Hi Giulio,
I'm curious about how you manage your PL/SQL development and identify
packages and its versions and how do you manage concurrency. Until were I
know Oracle won't manage concurrency when two developers edit the same
package and once broke, the entire package stop working.
I used, in a past project, to ask the team to save packages as text files in
the repository and lock them to signalize someone is working on it. The live
package being edited should be edited with a different name like
"package_name_plus_developers_user_name" so that the package won't be broken
for other developers. Once the developer finished to edit the package and
make his own tests he would increase its version number (a simple comment in
the package interface and body), write a small changelog as a comment,
submit it to the repository, release de lock, update package with its real
name and delete the temporary one.
Sounds a little troublesome, but I couldn't think of a better process. Later
I included a version function inside each package so I could check packages
versions using an sql script.
"It seems they do not want to tag everything in trunk because that would be
like a major release (apparently it would include those table and data
things). Maybe we could re-organised the code to separate the packages from
the data and then we could tag the packages, which is more what they want."
Yes, you can have a BTT (branch, tags, trunk) structure for the database DDL
including packages, another BTT for documentation and another for source
code and manage their evolution separately if it fits better to your team.
Instead of one big configuration item now you have 3 to care about. The
database definition versions can have patches to update a live database from
version a to version b.
But anyway you will need to build your high level configuration item, which
will represent your entire software package including database, code,
documentation etc in its specific compatible version. This set, you can
release software.
Your installer can check each configuration item version to decide what to
do (database version updates, documentation version). There are things that
can be just deleted and overwritten (like binaries and help), but database
will usually need to be patched. Having the database and plsql code as an
independent configuration item sounds good to me.
I'm not sure I'm helping... lol
Luiz Guilherme M. Kimel
-----Mensagem original-----
De: Giulio Troccoli [mailto:Giulio.Troccoli_at_uk.linedata.com]
Enviada em: sexta-feira, 29 de outubro de 2010 10:11
Para: users_at_subversion.apache.org
Assunto: Moving to Subversion for PL-SQL development
First of all let me tell you that I don't know much of how PL-SQL
development works so I might say something really obvious to you or more
likely just wrong. Please forgive me.
I have a team that uses StarTeam as their VCS and we are now working on
moving the project to Subversion. We are planning to use an importer for the
initial load of the repository which seems to do what they want (I'm not
looking after that part).
I have a problem though with their releasing process.
As I understand it, a major release is formed by all the packages and
scripts, plus some table initialisation and sometime some data (I presume
for defaults and stuff like that). Minor releases are done with patches
which included only the packages that have changed from the previous patch.
So, if I want 5.4.0 (major release), I get everything. I unpack the kit,
install it, run it, whatever it take and I'm done. If I am already on 5.4.0
and I want 5.4.3 (a minor release) I will be sent 3 patches: to 5.4.1, then
5.4.2 and finally 5.4.3. Apparently I just need to unzip them and I'm done.
Now, I might not be clear in the above process, so if someone with more
experience with PL-SQL development and release wants to correct me, please
do. I know there isn't one way to do things, but it's more likely that I
understood wrong than we are doing it in a special way.
Anyway, if I am right, I'm struggling to come up with a process using
Subversion. It seems they do not want to tag everything in trunk because
that would be like a major release (apparently it would include those table
and data things). Maybe we could re-organised the code to separate the
packages from the data and then we could tag the packages, which is more
what they want. And this way, to go to 5.4.3 I won't need 5.4.1 and 5.4.2 at
all, which in my opinion is even better.
In the end what I am looking for with this email is some advice on how to
proceed from people with more experience than me in projects using PL-SQL.
Thanks
Giulio Troccoli
Linedata Limited
Registered Office: 85 Gracechurch St., London, EC3V 0AA
Registered in England and Wales No 3475006 VAT Reg No 710 3140 03
Received on 2010-10-29 16:13:18 CEST