Re: what are correct steps to programmatically write/read large objects to/from a data base table?

Ian Bell wrote:
> I have just started working with PostgreSQL (v10.3) running under Window 10.
> Also I am new to SQL in general. I am writing test code to see how to write/read
> large objects and have a question about the order/number of steps when
> executing code written in C/C++ and C#.
>
> I want to write and read binary data as large objects.
>
> 1) In PSQL, I do this using the following commands:
>
> create database mydb;
> \c mydb;
> create table test( name text, mylargevariable oid);
> insert into test( name, mylargevariable) values( ‘n1’, lo_import(‘f:/temp/test.txt’) );
> select convert_from( lo_get(test.large), ‘UTF8’) from test;
>
> The point of this example is that the large object id value (‘OID’) is stored in the
> ‘mylargevariable’ field of the table ‘test’. My question is about how to programmatically
> do this using code written in C/C++ and C#. I address this in my next point.
>
> 2) The PostgreSQL sample code illustrates how to write/read a large object in C/C++
> while the NPGSQL sample code illustrates how to write/read large objects in C#.
> I am able to successfully compile and run both sets of sample code.
> Here are my questions:
>
> a. My understanding is these examples write the binary data as a large object to the
> ‘pg_largeobject’ table. Would you please confirm this is correct?
>
> b. If I want to be able to retrieve the binary data associated with a large object stored
> in the ‘pg_largeobject’ table at a later time (i.e. start a new process or run a
> new application) then I will have to persist/save/write the OID’s referencing the
> large object to a table when the large object is first created. So the pseudo code
> for both C/C++ and C# programs would be as follow:
>
> i. Create my database and create my own table which will store large object OIDs
> ii. Create a large object OID (e.g. call the function ‘lo_creat’ in the C libpg library)
> iii. Write the binary data to this large object (e.g. call the function ‘lo_write’ in the
> C libpg library)
> iv. Write the large object OID generated in step ii to a PostgreSQL table that I created in step i
>
> Would you please confirm these steps and their order are correct? The reason for asking
> this question is I want to confirm that i) while non-large objects require only one write
> to the data base ii) in contrast, large-objects require two writes (i.e. the first to write
> the large object, the second to insert/write the large object OID to a table field).
> I actually have written test code that does just this and it seems to be working.
> However, given my lack of experience with PostgreSQL and SQL in general, I would like
> confirmation the writing/reading binary data large object is this simple and that I have not
> missed something important.

That is all correct, and I think you have understood large objects well.

Have a look at the "lo" contrib: https://www.postgresql.org/docs/current/static/lo.html
Some of its functionality might help you to manage the large objects.

I'd like to remark that adding large "bytea" values to a table also affects
two tables: The table itself and the "TOAST table" where the bytea will
be stored out of line. But of course it is only one client-server
round trip.

Yours,
Laurenz Albe
--
Cybertec | https://www.cybertec-postgresql.com