. What I 
understand now is that the data is sent to the client and stored there until 
the client receives a message from the server to display it. The server 
sends the display message from within 'ReadyForQuery' but I can't figure it 
out.

What I expect is that when I call 'exec_simple_query' several times, the 
server sends one row only for each call. But since they are separate calls, 
the client overwrites the previous results or saves the new row in a new 
place in memory such that it displays the last row only when the server 
invokes 'ReadyForQuery'.

I wonder if I am on the right track or not and how to know such kind of 
message sent from the server?

Regards
Islam Hegazy

----- Original Message ----- 
From: "Tom Lane" <tgl@sss.pgh.pa.us>
To: "Islam Hegazy" <islheg@gmail.com>
Cc: <pgsql-hackers@postgresql.org>
Sent: Tuesday, April 17, 2007 1:44 AM
Subject: Re: [HACKERS] modifying the table function


> "Islam Hegazy" <islheg@gmail.com> writes:
>> My question is how to inform the client that there is a tuple to display 
>> =
>> and return back to the backend to continue the query execution?
>
> I'd suggest you start by reading
> http://developer.postgresql.org/pgdocs/postgres/protocol.html
> and then develop a clear specification at that level of what you
> think should happen.  Perhaps after that exercise it will be clearer
> how to change the code.  "Think first, program later."
>
> regards, tom lane 

"Islam Hegazy" <islheg@gmail.com> writes:
> I wonder if I am on the right track or not and how to know such kind of 
> message sent from the server?

Seems like you're doing it the hard way.  Wouldn't it be easier to fix
the client to display data before it's received the whole query result?
        regards, tom lane

Hi again

It seems now that I am one step away from the end. So far I have succeeded 
in returing row by row from the backend to the frontend, knew this from 
debugging.

Now comes the point of displaying them directly not to wait till the end of 
the query. These are the steps I took:

1) redefined 'PrintQueryResults' in common.c to be extern (not static as the 
initial definition) to be able to use it elsewhere
2) added a declaration for ''PrintQueryResults' in common.h, to tell other 
files about it
3) removed 'PrintQueryResults' invocation from 'SendQuery' common.c
4) added #include "../bin/psql/common.h" to fe-exec.c
5) called 'PrintQueryResults' from within 'PQexecFinish', last statment in 
the while loop

when I gmake the project I receive the following error:
"
../../../src/interfaces/libpq/libpq.so: undefined reference to 
`PrintQueryTuples'
collect2: ld returned 1 exit status
gmake[3]: *** [initdb] Error 1
gmake[3]: Leaving directory 
`/home/grads/imehegaz/postgresql-8.2.3-b/src/bin/initdb'
"

I wonder what does this error mean and how to solve it?

Regards
Islam Hegazy


----- Original Message ----- 
From: "Tom Lane" <tgl@sss.pgh.pa.us>
To: "Islam Hegazy" <islheg@gmail.com>
Cc: <pgsql-hackers@postgresql.org>
Sent: Wednesday, April 18, 2007 6:38 PM
Subject: Re: [HACKERS] modifying the table function


> "Islam Hegazy" <islheg@gmail.com> writes:
>> I wonder if I am on the right track or not and how to know such kind of
>> message sent from the server?
>
> Seems like you're doing it the hard way.  Wouldn't it be easier to fix
> the client to display data before it's received the whole query result?
>
> regards, tom lane